Compare commits

..

5 Commits

Author SHA1 Message Date
Gaurav Tewari
55d5ddb9ba feat: ai trace view cloumn 2026-08-04 16:14:18 +05:30
Gaurav Tewari
fead012459 chore: initial commit 2026-08-04 16:08:42 +05:30
Cursor Agent
5f50dcd349 poc(ai-o11y): Trace View column registry + client-side visibility
Hardcode a Trace View column whitelist (AI O11y TDD shape), own visibility
in a separate localStorage key, and wire Options → Edit columns via
FieldsSelector.availableFields so toggling is instant with no selectFields
refetch. Demoed on existing Trace View.

Co-authored-by: Gaurav Tewari <gauravtewari111@gmail.com>
2026-08-04 09:42:31 +00:00
Srikanth Chekuri
957580ec7c fix(alertmanager): resolve email SMTP settings from env (#12281)
* fix(alertmanager): resolve email SMTP settings from env

* refactor(alertmanager): remove dead legacy config and tidy integration test helpers

* test(alertmanager): scope smtp rotation test to env rotation only

* chore: drop redundant test

* chore: log and warn; do not persist global config
2026-08-04 08:53:34 +00:00
Tushar Vats
77c1b601be feat(logs): implement search() across all log fields (#12244)
Some checks failed
build-staging / prepare (push) Has been cancelled
build-staging / js-build (push) Has been cancelled
build-staging / go-build (push) Has been cancelled
build-staging / staging (push) Has been cancelled
cacheci / tests (push) Has been cancelled
Release Drafter / update_release_draft (push) Has been cancelled
- Replace the search() stub: the logs condition builder fans a case-insensitive
  match of the term across every searchable column (log columns, body/body_v2,
  attribute + resource maps).
- Grammar: searchCall takes a valueList (parser regenerated), so the scoped form
  search('term', body, resource) parses and narrows the fan-out to those contexts.
- The visitor emits FilterOperatorSearch and flags the statement scan-heavy; the
  statement builder attaches a CostGuard the querier enforces via EXPLAIN ESTIMATE
  against a per-shard budget, cumulative across buckets on the window-list path.
- body_v2 gets its own, lower budget (search_max_scan_rows_json_body): toString()
  rebuilds every document and no skip index prunes.
- Logs-only: other signals reject search().
- Unit tests for the per-context fan-out SQL and both budgets; integration suites
  run the same matrix over the legacy body and over body_v2.
2026-08-03 19:03:38 +00:00
80 changed files with 4212 additions and 426 deletions

View File

@@ -39,6 +39,8 @@ jobs:
matrix:
suite:
- alerts
- alertmanager
- alertmanagerrotation
- basepath
- callbackauthn
- cloudintegrations

View File

@@ -15477,72 +15477,6 @@ paths:
summary: Lock dashboard (v2)
tags:
- dashboard
/api/v2/dashboards/{id}/migrate:
post:
deprecated: false
description: 'This endpoint retries the v1→v2 (Perses) migration on a dashboard
still stored in the v1 schema and returns the v2-shape result. It is idempotent:
a dashboard already in the v2 schema is returned unchanged.'
operationId: MigrateDashboardV2
parameters:
- in: path
name: id
required: true
schema:
type: string
responses:
"200":
content:
application/json:
schema:
properties:
data:
$ref: '#/components/schemas/DashboardtypesGettableDashboardV2'
status:
type: string
required:
- status
- data
type: object
description: OK
"400":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Bad Request
"401":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Unauthorized
"403":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Forbidden
"404":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Not Found
"500":
content:
application/json:
schema:
$ref: '#/components/schemas/RenderErrorResponse'
description: Internal Server Error
security:
- api_key:
- EDITOR
- tokenizer:
- EDITOR
summary: Migrate dashboard to v2
tags:
- dashboard
/api/v2/factor_password/forgot:
post:
deprecated: false

View File

@@ -276,10 +276,6 @@ func (module *module) GetV2(ctx context.Context, orgID valuer.UUID, id valuer.UU
return module.pkgDashboardModule.GetV2(ctx, orgID, id)
}
func (module *module) MigrateV2(ctx context.Context, orgID valuer.UUID, id valuer.UUID) (*dashboardtypes.DashboardV2, error) {
return module.pkgDashboardModule.MigrateV2(ctx, orgID, id)
}
func (module *module) UpdateV2(ctx context.Context, orgID valuer.UUID, id valuer.UUID, updatedBy string, updatable dashboardtypes.UpdatableDashboardV2) (*dashboardtypes.DashboardV2, error) {
return module.pkgDashboardModule.UpdateV2(ctx, orgID, id, updatedBy, updatable)
}

View File

@@ -52,8 +52,6 @@ import type {
ListDashboardsV2200,
ListDashboardsV2Params,
LockDashboardV2PathParameters,
MigrateDashboardV2200,
MigrateDashboardV2PathParameters,
PatchDashboardV2200,
PatchDashboardV2PathParameters,
PinDashboardV2PathParameters,
@@ -1806,85 +1804,6 @@ export const useLockDashboardV2 = <
> => {
return useMutation(getLockDashboardV2MutationOptions(options));
};
/**
* This endpoint retries the v1→v2 (Perses) migration on a dashboard still stored in the v1 schema and returns the v2-shape result. It is idempotent: a dashboard already in the v2 schema is returned unchanged.
* @summary Migrate dashboard to v2
*/
export const migrateDashboardV2 = (
{ id }: MigrateDashboardV2PathParameters,
signal?: AbortSignal,
) => {
return GeneratedAPIInstance<MigrateDashboardV2200>({
url: `/api/v2/dashboards/${id}/migrate`,
method: 'POST',
signal,
});
};
export const getMigrateDashboardV2MutationOptions = <
TError = ErrorType<RenderErrorResponseDTO>,
TContext = unknown,
>(options?: {
mutation?: UseMutationOptions<
Awaited<ReturnType<typeof migrateDashboardV2>>,
TError,
{ pathParams: MigrateDashboardV2PathParameters },
TContext
>;
}): UseMutationOptions<
Awaited<ReturnType<typeof migrateDashboardV2>>,
TError,
{ pathParams: MigrateDashboardV2PathParameters },
TContext
> => {
const mutationKey = ['migrateDashboardV2'];
const { mutation: mutationOptions } = options
? options.mutation &&
'mutationKey' in options.mutation &&
options.mutation.mutationKey
? options
: { ...options, mutation: { ...options.mutation, mutationKey } }
: { mutation: { mutationKey } };
const mutationFn: MutationFunction<
Awaited<ReturnType<typeof migrateDashboardV2>>,
{ pathParams: MigrateDashboardV2PathParameters }
> = (props) => {
const { pathParams } = props ?? {};
return migrateDashboardV2(pathParams);
};
return { mutationFn, ...mutationOptions };
};
export type MigrateDashboardV2MutationResult = NonNullable<
Awaited<ReturnType<typeof migrateDashboardV2>>
>;
export type MigrateDashboardV2MutationError = ErrorType<RenderErrorResponseDTO>;
/**
* @summary Migrate dashboard to v2
*/
export const useMigrateDashboardV2 = <
TError = ErrorType<RenderErrorResponseDTO>,
TContext = unknown,
>(options?: {
mutation?: UseMutationOptions<
Awaited<ReturnType<typeof migrateDashboardV2>>,
TError,
{ pathParams: MigrateDashboardV2PathParameters },
TContext
>;
}): UseMutationResult<
Awaited<ReturnType<typeof migrateDashboardV2>>,
TError,
{ pathParams: MigrateDashboardV2PathParameters },
TContext
> => {
return useMutation(getMigrateDashboardV2MutationOptions(options));
};
/**
* This endpoint returns the sanitized v2-shape dashboard data for public access. Each panel query is reduced to a safe field subset, so filters and raw query strings are not exposed.
* @summary Get public dashboard data (v2)

View File

@@ -11164,17 +11164,6 @@ export type UnlockDashboardV2PathParameters = {
export type LockDashboardV2PathParameters = {
id: string;
};
export type MigrateDashboardV2PathParameters = {
id: string;
};
export type MigrateDashboardV2200 = {
data: DashboardtypesGettableDashboardV2DTO;
/**
* @type string
*/
status: string;
};
export type GetFeatures200 = {
/**
* @type array

View File

@@ -31,6 +31,13 @@ interface FieldsSelectorProps {
width?: number;
height?: number;
defaultPosition?: { x: number; y: number };
/**
* Caller-supplied field list. When provided, Other Fields offers exactly
* these and key discovery is skipped — use it when the set of selectable
* fields is known up front rather than fetched. Omit to discover keys from
* the API for `signal`.
*/
availableFields?: TelemetryFieldKey[];
}
type FieldsSelectorContentProps = Omit<FieldsSelectorProps, 'isOpen'>;
@@ -49,6 +56,7 @@ function FieldsSelectorContent({
width = DEFAULT_PANEL_WIDTH,
height,
defaultPosition,
availableFields,
}: FieldsSelectorContentProps): JSX.Element {
const resolvedHeight =
height ?? window.innerHeight - DEFAULT_PANEL_HEIGHT_OFFSET;
@@ -153,6 +161,7 @@ function FieldsSelectorContent({
addedFields={draftFields}
onAdd={handleAdd}
isAtLimit={isAtLimit}
availableFields={availableFields}
/>
{hasUnsavedChanges && (

View File

@@ -21,6 +21,11 @@ interface OtherFieldsProps {
addedFields: TelemetryFieldKey[];
onAdd: (field: TelemetryFieldKey) => void;
isAtLimit: boolean;
/**
* Caller-supplied field list. When provided, key discovery is skipped and
* these are filtered locally by the search term instead.
*/
availableFields?: TelemetryFieldKey[];
}
function OtherFields({
@@ -29,7 +34,10 @@ function OtherFields({
addedFields,
onAdd,
isAtLimit,
availableFields,
}: OtherFieldsProps): JSX.Element {
const useRegistry = Boolean(availableFields);
const { data, isFetching } = useGetQueryKeySuggestions(
{
signal,
@@ -41,11 +49,34 @@ function OtherFields({
signal,
debouncedInputValue,
],
enabled: true,
enabled: !useRegistry,
},
);
const otherFields: TelemetryFieldKey[] = useMemo(() => {
const addedIds = new Set(
addedFields.map((f) => f.key ?? buildCompositeKey(f.name, f.fieldContext)),
);
if (useRegistry && availableFields) {
const search = debouncedInputValue.trim().toLowerCase();
return availableFields
.filter((attr) => {
const id = attr.key ?? buildCompositeKey(attr.name, attr.fieldContext);
if (addedIds.has(id)) {
return false;
}
if (!search) {
return true;
}
return attr.name.toLowerCase().includes(search);
})
.map((attr) => ({
...attr,
key: attr.key ?? buildCompositeKey(attr.name, attr.fieldContext),
}));
}
const suggestions = Object.values(data?.data.data.keys || {}).flat();
// Normalize: synthesize `key` once so downstream reads can trust it.
const normalizedSuggestions: TelemetryFieldKey[] = suggestions.map(
@@ -57,15 +88,12 @@ function OtherFields({
fieldDataType: attr.fieldDataType,
}),
);
const addedIds = new Set(
addedFields.map((f) => f.key ?? buildCompositeKey(f.name, f.fieldContext)),
);
return normalizedSuggestions.filter(
(attr) => !addedIds.has(attr.key as string),
);
}, [data, addedFields]);
}, [data, addedFields, availableFields, debouncedInputValue, useRegistry]);
if (isFetching) {
if (!useRegistry && isFetching) {
return (
<div className={cx(styles.section, styles.sectionOther)}>
<div className={styles.sectionHeader}>OTHER FIELDS</div>

View File

@@ -11,6 +11,8 @@ export enum LOCALSTORAGE {
TRACES_LIST_OPTIONS = 'TRACES_LIST_OPTIONS',
GRAPH_VISIBILITY_STATES = 'GRAPH_VISIBILITY_STATES',
TRACES_LIST_COLUMNS = 'TRACES_LIST_COLUMNS',
/** AI Observability Trace View column visibility. */
AI_TRACE_VIEW_COLUMNS = 'AI_TRACE_VIEW_COLUMNS',
LOGS_LIST_COLUMNS = 'LOGS_LIST_COLUMNS',
LOGS_LIST_COLUMN_SIZING = 'LOGS_LIST_COLUMN_SIZING',
LOGGED_IN_USER_NAME = 'LOGGED_IN_USER_NAME',

View File

@@ -0,0 +1,178 @@
import { Typography } from '@signozhq/ui/typography';
import { LOCALSTORAGE } from 'constants/localStorage';
import {
BASE_TRACE_VIEW_COLUMNS,
renderTraceDurationCell,
TraceViewColumn,
} from 'container/TracesExplorer/TracesView/configs';
import { TraceViewColumnSelection } from 'container/TracesExplorer/TracesView/useTraceViewColumns';
function isBlank(value: unknown): boolean {
return value === undefined || value === null || value === '';
}
function renderCountCell(value: unknown): JSX.Element {
if (isBlank(value)) {
return <Typography></Typography>;
}
const count = Number(value);
return (
<Typography>
{Number.isFinite(count) ? count.toLocaleString() : String(value)}
</Typography>
);
}
function renderCostCell(value: unknown): JSX.Element {
if (isBlank(value)) {
return <Typography></Typography>;
}
const cost = Number(value);
return (
<Typography>
{Number.isFinite(cost) ? `$${cost.toFixed(4)}` : String(value)}
</Typography>
);
}
/**
* Trace-level gen_ai aggregates. The query-range response does not carry these
* yet, so they render as em dashes until the AI trace API lands — visible but
* empty is intentional, it lets the column set be reviewed ahead of the data.
*/
const AI_ONLY_COLUMNS: TraceViewColumn[] = [
{
field: {
name: 'input_tokens',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Input Tokens',
render: renderCountCell,
},
{
field: {
name: 'output_tokens',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Output Tokens',
render: renderCountCell,
},
{
field: {
name: 'total_tokens',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Total Tokens',
render: renderCountCell,
},
{
field: {
name: 'llm_call_count',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'LLM Calls',
render: renderCountCell,
},
{
field: {
name: 'tool_call_count',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Tool Calls',
render: renderCountCell,
},
{
field: {
name: 'distinct_tool_count',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Distinct Tools',
render: renderCountCell,
},
{
field: {
name: 'estimated_cost_usd',
fieldContext: 'trace',
fieldDataType: 'float64',
},
title: 'Est. Cost (USD)',
render: renderCostCell,
},
{
field: {
name: 'max_llm_latency_ns',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Max LLM Latency',
render: renderTraceDurationCell,
},
{
field: {
name: 'last_activity_time',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Last Activity',
},
{
field: { name: 'start_time', fieldContext: 'trace', fieldDataType: 'int64' },
title: 'Start Time',
},
{
field: { name: 'end_time', fieldContext: 'trace', fieldDataType: 'int64' },
title: 'End Time',
},
{
field: {
name: 'trace_duration_nano',
fieldContext: 'trace',
fieldDataType: 'int64',
},
title: 'Trace Duration',
render: renderTraceDurationCell,
},
{
field: { name: 'error_count', fieldContext: 'trace', fieldDataType: 'int64' },
title: 'Errors',
render: renderCountCell,
},
{
field: {
name: 'root_span_name',
fieldContext: 'trace',
fieldDataType: 'string',
},
title: 'Root Span Name',
},
{
field: { name: 'input', fieldContext: 'trace', fieldDataType: 'string' },
title: 'Input',
},
{
field: { name: 'output', fieldContext: 'trace', fieldDataType: 'string' },
title: 'Output',
},
];
export const AI_TRACE_VIEW_COLUMNS: TraceViewColumn[] = [
...BASE_TRACE_VIEW_COLUMNS,
...AI_ONLY_COLUMNS,
];
/**
* Hand this to `<TracesView columnSelection={…} />` to get the AI column set
* plus the Options → Edit columns picker. Module-level so its identity is
* stable across renders.
*/
export const AI_TRACE_VIEW_COLUMN_SELECTION: TraceViewColumnSelection = {
columns: AI_TRACE_VIEW_COLUMNS,
storageKey: LOCALSTORAGE.AI_TRACE_VIEW_COLUMNS,
defaultVisible: BASE_TRACE_VIEW_COLUMNS.map((column) => column.field.name),
};

View File

@@ -11,4 +11,12 @@
align-items: center;
gap: 4px;
cursor: pointer;
// Rendered as a <button> for keyboard access — strip the native chrome so it
// still reads as the inline text trigger it was.
padding: 0;
border: none;
background: none;
color: inherit;
font: inherit;
}

View File

@@ -5,6 +5,7 @@ import FieldsSelector from 'components/FieldsSelector';
import Controls, { ControlsProps } from 'container/Controls';
import { OptionsMenuConfig } from 'container/OptionsMenu/types';
import useQueryPagination from 'hooks/queryPagination/useQueryPagination';
import { TelemetryFieldKey } from 'types/api/v5/queryRange';
import { DataSource } from 'types/common/queryBuilder';
import styles from './Controls.module.scss';
@@ -15,6 +16,7 @@ function TraceExplorerControls({
perPageOptions,
config,
showSizeChanger = true,
availableFields,
}: TraceExplorerControlsProps): JSX.Element | null {
const { t } = useTranslation(['trace']);
const [isFieldsSelectorOpen, setIsFieldsSelectorOpen] = useState(false);
@@ -30,13 +32,15 @@ function TraceExplorerControls({
<div className={styles.container}>
{config?.fieldsSelector && (
<>
<div
<button
type="button"
className={styles.optionsTrigger}
onClick={(): void => setIsFieldsSelectorOpen(true)}
data-testid="trace-view-options-trigger"
>
{t('options_menu.options')}
<Settings size="md" />
</div>
</button>
<FieldsSelector
isOpen={isFieldsSelectorOpen}
title="Edit columns"
@@ -44,6 +48,7 @@ function TraceExplorerControls({
onFieldsChange={config.fieldsSelector.onFieldsChange}
onClose={(): void => setIsFieldsSelectorOpen(false)}
signal={DataSource.TRACES}
availableFields={availableFields}
/>
</>
)}
@@ -63,20 +68,20 @@ function TraceExplorerControls({
);
}
TraceExplorerControls.defaultProps = {
config: null,
};
type TraceExplorerControlsProps = Pick<
ControlsProps,
'isLoading' | 'totalCount' | 'perPageOptions'
> & {
config?: OptionsMenuConfig | null;
showSizeChanger?: boolean;
/** Forwarded to FieldsSelector — see `availableFields` there. */
availableFields?: TelemetryFieldKey[];
};
TraceExplorerControls.defaultProps = {
config: null,
showSizeChanger: true,
availableFields: undefined,
};
export default memo(TraceExplorerControls);

View File

@@ -4,47 +4,105 @@ import { Typography } from '@signozhq/ui/typography';
import ROUTES from 'constants/routes';
import { getMs } from 'container/Trace/Filters/Panel/PanelBody/Duration/util';
import { DEFAULT_PER_PAGE_OPTIONS } from 'hooks/queryPagination';
import { TelemetryFieldKey } from 'types/api/v5/queryRange';
import { ListItem } from 'types/api/widgets/getQuery';
export const PER_PAGE_OPTIONS: number[] = [10, ...DEFAULT_PER_PAGE_OPTIONS];
export const columns: ColumnsType<ListItem['data']> = [
/**
* One Trace View column: the telemetry field it reads, its header, and how the
* cell renders. Callers own their column sets — Trace View has no knowledge of
* any particular product's columns.
*/
export interface TraceViewColumn {
field: TelemetryFieldKey;
title: string;
/** Defaults to `renderTraceCellValue`. */
render?: (value: unknown) => JSX.Element;
}
function isBlank(value: unknown): boolean {
return value === undefined || value === null || value === '';
}
/** Fallback cell: em dash when empty, otherwise stringified. */
export function renderTraceCellValue(value: unknown): JSX.Element {
if (isBlank(value)) {
return <Typography></Typography>;
}
return <Typography>{String(value)}</Typography>;
}
/** Nanosecond duration rendered as milliseconds. */
export function renderTraceDurationCell(value: unknown): JSX.Element {
if (isBlank(value)) {
return <Typography></Typography>;
}
return <Typography>{getMs(String(value))}ms</Typography>;
}
function renderTraceIdCell(value: unknown): JSX.Element {
if (isBlank(value)) {
return <Typography></Typography>;
}
return (
<Link
to={generatePath(ROUTES.TRACE_DETAIL, {
id: String(value),
})}
data-testid="trace-id"
>
{String(value)}
</Link>
);
}
/**
* The root-span columns Trace View renders when the caller configures no
* column selection. Matches the pre-selection behaviour exactly.
*/
export const BASE_TRACE_VIEW_COLUMNS: TraceViewColumn[] = [
{
field: {
name: 'service.name',
fieldContext: 'resource',
fieldDataType: 'string',
},
title: 'Root Service Name',
dataIndex: 'service.name',
key: 'serviceName',
},
{
field: { name: 'name', fieldContext: 'span', fieldDataType: 'string' },
title: 'Root Operation Name',
dataIndex: 'name',
key: 'name',
},
{
field: {
name: 'duration_nano',
fieldContext: 'span',
fieldDataType: 'int64',
},
title: 'Root Duration (in ms)',
dataIndex: 'duration_nano',
key: 'durationNano',
render: (duration: number): JSX.Element => (
<Typography>{getMs(String(duration))}ms</Typography>
),
render: renderTraceDurationCell,
},
{
field: { name: 'span_count', fieldContext: 'trace', fieldDataType: 'int64' },
title: 'No of Spans',
dataIndex: 'span_count',
key: 'span_count',
},
{
field: { name: 'trace_id', fieldContext: 'span', fieldDataType: 'string' },
title: 'TraceID',
dataIndex: 'trace_id',
key: 'traceID',
render: (traceID: string): JSX.Element => (
<Link
to={generatePath(ROUTES.TRACE_DETAIL, {
id: traceID,
})}
data-testid="trace-id"
>
{traceID}
</Link>
),
render: renderTraceIdCell,
},
];
/** Build antd columns, preserving the order given. */
export function buildTraceViewColumns(
columns: TraceViewColumn[],
): ColumnsType<ListItem['data']> {
return columns.map(({ field, title, render }) => ({
title,
dataIndex: field.name,
key: field.name,
render: (value: unknown): JSX.Element =>
(render ?? renderTraceCellValue)(value),
}));
}

View File

@@ -34,14 +34,26 @@ import DOCLINKS from 'utils/docLinks';
import TraceExplorerControls from '../Controls';
import { TracesLoading } from '../TraceLoading/TraceLoading';
import { columns, PER_PAGE_OPTIONS } from './configs';
import {
BASE_TRACE_VIEW_COLUMNS,
buildTraceViewColumns,
PER_PAGE_OPTIONS,
} from './configs';
import { ActionsContainer, Container } from './styles';
import useTraceViewColumns, {
TraceViewColumnSelection,
} from './useTraceViewColumns';
interface TracesViewProps {
isFilterApplied: boolean;
setWarning: Dispatch<SetStateAction<Warning | undefined>>;
setIsLoadingQueries: Dispatch<SetStateAction<boolean>>;
queryKeyRef?: MutableRefObject<any>;
/**
* Opt in to user-editable columns. Omit for the base root-span columns with
* no Options → Edit columns picker.
*/
columnSelection?: TraceViewColumnSelection;
}
function TracesView({
@@ -49,6 +61,7 @@ function TracesView({
setWarning,
setIsLoadingQueries,
queryKeyRef,
columnSelection,
}: TracesViewProps): JSX.Element {
const { stagedQuery, panelType } = useQueryBuilder();
@@ -62,6 +75,32 @@ function TracesView({
QueryParams.pagination,
);
// Column visibility is owned here rather than by useOptionsMenu, whose
// TRACES_LIST_OPTIONS storage is already claimed by List View.
const { visibleColumns, selectedFields, availableFields, onFieldsChange } =
useTraceViewColumns(columnSelection);
const fieldsSelectorConfig = useMemo(
() =>
columnSelection
? {
fieldsSelector: {
value: selectedFields,
onFieldsChange,
},
}
: null,
[columnSelection, selectedFields, onFieldsChange],
);
const tableColumns = useMemo(
() =>
buildTraceViewColumns(
columnSelection ? visibleColumns : BASE_TRACE_VIEW_COLUMNS,
),
[columnSelection, visibleColumns],
);
const transformedQuery = useMemo(
() => getListViewQuery(stagedQuery || initialQueriesMap.traces),
[stagedQuery],
@@ -76,6 +115,8 @@ function TracesView({
stagedQuery,
panelType,
paginationQueryData,
// Column visibility is deliberately absent: it is client-side only, and
// this array doubles as the parent's cancelQueries handle.
],
[
globalSelectedTime,
@@ -100,6 +141,8 @@ function TracesView({
params: {
dataSource: 'traces',
},
// No selectColumns: the backend returns all columns and visibility is
// resolved client-side, so toggling a column is refetch-free.
tableParams: {
pagination: paginationQueryData,
},
@@ -162,6 +205,8 @@ function TracesView({
isLoading={isLoading}
totalCount={responseData?.length || 0}
perPageOptions={PER_PAGE_OPTIONS}
config={fieldsSelectorConfig}
availableFields={columnSelection ? availableFields : undefined}
/>
</div>
</ActionsContainer>
@@ -190,7 +235,7 @@ function TracesView({
{(tableData || []).length !== 0 && (
<ResizeTable
loading={isLoading}
columns={columns}
columns={tableColumns}
tableLayout="fixed"
dataSource={tableData}
scroll={{ x: true }}
@@ -203,6 +248,7 @@ function TracesView({
TracesView.defaultProps = {
queryKeyRef: undefined,
columnSelection: undefined,
};
export default memo(TracesView);

View File

@@ -0,0 +1,123 @@
import { useCallback, useMemo, useState } from 'react';
import getLocalStorageKey from 'api/browser/localstorage/get';
import setLocalStorageKey from 'api/browser/localstorage/set';
import { LOCALSTORAGE } from 'constants/localStorage';
import { TelemetryFieldKey } from 'types/api/v5/queryRange';
import { TraceViewColumn } from './configs';
/**
* Opt-in column selection for Trace View. Passing one enables the Options →
* Edit columns picker; omitting it leaves Trace View on its base columns with
* no picker at all.
*
* Pass a module-level constant, not an inline literal — a fresh object each
* render invalidates the memoised column set on every pass.
*/
export interface TraceViewColumnSelection {
/** Every column offered, in default display order. */
columns: TraceViewColumn[];
/** Where this caller persists visibility. Must be unique per view. */
storageKey: LOCALSTORAGE;
/** Field names visible before the user customises anything. */
defaultVisible: string[];
}
interface UseTraceViewColumnsReturn {
/** Columns to render, in the user's persisted order. */
visibleColumns: TraceViewColumn[];
/** Visible fields, for the picker's "added" list. */
selectedFields: TelemetryFieldKey[];
/** Every offered field, for the picker's "other" list. */
availableFields: TelemetryFieldKey[];
onFieldsChange: (fields: TelemetryFieldKey[]) => void;
}
function readStoredKeys(
selection: TraceViewColumnSelection | undefined,
): string[] {
if (!selection) {
return [];
}
const { storageKey, columns, defaultVisible } = selection;
const raw = getLocalStorageKey(storageKey);
if (!raw) {
return defaultVisible;
}
try {
const parsed = JSON.parse(raw) as string[];
if (!Array.isArray(parsed) || parsed.length === 0) {
return defaultVisible;
}
// Drop anything the caller no longer offers, so a stale localStorage
// entry from an earlier column set self-heals instead of rendering blank.
const allowed = new Set(columns.map((column) => column.field.name));
const filtered = parsed.filter((key) => allowed.has(key));
return filtered.length > 0 ? filtered : defaultVisible;
} catch {
return defaultVisible;
}
}
/**
* Owns Trace View column visibility client-side: toggling reorders or hides
* columns without touching the query, so it never triggers a refetch.
*/
function useTraceViewColumns(
selection?: TraceViewColumnSelection,
): UseTraceViewColumnsReturn {
const [visibleKeys, setVisibleKeys] = useState<string[]>(() =>
readStoredKeys(selection),
);
const offeredColumns = selection?.columns;
const columnsByName = useMemo(
() =>
new Map((offeredColumns ?? []).map((column) => [column.field.name, column])),
[offeredColumns],
);
// Ordered by visibleKeys, so reordering in the picker moves the column.
const visibleColumns = useMemo(
() =>
visibleKeys
.map((key) => columnsByName.get(key))
.filter((column): column is TraceViewColumn => Boolean(column)),
[columnsByName, visibleKeys],
);
const selectedFields = useMemo(
() => visibleColumns.map((column) => column.field),
[visibleColumns],
);
const availableFields = useMemo(
() => (offeredColumns ?? []).map((column) => column.field),
[offeredColumns],
);
const onFieldsChange = useCallback(
(fields: TelemetryFieldKey[]): void => {
if (!selection) {
return;
}
const nextKeys = fields.map((field) => field.name);
const keys = nextKeys.length > 0 ? nextKeys : selection.defaultVisible;
setVisibleKeys(keys);
setLocalStorageKey(selection.storageKey, JSON.stringify(keys));
},
[selection],
);
return {
visibleColumns,
selectedFields,
availableFields,
onFieldsChange,
};
}
export default useTraceViewColumns;

View File

@@ -17,6 +17,7 @@ import { usePageActions } from 'container/AIAssistant/pageActions/usePageActions
import ExplorerOptionWrapper from 'container/ExplorerOptions/ExplorerOptionWrapper';
import { useOptionsMenu } from 'container/OptionsMenu';
import LeftToolbarActions from 'container/QueryBuilder/components/ToolbarActions/LeftToolbarActions';
import { AI_TRACE_VIEW_COLUMN_SELECTION } from 'container/AIObservability/TraceView/aiTraceViewColumns';
import RightToolbarActions from 'container/QueryBuilder/components/ToolbarActions/RightToolbarActions';
import Toolbar from 'container/Toolbar/Toolbar';
import {
@@ -330,6 +331,10 @@ function TracesExplorer(): JSX.Element {
setWarning={setWarning}
setIsLoadingQueries={setIsLoadingQueries}
queryKeyRef={listQueryKeyRef}
// DEMO ONLY — surfaces the AI Observability column set here because
// the AI Explorer does not exist yet. Drop this prop (and the import)
// to return Trace View to its five base columns.
columnSelection={AI_TRACE_VIEW_COLUMN_SELECTION}
/>
</div>
)}

View File

@@ -112,15 +112,12 @@ functionCall
;
/*
* Full-text search call: search('needle')
*
* Uses the shared functionParamList so future scoped forms like
* search(body, 'abc') / search(attribute, 'abc') need no grammar change. Today
* only a single needle is supported. Unlike bare/quoted free text (`fullText`),
* which only targets the body column, search() fans out across every field.
* Full-text search: search('term') or scoped search('term', body, ...).
* First param is the search term; the rest are field-context scopes (body/attribute/
* resource/log), quoted or bare. Handled in the visitor — no grammar change.
*/
searchCall
: SEARCH LPAREN functionParamList RPAREN
: SEARCH LPAREN valueList RPAREN
;
// Function parameters can be keys, single scalar values, or arrays

View File

@@ -241,9 +241,12 @@ func (server *Server) PutAlerts(ctx context.Context, postableAlerts alertmanager
}
func (server *Server) SetConfig(ctx context.Context, alertmanagerConfig *alertmanagertypes.Config) error {
config := alertmanagerConfig.AlertmanagerConfig()
resolved, err := alertmanagerConfig.Resolved()
if err != nil {
return err
}
config := resolved.AlertmanagerConfig()
var err error
// Load SigNoz's alertmanager notification templates from the configured
// globs. The upstream default templates (default.tmpl, email.tmpl) are
// always loaded from the embedded alertmanager assets inside FromGlobs, so
@@ -275,7 +278,7 @@ func (server *Server) SetConfig(ctx context.Context, alertmanagerConfig *alertma
server.logger.InfoContext(ctx, "skipping creation of receiver not referenced by any route", slog.String("receiver", rcv.Name))
continue
}
extendedRcv, err := alertmanagerConfig.GetReceiver(rcv.Name)
extendedRcv, err := resolved.GetReceiver(rcv.Name)
if err != nil {
return err
}
@@ -350,7 +353,7 @@ func (server *Server) SetConfig(ctx context.Context, alertmanagerConfig *alertma
go server.dispatcher.Run()
go server.inhibitor.Run()
server.alertmanagerConfig = alertmanagerConfig
server.alertmanagerConfig = resolved
return nil
}

View File

@@ -1,7 +1,6 @@
package alertmanager
import (
"net/url"
"strings"
"time"
@@ -26,11 +25,6 @@ type Signoz struct {
alertmanagerserver.Config `mapstructure:",squash" yaml:",squash"`
}
type Legacy struct {
// ApiURL is the URL of the legacy signoz alertmanager.
ApiURL *url.URL `mapstructure:"api_url"`
}
func NewConfigFactory() factory.ConfigFactory {
return factory.NewConfigFactory(factory.MustNewName("alertmanager"), newConfig)
}

View File

@@ -167,6 +167,10 @@ func (provider *provider) UpdateChannelByReceiverAndID(ctx context.Context, orgI
return err
}
if err := config.SetGlobalConfig(provider.config.Signoz.Global); err != nil {
return err
}
if err := config.UpdateReceiver(receiver); err != nil {
return err
}
@@ -217,6 +221,10 @@ func (provider *provider) CreateChannel(ctx context.Context, orgID string, recei
return nil, err
}
if err := config.SetGlobalConfig(provider.config.Signoz.Global); err != nil {
return nil, err
}
if err := config.CreateReceiver(receiver); err != nil {
return nil, err
}

View File

@@ -85,23 +85,6 @@ func (provider *provider) addDashboardRoutes(router *mux.Router) error {
return err
}
if err := router.Handle("/api/v2/dashboards/{id}/migrate", handler.New(provider.authzMiddleware.EditAccess(provider.dashboardHandler.MigrateV2), handler.OpenAPIDef{
ID: "MigrateDashboardV2",
Tags: []string{"dashboard"},
Summary: "Migrate dashboard to v2",
Description: "This endpoint retries the v1→v2 (Perses) migration on a dashboard still stored in the v1 schema and returns the v2-shape result. It is idempotent: a dashboard already in the v2 schema is returned unchanged.",
Request: nil,
RequestContentType: "",
Response: new(dashboardtypes.GettableDashboardV2),
ResponseContentType: "application/json",
SuccessStatusCode: http.StatusOK,
ErrorStatusCodes: []int{http.StatusBadRequest, http.StatusNotFound},
Deprecated: false,
SecuritySchemes: newSecuritySchemes(types.RoleEditor),
})).Methods(http.MethodPost).GetError(); err != nil {
return err
}
if err := router.Handle("/api/v2/dashboards/{id}", handler.New(provider.authzMiddleware.ViewAccess(provider.dashboardHandler.GetV2), handler.OpenAPIDef{
ID: "GetDashboardV2",
Tags: []string{"dashboard"},

View File

@@ -29,13 +29,18 @@ func New(t *testing.T) flagger.Flagger {
// WithUseJSONBody returns a Flagger with use_json_body set to the given value.
func WithUseJSONBody(t *testing.T, enabled bool) flagger.Flagger {
return WithBooleanFlags(t, map[string]bool{
flagger.FeatureUseJSONBody.String(): enabled,
})
}
// WithBooleanFlags returns a Flagger with the given boolean flags, keyed by feature name.
func WithBooleanFlags(t *testing.T, flags map[string]bool) flagger.Flagger {
t.Helper()
registry := flagger.MustNewRegistry()
cfg := flagger.Config{}
if enabled {
cfg.Config.Boolean = map[string]bool{
flagger.FeatureUseJSONBody.String(): true,
}
if len(flags) > 0 {
cfg.Config.Boolean = flags
}
fl, err := flagger.New(
context.Background(),

View File

@@ -63,9 +63,6 @@ type Module interface {
GetV2(ctx context.Context, orgID valuer.UUID, id valuer.UUID) (*dashboardtypes.DashboardV2, error)
// MigrateV2 retries the v1→v2 migration on a dashboard still stored in the v1 schema.
MigrateV2(ctx context.Context, orgID valuer.UUID, id valuer.UUID) (*dashboardtypes.DashboardV2, error)
ListV2(ctx context.Context, orgID valuer.UUID, params *dashboardtypes.ListDashboardsV2Params) (*dashboardtypes.ListableDashboardV2, error)
ListForUserV2(ctx context.Context, orgID valuer.UUID, userID valuer.UUID, params *dashboardtypes.ListDashboardsV2Params) (*dashboardtypes.ListableDashboardForUserV2, error)
@@ -135,8 +132,6 @@ type Handler interface {
GetV2(http.ResponseWriter, *http.Request)
MigrateV2(http.ResponseWriter, *http.Request)
ListV2(http.ResponseWriter, *http.Request)
ListForUserV2(http.ResponseWriter, *http.Request)

View File

@@ -207,38 +207,6 @@ func (handler *handler) GetV2(rw http.ResponseWriter, r *http.Request) {
render.Success(rw, http.StatusOK, dashboard.ToGettableDashboardV2())
}
func (handler *handler) MigrateV2(rw http.ResponseWriter, r *http.Request) {
ctx, cancel := context.WithTimeout(r.Context(), 10*time.Second)
defer cancel()
claims, err := authtypes.ClaimsFromContext(ctx)
if err != nil {
render.Error(rw, err)
return
}
orgID := valuer.MustNewUUID(claims.OrgID)
id := mux.Vars(r)["id"]
if id == "" {
render.Error(rw, errors.Newf(errors.TypeInvalidInput, errors.CodeInvalidInput, "id is missing in the path"))
return
}
dashboardID, err := valuer.NewUUID(id)
if err != nil {
render.Error(rw, err)
return
}
dashboard, err := handler.module.MigrateV2(ctx, orgID, dashboardID)
if err != nil {
render.Error(rw, err)
return
}
render.Success(rw, http.StatusOK, dashboard.ToGettableDashboardV2())
}
func (handler *handler) LockV2(rw http.ResponseWriter, r *http.Request) {
handler.lockUnlockV2(rw, r, true)
}

View File

@@ -4,7 +4,6 @@ import (
"context"
"github.com/SigNoz/signoz/pkg/errors"
"github.com/SigNoz/signoz/pkg/transition"
"github.com/SigNoz/signoz/pkg/types/coretypes"
"github.com/SigNoz/signoz/pkg/types/dashboardtypes"
"github.com/SigNoz/signoz/pkg/types/tagtypes"
@@ -122,51 +121,6 @@ func (module *module) GetV2(ctx context.Context, orgID valuer.UUID, id valuer.UU
return storable.ToDashboardV2(tags)
}
// MigrateV2 retries the v1→v2 migration on a dashboard still stored as v1 (one the
// bulk 103 migration skipped or failed). Idempotent: an already-v2 one is unchanged.
func (module *module) MigrateV2(ctx context.Context, orgID valuer.UUID, id valuer.UUID) (*dashboardtypes.DashboardV2, error) {
storable, err := module.store.Get(ctx, orgID, id)
if err != nil {
return nil, err
}
// Already migrated: return as-is.
if storable.IsV2() {
tags, err := module.tagModule.ListForResource(ctx, orgID, coretypes.KindDashboard, id)
if err != nil {
return nil, err
}
return storable.ToDashboardV2(tags)
}
// v1→v2 needs v5-shaped queries; run v4→v5 in place first.
transition.NewDashboardMigrateV5(module.settings.Logger(), nil, nil).Migrate(ctx, storable.Data)
v2, err := storable.ConvertV1ToV2()
if err != nil {
return nil, err
}
err = module.store.RunInTx(ctx, func(ctx context.Context) error {
resolvedTags, err := module.tagModule.SyncTags(ctx, orgID, coretypes.KindDashboard, v2.ID, tagtypes.NewPostableTagsFromTags(v2.Tags))
if err != nil {
return err
}
v2.Tags = resolvedTags
storableV2, err := v2.ToStorableDashboard()
if err != nil {
return err
}
return module.store.Update(ctx, orgID, storableV2)
})
if err != nil {
return nil, err
}
return v2, nil
}
func (module *module) UpdateV2(ctx context.Context, orgID valuer.UUID, id valuer.UUID, updatedBy string, updatable dashboardtypes.UpdatableDashboardV2) (*dashboardtypes.DashboardV2, error) {
if err := updatable.Validate(); err != nil {
return nil, err

View File

@@ -35,7 +35,7 @@ func (c *conditionBuilder) ConditionFor(
sb *sqlbuilder.SelectBuilder,
) ([]string, []string, error) {
// has/hasAny/hasAll/hasToken are logs-body-only; reject for rule state history.
// has/hasAny/hasAll/hasToken/search are logs-only functions; reject for rule state history.
if err := querybuilder.NewFunctionUnsupportedError(operator); err != nil {
return nil, nil, err
}

File diff suppressed because one or more lines are too long

View File

@@ -137,8 +137,8 @@ func filterqueryParserInit() {
0, 0, 191, 19, 1, 0, 0, 0, 192, 190, 1, 0, 0, 0, 193, 194, 7, 2, 0, 0,
194, 21, 1, 0, 0, 0, 195, 196, 7, 3, 0, 0, 196, 197, 5, 1, 0, 0, 197, 198,
3, 26, 13, 0, 198, 199, 5, 2, 0, 0, 199, 23, 1, 0, 0, 0, 200, 201, 5, 27,
0, 0, 201, 202, 5, 1, 0, 0, 202, 203, 3, 26, 13, 0, 203, 204, 5, 2, 0,
0, 204, 25, 1, 0, 0, 0, 205, 210, 3, 28, 14, 0, 206, 207, 5, 5, 0, 0, 207,
0, 0, 201, 202, 5, 1, 0, 0, 202, 203, 3, 18, 9, 0, 203, 204, 5, 2, 0, 0,
204, 25, 1, 0, 0, 0, 205, 210, 3, 28, 14, 0, 206, 207, 5, 5, 0, 0, 207,
209, 3, 28, 14, 0, 208, 206, 1, 0, 0, 0, 209, 212, 1, 0, 0, 0, 210, 208,
1, 0, 0, 0, 210, 211, 1, 0, 0, 0, 211, 27, 1, 0, 0, 0, 212, 210, 1, 0,
0, 0, 213, 217, 3, 34, 17, 0, 214, 217, 3, 32, 16, 0, 215, 217, 3, 30,
@@ -2945,7 +2945,7 @@ type ISearchCallContext interface {
// Getter signatures
SEARCH() antlr.TerminalNode
LPAREN() antlr.TerminalNode
FunctionParamList() IFunctionParamListContext
ValueList() IValueListContext
RPAREN() antlr.TerminalNode
// IsSearchCallContext differentiates from other interfaces.
@@ -2992,10 +2992,10 @@ func (s *SearchCallContext) LPAREN() antlr.TerminalNode {
return s.GetToken(FilterQueryParserLPAREN, 0)
}
func (s *SearchCallContext) FunctionParamList() IFunctionParamListContext {
func (s *SearchCallContext) ValueList() IValueListContext {
var t antlr.RuleContext
for _, ctx := range s.GetChildren() {
if _, ok := ctx.(IFunctionParamListContext); ok {
if _, ok := ctx.(IValueListContext); ok {
t = ctx.(antlr.RuleContext)
break
}
@@ -3005,7 +3005,7 @@ func (s *SearchCallContext) FunctionParamList() IFunctionParamListContext {
return nil
}
return t.(IFunctionParamListContext)
return t.(IValueListContext)
}
func (s *SearchCallContext) RPAREN() antlr.TerminalNode {
@@ -3064,7 +3064,7 @@ func (p *FilterQueryParser) SearchCall() (localctx ISearchCallContext) {
}
{
p.SetState(202)
p.FunctionParamList()
p.ValueList()
}
{
p.SetState(203)

View File

@@ -20,6 +20,8 @@ import (
"github.com/SigNoz/signoz/pkg/valuer"
)
const estimateTimeout = 5 * time.Second
const traceOutsideRangeWarn = "Query %s references a trace_id that exists between %s and %s (UTC) but lies outside the selected time range; adjust the time range to see results"
type builderQuery[T any] struct {
@@ -247,6 +249,10 @@ func (q *builderQuery[T]) Execute(ctx context.Context) (*qbtypes.Result, error)
return nil, err
}
if err := q.enforceEstimate(ctx, stmt); err != nil {
return nil, err
}
// Execute the query with proper context for partial value detection
result, err := q.executeWithContext(ctx, stmt.Query, stmt.Args)
if err != nil {
@@ -258,6 +264,65 @@ func (q *builderQuery[T]) Execute(ctx context.Context) (*qbtypes.Result, error)
return result, nil
}
// estimateRows returns the per-shard EXPLAIN ESTIMATE scan rows for a cost-guarded
// statement. guarded=false means nothing to enforce; a non-nil error means reject.
// Callers own the budget comparison (per-statement or cumulative).
func (q *builderQuery[T]) estimateRows(ctx context.Context, stmt *qbtypes.Statement) (int64, bool, error) {
if stmt.CostGuard == nil || stmt.CostGuard.MaxScanRows <= 0 {
return 0, false, nil
}
estCtx, cancel := context.WithTimeout(ctx, estimateTimeout)
defer cancel()
entries, err := q.telemetryStore.Estimate(estCtx, stmt.Query, stmt.Args...)
if err != nil {
if ctx.Err() != nil {
return 0, true, ctx.Err()
}
if errors.Is(estCtx.Err(), context.DeadlineExceeded) {
return 0, true, errors.NewInvalidInputf(errors.CodeInvalidInput,
"This query is too broad to plan within %s", estimateTimeout).
WithSuggestions(costGuardSuggestions(stmt.CostGuard.Warning)...)
}
return 0, true, err
}
var rows int64
for _, e := range entries {
rows += e.Rows
}
return rows, true, nil
}
// enforceEstimate rejects a scan-heavy statement whose estimate exceeds its own
// budget, before executing. Budget 0 disables.
func (q *builderQuery[T]) enforceEstimate(ctx context.Context, stmt *qbtypes.Statement) error {
rows, guarded, err := q.estimateRows(ctx, stmt)
if err != nil {
return err
}
if !guarded {
return nil
}
if budget := stmt.CostGuard.MaxScanRows; rows > budget {
return errors.NewInvalidInputf(errors.CodeInvalidInput,
"This query would scan about %d rows per shard in this range, over the per-shard limit of %d", rows, budget).
WithSuggestions(costGuardSuggestions(stmt.CostGuard.Warning)...)
}
return nil
}
// costGuardSuggestions leads with the requirement's advisory (e.g. the search() hint),
// then how to get under budget.
func costGuardSuggestions(advisory string) []string {
suggestions := make([]string, 0, 2)
if advisory != "" {
suggestions = append(suggestions, advisory)
}
return append(suggestions, "Narrow the time range or add a more selective filter.")
}
// narrowWindowByTraceID inspects the filter for trace_id predicates and clamps
// [fromMS,toMS] to the time range stored in signoz_traces.distributed_trace_summary.
// Returns the (possibly narrowed) window, overlap=false when the trace lies
@@ -491,6 +556,10 @@ func (q *builderQuery[T]) executeWindowList(ctx context.Context) (*qbtypes.Resul
var warnings []string
var warningsDocURL string
// Cumulative across visited buckets: the budget bounds the whole query's per-shard
// scan, not each bucket independently.
var estimatedScan int64
for _, r := range buckets {
q.spec.Offset = 0
q.spec.Limit = need
@@ -501,6 +570,18 @@ func (q *builderQuery[T]) executeWindowList(ctx context.Context) (*qbtypes.Resul
}
warnings = stmt.Warnings
warningsDocURL = stmt.WarningsDocURL
rowsEst, guarded, err := q.estimateRows(ctx, stmt)
if err != nil {
return nil, err
}
if guarded {
estimatedScan += rowsEst
if budget := stmt.CostGuard.MaxScanRows; estimatedScan > budget {
return nil, errors.NewInvalidInputf(errors.CodeInvalidInput,
"This query would scan about %d rows per shard across the time range, over the per-shard limit of %d", estimatedScan, budget).
WithSuggestions(costGuardSuggestions(stmt.CostGuard.Warning)...)
}
}
// Execute with proper context for partial value detection
res, err := q.executeWithContext(ctx, stmt.Query, stmt.Args)
if err != nil {

View File

@@ -8,6 +8,9 @@ const (
// BodyFullTextSearchDefaultWarning is emitted when a full-text search or "body" searches are hit
// with New JSON Body enhancements.
BodyFullTextSearchDefaultWarning = "Full text searches default to `body.message:string`. Use `body.<key>` to search a different field inside body"
// SearchWarning is emitted on every search() call — it scans all fields.
SearchWarning = "search() runs across all fields and can be slow and expensive. Prefer a specific field, e.g. `<context>.<field_key>:<type>`"
)
var (

View File

@@ -161,14 +161,16 @@ func inferDataTypesFromList(values []any) []telemetrytypes.FieldDataType {
return out
}
// NewFunctionUnsupportedError returns the error for a has/hasAny/hasAll/hasToken operator
// on a builder that doesn't support it (logs body only), or nil for other operators.
// NewFunctionUnsupportedError returns the error for a has/hasAny/hasAll/hasToken/search
// operator on a builder that doesn't support it (logs only), or nil for other operators.
func NewFunctionUnsupportedError(operator qbtypes.FilterOperator) error {
switch operator {
case qbtypes.FilterOperatorHasToken:
return errors.NewInvalidInputf(errors.CodeInvalidInput, "function `hasToken` only supports body field as first parameter").WithUrl(hasTokenFunctionDocURL)
case qbtypes.FilterOperatorHas, qbtypes.FilterOperatorHasAny, qbtypes.FilterOperatorHasAll:
return errors.NewInvalidInputf(errors.CodeInvalidInput, "function `%s` supports only body JSON search", operator.FunctionName()).WithUrl(functionBodyJSONSearchDocURL)
case qbtypes.FilterOperatorSearch:
return errors.NewInvalidInputf(errors.CodeInvalidInput, "function `search` is only supported for logs")
default:
return nil
}

View File

@@ -43,6 +43,8 @@ type filterExpressionVisitor struct {
keysWithWarnings map[string]bool
startNs uint64
endNs uint64
requiresCostGuard bool
}
type FilterExprVisitorOpts struct {
@@ -81,9 +83,10 @@ func newFilterExpressionVisitor(opts FilterExprVisitorOpts) *filterExpressionVis
}
type PreparedWhereClause struct {
WhereClause *sqlbuilder.WhereClause
Warnings []string
WarningsDocURL string
WhereClause *sqlbuilder.WhereClause
Warnings []string
WarningsDocURL string
RequiresCostGuard bool
}
func (p PreparedWhereClause) IsEmpty() bool {
@@ -165,12 +168,12 @@ func PrepareWhereClause(query string, opts FilterExprVisitorOpts) (PreparedWhere
// Return empty where clause so callers can skip the WHERE clause
if cond == "" || cond == SkipConditionLiteral {
return PreparedWhereClause{WhereClause: nil, Warnings: visitor.warnings, WarningsDocURL: visitor.mainWarnURL}, nil
return PreparedWhereClause{WhereClause: nil, Warnings: visitor.warnings, WarningsDocURL: visitor.mainWarnURL, RequiresCostGuard: visitor.requiresCostGuard}, nil
}
whereClause := sqlbuilder.NewWhereClause().AddWhereExpr(visitor.builder.Args, cond)
return PreparedWhereClause{WhereClause: whereClause, Warnings: visitor.warnings, WarningsDocURL: visitor.mainWarnURL}, nil
return PreparedWhereClause{WhereClause: whereClause, Warnings: visitor.warnings, WarningsDocURL: visitor.mainWarnURL, RequiresCostGuard: visitor.requiresCostGuard}, nil
}
// Visit dispatches to the specific visit method based on node type.
@@ -776,11 +779,77 @@ func normalizeFunctionValue(operator qbtypes.FilterOperator, functionName string
return valueParams, nil
}
// VisitSearchCall handles search('needle'). The search() function is parsed but
// not yet implemented; reject it with a clear invalid-input error.
// VisitSearchCall handles search('term'[, body, resource, …]): a case-insensitive
// search term plus optional field-context scopes, ORing one FilterOperatorSearch per
// scope (no scope = keyless, covering every field).
func (v *filterExpressionVisitor) VisitSearchCall(ctx *grammar.SearchCallContext) any {
v.errors = append(v.errors, "function `search` is not yet supported")
return ErrorConditionLiteral
// Flag scan-heavy so the statement builder attaches the cost guard.
v.requiresCostGuard = true
valueList := ctx.ValueList()
if valueList == nil {
v.errors = append(v.errors, "function `search` expects a search term, e.g. search('error')")
return ErrorConditionLiteral
}
params := valueList.AllValue()
if len(params) == 0 {
v.errors = append(v.errors, "function `search` expects a search term, e.g. search('error')")
return ErrorConditionLiteral
}
searchText, ok := searchParamText(params[0])
if !ok {
v.errors = append(v.errors, "function `search` expects a search term as its first argument, e.g. search('error')")
return ErrorConditionLiteral
}
var fieldContexts []telemetrytypes.FieldContext
if len(params) == 1 {
fieldContexts = []telemetrytypes.FieldContext{telemetrytypes.FieldContextUnspecified}
} else {
for _, p := range params[1:] {
scopeText, sok := searchParamText(p)
if !sok {
v.errors = append(v.errors, "function `search` expects each scope to be a context, e.g. search('error', body, resource)")
return ErrorConditionLiteral
}
fc, fok := telemetrytypes.FieldContextFromText(scopeText)
if !fok {
v.errors = append(v.errors, fmt.Sprintf("invalid search scope %q; expected a field context: body, attribute, resource, or log", scopeText))
return ErrorConditionLiteral
}
fieldContexts = append(fieldContexts, fc)
}
}
var conds []string
for _, fieldContext := range fieldContexts {
key := telemetrytypes.NewTelemetryFieldKey("", fieldContext, telemetrytypes.FieldDataTypeUnspecified)
scoped, cok := v.buildConditions(key, nil, qbtypes.FilterOperatorSearch, searchText)
if !cok {
return ErrorConditionLiteral
}
conds = append(conds, scoped...)
}
if len(conds) == 0 {
return SkipConditionLiteral
}
if len(conds) == 1 {
return conds[0]
}
return v.builder.Or(conds...)
}
// searchParamText returns an argument's raw token text (quoted or bare) rather than its
// visited value, so a bare word stays literal and search(1000000) isn't "1e+06".
func searchParamText(val grammar.IValueContext) (string, bool) {
if val == nil {
return "", false
}
if val.QUOTED_TEXT() != nil {
return trimQuotes(val.QUOTED_TEXT().GetText()), true
}
return val.GetText(), true
}
// VisitFunctionParamList handles the parameter list for function calls.

View File

@@ -233,6 +233,7 @@ func NewSQLMigrationProviderFactories(
sqlmigration.NewFillDashboardMeterSourceFactory(sqlstore, dashboardStore),
sqlmigration.NewUpdateRoleTransactionGroupsFactory(),
sqlmigration.NewFillDashboardSpecCollectionsFactory(sqlstore, dashboardStore),
sqlmigration.NewScrubEmailChannelTransportFactory(sqlstore),
)
}

View File

@@ -0,0 +1,274 @@
package sqlmigration
import (
"context"
"crypto/md5"
"encoding/json"
"fmt"
"log/slog"
"github.com/SigNoz/signoz/pkg/errors"
"github.com/SigNoz/signoz/pkg/factory"
"github.com/SigNoz/signoz/pkg/sqlstore"
"github.com/uptrace/bun"
"github.com/uptrace/bun/migrate"
)
type scrubEmailChannelTransport struct {
sqlstore sqlstore.SQLStore
logger *slog.Logger
}
type alertmanagerConfigScrubRow struct {
bun.BaseModel `bun:"table:alertmanager_config"`
ID string `bun:"id"`
Config string `bun:"config"`
}
type notificationChannelScrubRow struct {
bun.BaseModel `bun:"table:notification_channel"`
ID string `bun:"id"`
Data string `bun:"data"`
}
var emailTransportKeys = []string{
"from",
"hello",
"smarthost",
"auth_username",
"auth_password",
"auth_password_file",
"auth_secret",
"auth_secret_file",
"auth_identity",
"require_tls",
"tls_config",
"force_implicit_tls",
}
var globalSMTPKeys = []string{
"smtp_from",
"smtp_hello",
"smtp_smarthost",
"smtp_auth_username",
"smtp_auth_password",
"smtp_auth_password_file",
"smtp_auth_secret",
"smtp_auth_secret_file",
"smtp_auth_identity",
"smtp_require_tls",
"smtp_tls_config",
"smtp_force_implicit_tls",
}
func NewScrubEmailChannelTransportFactory(sqlstore sqlstore.SQLStore) factory.ProviderFactory[SQLMigration, Config] {
return factory.NewProviderFactory(
factory.MustNewName("scrub_email_channel_transport"),
func(ctx context.Context, ps factory.ProviderSettings, c Config) (SQLMigration, error) {
return &scrubEmailChannelTransport{sqlstore: sqlstore, logger: ps.Logger}, nil
},
)
}
func (migration *scrubEmailChannelTransport) Register(migrations *migrate.Migrations) error {
if err := migrations.Register(migration.Up, migration.Down); err != nil {
return err
}
return nil
}
func (migration *scrubEmailChannelTransport) Up(ctx context.Context, db *bun.DB) error {
tx, err := db.BeginTx(ctx, nil)
if err != nil {
return err
}
defer func() {
_ = tx.Rollback()
}()
if err := migration.scrubConfigs(ctx, tx); err != nil {
return err
}
if err := migration.scrubChannels(ctx, tx); err != nil {
return err
}
return tx.Commit()
}
func (migration *scrubEmailChannelTransport) scrubConfigs(ctx context.Context, tx bun.Tx) error {
rows := make([]*alertmanagerConfigScrubRow, 0)
if err := tx.NewSelect().Model(&rows).Scan(ctx); err != nil {
return err
}
for _, row := range rows {
cfg := make(map[string]json.RawMessage)
if err := json.Unmarshal([]byte(row.Config), &cfg); err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config with unreadable config", slog.String("config_id", row.ID), errors.Attr(err))
continue
}
changed := false
if globalRaw, ok := cfg["global"]; ok && string(globalRaw) != "null" {
global := make(map[string]json.RawMessage)
if err := json.Unmarshal(globalRaw, &global); err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config with unreadable global", slog.String("config_id", row.ID), errors.Attr(err))
continue
}
if deleteKeys(global, globalSMTPKeys) {
newGlobal, err := json.Marshal(global)
if err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config, cannot marshal global", slog.String("config_id", row.ID), errors.Attr(err))
continue
}
cfg["global"] = newGlobal
changed = true
}
}
if receiversRaw, ok := cfg["receivers"]; ok && string(receiversRaw) != "null" {
receivers := make([]map[string]json.RawMessage, 0)
if err := json.Unmarshal(receiversRaw, &receivers); err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config with unreadable receivers", slog.String("config_id", row.ID), errors.Attr(err))
continue
}
receiversChanged := false
unreadable := false
for _, receiver := range receivers {
scrubbed, err := scrubEmailConfigs(receiver)
if err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config with unreadable email configs", slog.String("config_id", row.ID), errors.Attr(err))
unreadable = true
break
}
receiversChanged = receiversChanged || scrubbed
}
if unreadable {
continue
}
if receiversChanged {
newReceivers, err := json.Marshal(receivers)
if err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config, cannot marshal receivers", slog.String("config_id", row.ID), errors.Attr(err))
continue
}
cfg["receivers"] = newReceivers
changed = true
}
}
if !changed {
continue
}
newConfig, err := json.Marshal(cfg)
if err != nil {
migration.logger.WarnContext(ctx, "skipping alertmanager config, cannot marshal config", slog.String("config_id", row.ID), errors.Attr(err))
continue
}
if _, err := tx.NewUpdate().
Model((*alertmanagerConfigScrubRow)(nil)).
Set("config = ?", string(newConfig)).
Set("hash = ?", fmt.Sprintf("%x", md5.Sum(newConfig))).
Where("id = ?", row.ID).
Exec(ctx); err != nil {
return err
}
}
return nil
}
func (migration *scrubEmailChannelTransport) scrubChannels(ctx context.Context, tx bun.Tx) error {
rows := make([]*notificationChannelScrubRow, 0)
if err := tx.NewSelect().Model(&rows).Where("type = ?", "email").Scan(ctx); err != nil {
return err
}
for _, row := range rows {
receiver := make(map[string]json.RawMessage)
if err := json.Unmarshal([]byte(row.Data), &receiver); err != nil {
migration.logger.WarnContext(ctx, "skipping notification channel with unreadable data", slog.String("channel_id", row.ID), errors.Attr(err))
continue
}
scrubbed, err := scrubEmailConfigs(receiver)
if err != nil {
migration.logger.WarnContext(ctx, "skipping notification channel with unreadable email configs", slog.String("channel_id", row.ID), errors.Attr(err))
continue
}
if !scrubbed {
continue
}
newData, err := json.Marshal(receiver)
if err != nil {
migration.logger.WarnContext(ctx, "skipping notification channel, cannot marshal data", slog.String("channel_id", row.ID), errors.Attr(err))
continue
}
if _, err := tx.NewUpdate().
Model((*notificationChannelScrubRow)(nil)).
Set("data = ?", string(newData)).
Where("id = ?", row.ID).
Exec(ctx); err != nil {
return err
}
}
return nil
}
func scrubEmailConfigs(receiver map[string]json.RawMessage) (bool, error) {
emailConfigsRaw, ok := receiver["email_configs"]
if !ok || string(emailConfigsRaw) == "null" {
return false, nil
}
emailConfigs := make([]map[string]json.RawMessage, 0)
if err := json.Unmarshal(emailConfigsRaw, &emailConfigs); err != nil {
return false, err
}
changed := false
for _, emailConfig := range emailConfigs {
changed = deleteKeys(emailConfig, emailTransportKeys) || changed
}
if !changed {
return false, nil
}
newEmailConfigs, err := json.Marshal(emailConfigs)
if err != nil {
return false, err
}
receiver["email_configs"] = newEmailConfigs
return true, nil
}
func deleteKeys(m map[string]json.RawMessage, keys []string) bool {
deleted := false
for _, key := range keys {
if _, ok := m[key]; ok {
delete(m, key)
deleted = true
}
}
return deleted
}
func (migration *scrubEmailChannelTransport) Down(context.Context, *bun.DB) error {
return nil
}

View File

@@ -15,6 +15,12 @@ type SkipResourceFingerprint struct {
type Config struct {
// SkipResourceFingerprint configures when the resource fingerprint subquery is skipped in favor of main-table filtering.
SkipResourceFingerprint SkipResourceFingerprint `yaml:"skip_resource_fingerprint" mapstructure:"skip_resource_fingerprint"`
// SearchMaxScanRows caps per-shard rows a search() may scan, enforced by the querier
// via EXPLAIN ESTIMATE (0 disables).
SearchMaxScanRows int64 `yaml:"search_max_scan_rows" mapstructure:"search_max_scan_rows"`
// SearchMaxScanRowsJSONBody is the same budget for body_v2, where each row costs far
// more: toString() rebuilds every document and no skip index prunes (0 disables).
SearchMaxScanRowsJSONBody int64 `yaml:"search_max_scan_rows_json_body" mapstructure:"search_max_scan_rows_json_body"`
}
func NewConfig() Config {
@@ -23,6 +29,8 @@ func NewConfig() Config {
Enabled: false,
Threshold: 100000,
},
SearchMaxScanRows: 60_000_000,
SearchMaxScanRowsJSONBody: 6_000_000,
}
}
@@ -31,5 +39,11 @@ func (c Config) Validate() error {
if c.SkipResourceFingerprint.Enabled && c.SkipResourceFingerprint.Threshold == 0 {
return errors.NewInvalidInputf(errors.CodeInvalidInput, "skip_resource_fingerprint.threshold must be > 0 when enabled")
}
if c.SearchMaxScanRows < 0 {
return errors.NewInvalidInputf(errors.CodeInvalidInput, "search_max_scan_rows must not be negative, got %v", c.SearchMaxScanRows)
}
if c.SearchMaxScanRowsJSONBody < 0 {
return errors.NewInvalidInputf(errors.CodeInvalidInput, "search_max_scan_rows_json_body must not be negative, got %v", c.SearchMaxScanRowsJSONBody)
}
return nil
}

View File

@@ -0,0 +1,85 @@
package logsstatementbuilder
import (
"context"
"testing"
"time"
"github.com/SigNoz/signoz/pkg/flagger/flaggertest"
"github.com/SigNoz/signoz/pkg/instrumentation/instrumentationtest"
"github.com/SigNoz/signoz/pkg/querybuilder"
"github.com/SigNoz/signoz/pkg/statementbuilder"
"github.com/SigNoz/signoz/pkg/telemetryschema/logstelemetryschema"
qbtypes "github.com/SigNoz/signoz/pkg/types/querybuildertypes/querybuildertypesv5"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes/telemetrytypestest"
"github.com/SigNoz/signoz/pkg/valuer"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
// TestSearchCostGuard asserts Build attaches the CostGuard budget and its advisory.
func TestSearchCostGuard(t *testing.T) {
releaseTime := time.Date(2024, 1, 15, 10, 0, 0, 0, time.UTC)
ctx := context.Background()
start := uint64(releaseTime.Add(-5 * time.Minute).UnixMilli())
end := uint64(releaseTime.UnixMilli())
fl := flaggertest.WithBooleanFlags(t, map[string]bool{})
fm := logstelemetryschema.NewFieldMapper(fl)
cb := logstelemetryschema.NewConditionBuilder(fm, fl)
store := telemetrytypestest.NewMockMetadataStore()
store.KeysMap = logstelemetryschema.BuildCompleteFieldKeyMap(releaseTime)
rewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, fl)
sb := NewLogQueryStatementBuilder(
instrumentationtest.New().ToProviderSettings(),
store, fm, cb, rewriter, logstelemetryschema.DefaultFullTextColumn, fl, nil,
statementbuilder.Config{SearchMaxScanRows: 100000, SkipResourceFingerprint: statementbuilder.SkipResourceFingerprint{Enabled: false, Threshold: 100000}},
)
query := qbtypes.QueryBuilderQuery[qbtypes.LogAggregation]{
Signal: telemetrytypes.SignalLogs,
Filter: &qbtypes.Filter{Expression: "search('error')"},
Limit: 1,
}
stmt, err := sb.Build(ctx, valuer.UUID{}, start, end, qbtypes.RequestTypeRaw, query, nil)
require.NoError(t, err)
require.NotNil(t, stmt.CostGuard)
assert.Equal(t, int64(100000), stmt.CostGuard.MaxScanRows)
assert.Contains(t, stmt.Warnings, querybuilder.SearchWarning)
}
// TestSearchCostGuardJSONBody asserts body_v2 gets its own, lower budget.
func TestSearchCostGuardJSONBody(t *testing.T) {
releaseTime := time.Date(2024, 1, 15, 10, 0, 0, 0, time.UTC)
ctx := context.Background()
start := uint64(releaseTime.Add(-5 * time.Minute).UnixMilli())
end := uint64(releaseTime.UnixMilli())
fl := flaggertest.WithUseJSONBody(t, true)
fm := logstelemetryschema.NewFieldMapper(fl)
cb := logstelemetryschema.NewConditionBuilder(fm, fl)
store := telemetrytypestest.NewMockMetadataStore()
store.KeysMap = logstelemetryschema.BuildCompleteFieldKeyMap(releaseTime)
rewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, fm, cb, fl)
sb := NewLogQueryStatementBuilder(
instrumentationtest.New().ToProviderSettings(),
store, fm, cb, rewriter, logstelemetryschema.DefaultFullTextColumn, fl, nil,
statementbuilder.Config{
SearchMaxScanRows: 100000,
SearchMaxScanRowsJSONBody: 10000,
SkipResourceFingerprint: statementbuilder.SkipResourceFingerprint{Enabled: false, Threshold: 100000},
},
)
query := qbtypes.QueryBuilderQuery[qbtypes.LogAggregation]{
Signal: telemetrytypes.SignalLogs,
Filter: &qbtypes.Filter{Expression: "search('error')"},
Limit: 1,
}
stmt, err := sb.Build(ctx, valuer.UUID{}, start, end, qbtypes.RequestTypeRaw, query, nil)
require.NoError(t, err)
require.NotNil(t, stmt.CostGuard)
assert.Equal(t, int64(10000), stmt.CostGuard.MaxScanRows)
assert.Contains(t, stmt.Warnings, querybuilder.SearchWarning)
}

View File

@@ -41,14 +41,16 @@ type logQueryStatementBuilder struct {
fl flagger.Flagger
skipResourceFingerprintEnabled bool
fullTextColumn *telemetrytypes.TelemetryFieldKey
fullTextColumn *telemetrytypes.TelemetryFieldKey
searchMaxScanRows int64
searchMaxScanRowsJSONBody int64
}
var _ qbtypes.StatementBuilder[qbtypes.LogAggregation] = (*logQueryStatementBuilder)(nil)
// NewFactory returns a provider factory for the logs statement builder. Its New
// internalizes the FieldMapper, ConditionBuilder, and AggExprRewriter, and reads
// SkipResourceFingerprint from the config.
// SkipResourceFingerprint and the search() scan budgets from the config.
func NewFactory(
telemetryStore telemetrystore.TelemetryStore,
metadataStore telemetrytypes.MetadataStore,
@@ -94,7 +96,7 @@ func NewLogQueryStatementBuilder(
cfg.SkipResourceFingerprint.Threshold,
)
return &logQueryStatementBuilder{
b := &logQueryStatementBuilder{
logger: logsSettings.Logger(),
metadataStore: metadataStore,
fm: fieldMapper,
@@ -103,8 +105,11 @@ func NewLogQueryStatementBuilder(
aggExprRewriter: aggExprRewriter,
fl: fl,
skipResourceFingerprintEnabled: cfg.SkipResourceFingerprint.Enabled,
searchMaxScanRows: cfg.SearchMaxScanRows,
searchMaxScanRowsJSONBody: cfg.SearchMaxScanRowsJSONBody,
fullTextColumn: fullTextColumn,
}
return b
}
// Build builds a SQL query for logs based on the given parameters.
@@ -150,9 +155,26 @@ func (b *logQueryStatementBuilder) Build(
}
stmt.Warnings = append(stmt.Warnings, warnings...)
// Surface the guard's advisory to the user alongside the other warnings.
if stmt.CostGuard != nil && stmt.CostGuard.Warning != "" {
stmt.Warnings = append(stmt.Warnings, stmt.CostGuard.Warning)
}
return stmt, nil
}
// costGuardFor pairs the search() advisory with the budget for the body path taken —
// body_v2 has its own, lower one. nil when the statement needs no guard.
func (b *logQueryStatementBuilder) costGuardFor(ctx context.Context, orgID valuer.UUID, required bool) *qbtypes.CostGuard {
if !required {
return nil
}
maxScanRows := b.searchMaxScanRows
if b.fl.BooleanOrEmpty(ctx, flagger.FeatureUseJSONBody, featuretypes.NewFlaggerEvaluationContext(orgID)) {
maxScanRows = b.searchMaxScanRowsJSONBody
}
return &qbtypes.CostGuard{Warning: querybuilder.SearchWarning, MaxScanRows: maxScanRows}
}
func getKeySelectors(query qbtypes.QueryBuilderQuery[qbtypes.LogAggregation], bodyJSONEnabled bool) ([]*telemetrytypes.FieldKeySelector, []string) {
var keySelectors []*telemetrytypes.FieldKeySelector
var warnings []string
@@ -389,6 +411,7 @@ func (b *logQueryStatementBuilder) buildListQuery(
Args: finalArgs,
Warnings: preparedWhereClause.Warnings,
WarningsDocURL: preparedWhereClause.WarningsDocURL,
CostGuard: b.costGuardFor(ctx, orgID, preparedWhereClause.RequiresCostGuard),
}
return stmt, nil
@@ -555,6 +578,7 @@ func (b *logQueryStatementBuilder) buildTimeSeriesQuery(
Args: finalArgs,
Warnings: preparedWhereClause.Warnings,
WarningsDocURL: preparedWhereClause.WarningsDocURL,
CostGuard: b.costGuardFor(ctx, orgID, preparedWhereClause.RequiresCostGuard),
}
return stmt, nil
@@ -682,6 +706,7 @@ func (b *logQueryStatementBuilder) buildScalarQuery(
Args: finalArgs,
Warnings: preparedWhereClause.Warnings,
WarningsDocURL: preparedWhereClause.WarningsDocURL,
CostGuard: b.costGuardFor(ctx, orgID, preparedWhereClause.RequiresCostGuard),
}
return stmt, nil

View File

@@ -134,7 +134,7 @@ func (c *conditionBuilder) ConditionFor(
sb *sqlbuilder.SelectBuilder,
) ([]string, []string, error) {
// has/hasAny/hasAll/hasToken are logs-body-only; reject for audit.
// has/hasAny/hasAll/hasToken/search are logs-only functions; reject for audit.
if err := querybuilder.NewFunctionUnsupportedError(operator); err != nil {
return nil, nil, err
}

View File

@@ -3,6 +3,7 @@ package logstelemetryschema
import (
"context"
"fmt"
"regexp"
schema "github.com/SigNoz/signoz-otel-collector/cmd/signozschemamigrator/schema_migrator"
"github.com/SigNoz/signoz/pkg/errors"
@@ -27,6 +28,52 @@ func NewConditionBuilder(fm qbtypes.FieldMapper, fl flagger.Flagger) *conditionB
return &conditionBuilder{fm: fm, fl: fl}
}
// conditionForSearch ORs a case-insensitive match of the search term across the key
// context's searchable columns (unspecified context = every column).
func (c *conditionBuilder) conditionForSearch(
ctx context.Context,
orgID valuer.UUID,
key *telemetrytypes.TelemetryFieldKey,
value any,
sb *sqlbuilder.SelectBuilder,
) ([]string, []string, error) {
// QuoteMeta + LOWER on both sides, not (?i): a literal match that can still use the
// LOWER(toString(body_v2)) skip index.
term := regexp.QuoteMeta(fmt.Sprintf("%v", value))
useJSONBody := c.fl.BooleanOrEmpty(ctx, flagger.FeatureUseJSONBody, featuretypes.NewFlaggerEvaluationContext(orgID))
var conditions []string
for _, col := range searchColumns(key.FieldContext, useJSONBody) {
switch col.Type.GetType() {
case schema.ColumnTypeEnumMap:
keysExpr := fmt.Sprintf("mapKeys(%s)", col.Name)
valsExpr := fmt.Sprintf("mapValues(%s)", col.Name)
// match() needs a String array; cast non-string map values first.
if mc, ok := col.Type.(schema.MapColumnType); ok && mc.ValueType.GetType() != schema.ColumnTypeEnumString {
valsExpr = fmt.Sprintf("arrayMap(x -> toString(x), mapValues(%s))", col.Name)
}
conditions = append(conditions, sb.Or(
fmt.Sprintf("arrayExists(x -> match(LOWER(x), LOWER(%s)), %s)", sb.Var(term), keysExpr),
fmt.Sprintf("arrayExists(x -> match(LOWER(x), LOWER(%s)), %s)", sb.Var(term), valsExpr),
))
case schema.ColumnTypeEnumJSON:
conditions = append(conditions, fmt.Sprintf("match(LOWER(toString(%s)), LOWER(%s))", col.Name, sb.Var(term)))
case schema.ColumnTypeEnumString, schema.ColumnTypeEnumLowCardinality:
conditions = append(conditions, fmt.Sprintf("match(LOWER(%s), LOWER(%s))", col.Name, sb.Var(term)))
default:
return nil, nil, errors.NewInternalf(errors.CodeInternal, "search does not support the column type of %q", col.Name)
}
}
if len(conditions) == 0 {
return nil, nil, nil
}
// The advisory rides on CostGuard (set by the visitor), not warnings.
return []string{sb.Or(conditions...)}, nil, nil
}
// isBodyJSONSearch reports whether a key addresses a path within the body JSON. Only
// an explicit Body context qualifies; a bare, context-less `body` (e.g. full-text
// `count_distinct(body)` or `body EXISTS`) is a full-text match, not a `$.body` path.
@@ -408,6 +455,11 @@ func (c *conditionBuilder) ConditionFor(
matches := querybuilder.MatchingFieldKeys(key, fieldKeys)
skipResourceFilter := options.SkipResourceFilter
// search() resolves its own (optional) scope; handle it before key resolution.
if operator == qbtypes.FilterOperatorSearch {
return c.conditionForSearch(ctx, orgID, key, value, sb)
}
keys, warning := querybuilder.ResolveKeys(key, matches)
var warnings []string
if warning != "" {

View File

@@ -635,3 +635,37 @@ func (m *fieldMapper) existsExpressionFor(
}
return querybuilder.ExistsExpression(columns, key, tsStart, tsEnd, fieldExpression, exists)
}
// searchColumns is the single source of truth for the columns search() fans out across,
// by context; body is body_v2 when useJSONBody, else the legacy body string.
func searchColumns(fieldContext telemetrytypes.FieldContext, useJSONBody bool) []*schema.Column {
switch fieldContext {
case telemetrytypes.FieldContextLog:
return []*schema.Column{
logsV2Columns[LogsV2SeverityTextColumn],
logsV2Columns[LogsV2TraceIDColumn],
logsV2Columns[LogsV2SpanIDColumn],
}
case telemetrytypes.FieldContextBody:
if useJSONBody {
return []*schema.Column{logsV2Columns[LogsV2BodyV2Column]}
}
return []*schema.Column{logsV2Columns[LogsV2BodyColumn]}
case telemetrytypes.FieldContextAttribute:
return []*schema.Column{
logsV2Columns[LogsV2AttributesStringColumn],
logsV2Columns[LogsV2AttributesNumberColumn],
logsV2Columns[LogsV2AttributesBoolColumn],
}
case telemetrytypes.FieldContextResource:
return []*schema.Column{
logsV2Columns[LogsV2ResourcesStringColumn],
}
default:
columns := searchColumns(telemetrytypes.FieldContextLog, useJSONBody)
columns = append(columns, searchColumns(telemetrytypes.FieldContextBody, useJSONBody)...)
columns = append(columns, searchColumns(telemetrytypes.FieldContextAttribute, useJSONBody)...)
columns = append(columns, searchColumns(telemetrytypes.FieldContextResource, useJSONBody)...)
return columns
}
}

View File

@@ -0,0 +1,284 @@
package logstelemetryschema
import (
"context"
"testing"
"time"
"github.com/SigNoz/signoz/pkg/flagger"
"github.com/SigNoz/signoz/pkg/flagger/flaggertest"
"github.com/SigNoz/signoz/pkg/instrumentation/instrumentationtest"
"github.com/SigNoz/signoz/pkg/querybuilder"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/huandu/go-sqlbuilder"
"github.com/stretchr/testify/require"
)
// searchFanOut returns the WHERE fragment search() fans out to; bodyExpr differs
// between the legacy string body and the body_v2 JSON column.
func searchFanOut(bodyExpr string) string {
return "(match(LOWER(severity_text), LOWER(?)) OR match(LOWER(trace_id), LOWER(?)) OR match(LOWER(span_id), LOWER(?)) OR " +
bodyExpr + " OR " +
"(arrayExists(x -> match(LOWER(x), LOWER(?)), mapKeys(attributes_string)) OR arrayExists(x -> match(LOWER(x), LOWER(?)), mapValues(attributes_string))) OR " +
"(arrayExists(x -> match(LOWER(x), LOWER(?)), mapKeys(attributes_number)) OR arrayExists(x -> match(LOWER(x), LOWER(?)), arrayMap(x -> toString(x), mapValues(attributes_number)))) OR " +
"(arrayExists(x -> match(LOWER(x), LOWER(?)), mapKeys(attributes_bool)) OR arrayExists(x -> match(LOWER(x), LOWER(?)), arrayMap(x -> toString(x), mapValues(attributes_bool)))) OR " +
"(arrayExists(x -> match(LOWER(x), LOWER(?)), mapKeys(resources_string)) OR arrayExists(x -> match(LOWER(x), LOWER(?)), mapValues(resources_string))))"
}
// searchArgs returns v once per bound parameter search() emits — one per searchable
// column expression (currently 12).
func searchArgs(v any) []any {
const searchColumnParams = 12
args := make([]any, searchColumnParams)
for i := range args {
args[i] = v
}
return args
}
// TestFilterExprSearch covers search('term') fanning out across every searchable
// column via FilterOperatorSearch.
func TestFilterExprSearch(t *testing.T) {
releaseTime := time.Date(2024, 1, 15, 10, 0, 0, 0, time.UTC)
inWindowStart := uint64(releaseTime.Add(-5 * time.Minute).UnixNano())
inWindowEnd := uint64(releaseTime.Add(5 * time.Minute).UnixNano())
legacyBody := "match(LOWER(body), LOWER(?))"
jsonBody := "match(LOWER(toString(body_v2)), LOWER(?))"
// Single-context scope fragments (the fan-out narrowed to one context).
logScope := "(match(LOWER(severity_text), LOWER(?)) OR match(LOWER(trace_id), LOWER(?)) OR match(LOWER(span_id), LOWER(?)))"
resourceScope := "(arrayExists(x -> match(LOWER(x), LOWER(?)), mapKeys(resources_string)) OR arrayExists(x -> match(LOWER(x), LOWER(?)), mapValues(resources_string)))"
serviceNameEq := "(multiIf(resource.`service.name` IS NOT NULL, resource.`service.name`::String, mapContains(resources_string, 'service.name'), resources_string['service.name'], NULL) = ? " +
"AND multiIf(resource.`service.name` IS NOT NULL, resource.`service.name`::String, mapContains(resources_string, 'service.name'), resources_string['service.name'], NULL) IS NOT NULL)"
testCases := []struct {
name string
query string
jsonBodyEnabled bool
fullTextColumn *telemetrytypes.TelemetryFieldKey
startNs uint64
endNs uint64
shouldPass bool
expectedQuery string
expectedArgs []any
expectWarning bool
expectedErrorContains string
}{
{
name: "quoted, legacy body",
query: "search('error')",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(legacyBody),
expectedArgs: searchArgs("error"),
expectWarning: true,
},
{
name: "quoted, json body",
query: "search('error')",
jsonBodyEnabled: true,
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(jsonBody),
expectedArgs: searchArgs("error"),
expectWarning: true,
},
{
name: "bare word",
query: "search(timeout)",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(legacyBody),
expectedArgs: searchArgs("timeout"),
expectWarning: true,
},
{
name: "negated",
query: "NOT search('error')",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE NOT (" + searchFanOut(legacyBody) + ")",
expectedArgs: searchArgs("error"),
expectWarning: true,
},
{
name: "combined with field filter",
query: "search('error') AND service.name=\"api\"",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE (" + searchFanOut(legacyBody) + " AND " + serviceNameEq + ")",
expectedArgs: append(searchArgs("error"), "api"),
expectWarning: true,
},
{
// The builder caps no window; the querier's estimate gate bounds scan cost.
name: "wide window builds (estimate gate lives in querier)",
query: "search('error')",
fullTextColumn: DefaultFullTextColumn,
startNs: uint64(releaseTime.Add(-10 * time.Hour).UnixNano()),
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(legacyBody),
expectedArgs: searchArgs("error"),
expectWarning: true,
},
{
// fullTextColumn governs only bare/quoted free text, so search() must
// work when it is unset.
name: "independent of full text column",
query: "search('error')",
fullTextColumn: nil,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(legacyBody),
expectedArgs: searchArgs("error"),
expectWarning: true,
},
{
// The bare word is the literal search term; Normalize would strip "resource.".
name: "bare word with context prefix is not normalized",
query: "search(resource.deployment)",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(legacyBody),
expectedArgs: searchArgs("resource\\.deployment"),
expectWarning: true,
},
{
// Literal digits, not %v of a parsed float64 (which would scan "1e+06").
name: "numeric search term is not scientific notation",
query: "search(1000000)",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + searchFanOut(legacyBody),
expectedArgs: searchArgs("1000000"),
expectWarning: true,
},
{
name: "scoped to body, legacy",
query: "search('error', body)",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE (" + legacyBody + ")",
expectedArgs: []any{"error"},
expectWarning: true,
},
{
name: "scoped to body, json",
query: "search('error', body)",
jsonBodyEnabled: true,
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE (" + jsonBody + ")",
expectedArgs: []any{"error"},
expectWarning: true,
},
{
name: "scoped to resource (quoted scope)",
query: "search('error', 'resource')",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE (" + resourceScope + ")",
expectedArgs: []any{"error", "error"},
expectWarning: true,
},
{
name: "scoped to log fields",
query: "search('error', log)",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE " + logScope,
expectedArgs: []any{"error", "error", "error"},
expectWarning: true,
},
{
name: "scoped to multiple contexts",
query: "search('error', body, resource)",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: true,
expectedQuery: "WHERE ((" + legacyBody + ") OR (" + resourceScope + "))",
expectedArgs: []any{"error", "error", "error"},
expectWarning: true,
},
{
name: "invalid scope",
query: "search('error', 'timeout')",
fullTextColumn: DefaultFullTextColumn,
startNs: inWindowStart,
endNs: inWindowEnd,
shouldPass: false,
expectedErrorContains: "invalid search scope",
},
}
for _, tc := range testCases {
t.Run(tc.name, func(t *testing.T) {
fl := flaggertest.WithBooleanFlags(t, map[string]bool{
flagger.FeatureUseJSONBody.String(): tc.jsonBodyEnabled,
})
fm := NewFieldMapper(fl)
cb := NewConditionBuilder(fm, fl)
keys := BuildCompleteFieldKeyMap(releaseTime)
opts := querybuilder.FilterExprVisitorOpts{
Context: context.Background(),
Logger: instrumentationtest.New().Logger(),
FieldMapper: fm,
ConditionBuilder: cb,
FieldKeys: keys,
FullTextColumn: tc.fullTextColumn,
StartNs: tc.startNs,
EndNs: tc.endNs,
}
clause, err := querybuilder.PrepareWhereClause(tc.query, opts)
if !tc.shouldPass {
require.Error(t, err)
require.True(t, detailContains(err, tc.expectedErrorContains),
"error %v should contain %q", err, tc.expectedErrorContains)
return
}
require.NoError(t, err)
require.False(t, clause.IsEmpty())
sql, args := clause.WhereClause.BuildWithFlavor(sqlbuilder.ClickHouse)
require.Equal(t, tc.expectedQuery, sql)
require.Equal(t, tc.expectedArgs, args)
if tc.expectWarning {
// The visitor only flags the guard; the statement builder
// materializes it from config.
require.True(t, clause.RequiresCostGuard)
}
})
}
}

View File

@@ -157,7 +157,7 @@ func (c *conditionBuilder) ConditionFor(
sb *sqlbuilder.SelectBuilder,
) ([]string, []string, error) {
// has/hasAny/hasAll/hasToken are logs-body-only; reject for metrics.
// has/hasAny/hasAll/hasToken/search are logs-only functions; reject for metrics.
if err := querybuilder.NewFunctionUnsupportedError(operator); err != nil {
return nil, nil, err
}

View File

@@ -205,7 +205,7 @@ func (c *conditionBuilder) ConditionFor(
sb *sqlbuilder.SelectBuilder,
) ([]string, []string, error) {
// has/hasAny/hasAll/hasToken are logs-body-only; reject for traces.
// has/hasAny/hasAll/hasToken/search are logs-only functions; reject for traces.
if err := querybuilder.NewFunctionUnsupportedError(operator); err != nil {
return nil, nil, err
}

View File

@@ -35,12 +35,8 @@ func TestNewConfigFromChannels(t *testing.T) {
"email_configs": []any{map[string]any{
"send_resolved": false,
"to": "test@example.com",
"from": "alerts@example.com",
"hello": "localhost",
"smarthost": "smtp.example.com:587",
"require_tls": true,
"smarthost": "",
"html": "{{ template \"email.default.html\" . }}",
"tls_config": map[string]any{"insecure_skip_verify": false},
"threading": map[string]any{},
}},
},
@@ -63,7 +59,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"slack_configs": []any{map[string]any{
"send_resolved": true,
"api_url": "https://slack.com/api/test",
"app_url": "https://slack.com/api/chat.postMessage",
"channel": "#alerts",
"callback_id": "{{ template \"slack.default.callbackid\" . }}",
"color": "{{ if eq .Status \"firing\" }}danger{{ else }}good{{ end }}",
@@ -77,12 +72,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"title": "{{ template \"slack.default.title\" . }}",
"title_link": "{{ template \"slack.default.titlelink\" . }}",
"username": "{{ template \"slack.default.username\" . }}",
"http_config": map[string]any{
"tls_config": map[string]any{"insecure_skip_verify": false},
"follow_redirects": true,
"enable_http2": true,
"proxy_url": nil,
},
}},
},
},
@@ -104,7 +93,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"pagerduty_configs": []any{map[string]any{
"send_resolved": false,
"service_key": "test",
"url": "https://events.pagerduty.com/v2/enqueue",
"client": "{{ template \"pagerduty.default.client\" . }}",
"client_url": "{{ template \"pagerduty.default.clientURL\" . }}",
"description": "{{ template \"pagerduty.default.description\" .}}",
@@ -116,12 +104,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"num_resolved": "{{ .Alerts.Resolved | len }}",
"resolved": "{{ .Alerts.Resolved | toJson }}",
},
"http_config": map[string]any{
"tls_config": map[string]any{"insecure_skip_verify": false},
"follow_redirects": true,
"enable_http2": true,
"proxy_url": nil,
},
}},
},
},
@@ -148,7 +130,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"pagerduty_configs": []any{map[string]any{
"send_resolved": false,
"service_key": "test",
"url": "https://events.pagerduty.com/v2/enqueue",
"client": "{{ template \"pagerduty.default.client\" . }}",
"client_url": "{{ template \"pagerduty.default.clientURL\" . }}",
"description": "{{ template \"pagerduty.default.description\" .}}",
@@ -160,12 +141,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"num_resolved": "{{ .Alerts.Resolved | len }}",
"resolved": "{{ .Alerts.Resolved | toJson }}",
},
"http_config": map[string]any{
"tls_config": map[string]any{"insecure_skip_verify": false},
"follow_redirects": true,
"enable_http2": true,
"proxy_url": nil,
},
}},
},
{
@@ -173,7 +148,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"slack_configs": []any{map[string]any{
"send_resolved": true,
"api_url": "https://slack.com/api/test",
"app_url": "https://slack.com/api/chat.postMessage",
"channel": "#alerts",
"callback_id": "{{ template \"slack.default.callbackid\" . }}",
"color": "{{ if eq .Status \"firing\" }}danger{{ else }}good{{ end }}",
@@ -187,12 +161,6 @@ func TestNewConfigFromChannels(t *testing.T) {
"title": "{{ template \"slack.default.title\" . }}",
"title_link": "{{ template \"slack.default.titlelink\" . }}",
"username": "{{ template \"slack.default.username\" . }}",
"http_config": map[string]any{
"tls_config": map[string]any{"insecure_skip_verify": false},
"follow_redirects": true,
"enable_http2": true,
"proxy_url": nil,
},
}},
},
},

View File

@@ -117,6 +117,11 @@ func NewConfigFromStoreableConfig(sc *StoreableConfig) (*Config, error) {
return nil, err
}
// It must be replaced with an empty, non-nil global, upstream swaps nil for
// DefaultGlobalConfig, which would let a path that skips SetGlobalConfig pass
// validation and fail silently at delivery instead of failing fast here.
alertmanagerConfig.Global = &config.GlobalConfig{}
return &Config{
alertmanagerConfig: alertmanagerConfig,
customConfigs: customConfigs,
@@ -174,7 +179,7 @@ func newConfigFromString(s string) (*config.Config, map[string]customReceiverCon
return amConfig, customConfigs, nil
}
func newRawFromConfig(c *config.Config, customConfigs map[string]customReceiverConfigs) []byte {
func extendedReceivers(c *config.Config, customConfigs map[string]customReceiverConfigs) []*Receiver {
receivers := make([]*Receiver, len(c.Receivers))
for i := range c.Receivers {
base := c.Receivers[i]
@@ -185,7 +190,14 @@ func newRawFromConfig(c *config.Config, customConfigs map[string]customReceiverC
}
}
b, err := json.Marshal(storedConfig{Config: c, Receivers: receivers})
return receivers
}
func newRawFromConfig(c *config.Config, customConfigs map[string]customReceiverConfigs) []byte {
persistable := *c
persistable.Global = nil
b, err := json.Marshal(storedConfig{Config: &persistable, Receivers: extendedReceivers(c, customConfigs)})
if err != nil {
// Taking inspiration from the upstream. This is never expected to happen.
return []byte(fmt.Sprintf("<error creating config string: %s>", err))
@@ -206,6 +218,37 @@ func (c *Config) flush() {
c.storeableConfig.UpdatedAt = time.Now()
}
func (c *Config) Resolved() (*Config, error) {
raw, err := json.Marshal(storedConfig{Config: c.alertmanagerConfig, Receivers: extendedReceivers(c.alertmanagerConfig, c.customConfigs)})
if err != nil {
return nil, err
}
alertmanagerConfig, customConfigs, err := newConfigFromString(string(raw))
if err != nil {
return nil, err
}
if err := alertmanagerConfig.UnmarshalYAML(func(i interface{}) error { return nil }); err != nil {
return nil, err
}
storeableConfig := *c.storeableConfig
resolved := &Config{
alertmanagerConfig: alertmanagerConfig,
customConfigs: customConfigs,
storeableConfig: &storeableConfig,
}
resolved.applyNativeDefaults()
return resolved, nil
}
func (c *Config) validate() error {
_, err := c.Resolved()
return err
}
func (c *Config) CopyWithReset() (*Config, error) {
newConfig, err := NewDefaultConfig(
*c.alertmanagerConfig.Global,
@@ -271,6 +314,15 @@ func (c *Config) StoreableConfig() *StoreableConfig {
return c.storeableConfig
}
func cloneReceiver(receiver *Receiver) (*Receiver, error) {
raw, err := json.Marshal(receiver)
if err != nil {
return nil, err
}
return NewReceiver(string(raw))
}
func (c *Config) CreateReceiver(receiver *Receiver) error {
// check that receiver name is not already used
for _, existingReceiver := range c.alertmanagerConfig.Receivers {
@@ -279,16 +331,21 @@ func (c *Config) CreateReceiver(receiver *Receiver) error {
}
}
route, err := NewRouteFromReceiver(receiver)
owned, err := cloneReceiver(receiver)
if err != nil {
return err
}
route, err := NewRouteFromReceiver(owned)
if err != nil {
return err
}
c.alertmanagerConfig.Route.Routes = append(c.alertmanagerConfig.Route.Routes, route)
c.alertmanagerConfig.Receivers = append(c.alertmanagerConfig.Receivers, *receiver.Receiver)
c.setCustomConfigs(receiver)
c.alertmanagerConfig.Receivers = append(c.alertmanagerConfig.Receivers, *owned.Receiver)
c.setCustomConfigs(owned)
if err := c.alertmanagerConfig.UnmarshalYAML(func(i interface{}) error { return nil }); err != nil {
if err := c.validate(); err != nil {
return err
}
c.applyNativeDefaults()
@@ -313,16 +370,21 @@ func (c *Config) GetReceiver(name string) (*Receiver, error) {
}
func (c *Config) UpdateReceiver(receiver *Receiver) error {
owned, err := cloneReceiver(receiver)
if err != nil {
return err
}
// find and update receiver
for i, existingReceiver := range c.alertmanagerConfig.Receivers {
if existingReceiver.Name == receiver.Name {
c.alertmanagerConfig.Receivers[i] = *receiver.Receiver
c.setCustomConfigs(receiver)
if existingReceiver.Name == owned.Name {
c.alertmanagerConfig.Receivers[i] = *owned.Receiver
c.setCustomConfigs(owned)
break
}
}
if err := c.alertmanagerConfig.UnmarshalYAML(func(i interface{}) error { return nil }); err != nil {
if err := c.validate(); err != nil {
return err
}
c.applyNativeDefaults()

View File

@@ -330,6 +330,150 @@ func TestSetGlobalConfigPreservesSMTPRequireTLS(t *testing.T) {
}
}
func newSMTPGlobalConfig() GlobalConfig {
return GlobalConfig{
SMTPFrom: "alerts@example.com",
SMTPHello: "example.com",
SMTPSmarthost: config.HostPort{Host: "smtp.sendgrid.net", Port: "587"},
SMTPAuthUsername: "apikey",
SMTPAuthPassword: "operator-secret",
SMTPRequireTLS: true,
}
}
func newEmailTestConfig(t *testing.T) *Config {
t.Helper()
cfg, err := NewDefaultConfig(
newSMTPGlobalConfig(),
RouteConfig{GroupInterval: time.Minute, GroupWait: time.Minute, RepeatInterval: time.Minute},
"1",
)
require.NoError(t, err)
receiver, err := NewReceiver(`{"name":"email-receiver","email_configs":[{"to":"team@example.com"}]}`)
require.NoError(t, err)
require.NoError(t, cfg.CreateReceiver(receiver))
return cfg
}
func TestStoreableConfigCarriesNoSMTPSettings(t *testing.T) {
cfg := newEmailTestConfig(t)
raw := cfg.StoreableConfig().Config
assert.NotContains(t, raw, "operator-secret")
assert.NotContains(t, raw, "smtp.sendgrid.net")
assert.NotContains(t, raw, "apikey")
assert.NotContains(t, raw, "alerts@example.com")
assert.Equal(t, "operator-secret", string(cfg.alertmanagerConfig.Global.SMTPAuthPassword))
}
func TestStoreableConfigCarriesNoGlobal(t *testing.T) {
cfg := newEmailTestConfig(t)
stored := map[string]json.RawMessage{}
require.NoError(t, json.Unmarshal([]byte(cfg.StoreableConfig().Config), &stored))
assert.NotContains(t, stored, "global")
}
func TestSetGlobalConfigDoesNotChangeStoreableHash(t *testing.T) {
cfg := newEmailTestConfig(t)
hash := cfg.StoreableConfig().Hash
require.NoError(t, cfg.SetGlobalConfig(GlobalConfig{SMTPSmarthost: config.HostPort{Host: "smtp.other.net", Port: "2525"}, SMTPAuthPassword: "rotated-secret"}))
assert.Equal(t, hash, cfg.StoreableConfig().Hash)
assert.NotContains(t, cfg.StoreableConfig().Config, "rotated-secret")
}
func TestNewConfigFromStoreableConfigDiscardsStoredGlobal(t *testing.T) {
stored := &StoreableConfig{
Config: `{"global":{"resolve_timeout":"5m","smtp_smarthost":"email-smtp.us-east-1.amazonaws.com:587","smtp_auth_password":"old-secret","slack_api_url":"https://hooks.slack.com/services/T/B/X"},"route":{"receiver":"default-receiver"},"receivers":[{"name":"default-receiver"}]}`,
OrgID: "1",
}
cfg, err := NewConfigFromStoreableConfig(stored)
require.NoError(t, err)
assert.Equal(t, &config.GlobalConfig{}, cfg.alertmanagerConfig.Global)
}
func TestResolvedFillsEmailTransportFromGlobal(t *testing.T) {
cfg := newEmailTestConfig(t)
resolved, err := cfg.Resolved()
require.NoError(t, err)
receiver, err := resolved.GetReceiver("email-receiver")
require.NoError(t, err)
require.Len(t, receiver.EmailConfigs, 1)
got := receiver.EmailConfigs[0]
assert.Equal(t, "team@example.com", got.To)
assert.Equal(t, "smtp.sendgrid.net:587", got.Smarthost.String())
assert.Equal(t, "alerts@example.com", got.From)
assert.Equal(t, "apikey", got.AuthUsername)
assert.Equal(t, "operator-secret", string(got.AuthPassword))
require.NotNil(t, got.RequireTLS)
assert.True(t, *got.RequireTLS)
stored, err := cfg.GetReceiver("email-receiver")
require.NoError(t, err)
assert.Empty(t, stored.EmailConfigs[0].Smarthost.String())
assert.NotContains(t, cfg.StoreableConfig().Config, "operator-secret")
}
func TestStaleStoredSMTPSettingsAreReplacedOnLoad(t *testing.T) {
stored := &StoreableConfig{
Config: `{"global":{"resolve_timeout":"5m","smtp_from":"old@example.com","smtp_hello":"localhost","smtp_smarthost":"email-smtp.us-east-1.amazonaws.com:587","smtp_auth_username":"old-user","smtp_auth_password":"old-secret","smtp_require_tls":true},"route":{"receiver":"default-receiver","group_by":["ruleId"],"routes":[{"receiver":"email-receiver","continue":true,"matchers":["ruleId=~\"-1\""]}],"group_wait":"30s","group_interval":"5m","repeat_interval":"4h"},"receivers":[{"name":"default-receiver"},{"name":"email-receiver","email_configs":[{"send_resolved":false,"to":"team@example.com","from":"old@example.com","hello":"localhost","smarthost":"email-smtp.us-east-1.amazonaws.com:587","auth_username":"old-user","auth_password":"old-secret","require_tls":true}]}]}`,
OrgID: "1",
}
cfg, err := NewConfigFromStoreableConfig(stored)
require.NoError(t, err)
loaded, err := cfg.GetReceiver("email-receiver")
require.NoError(t, err)
require.Len(t, loaded.EmailConfigs, 1)
assert.Empty(t, loaded.EmailConfigs[0].Smarthost.String())
assert.Empty(t, string(loaded.EmailConfigs[0].AuthPassword))
require.NoError(t, cfg.SetGlobalConfig(newSMTPGlobalConfig()))
resolved, err := cfg.Resolved()
require.NoError(t, err)
receiver, err := resolved.GetReceiver("email-receiver")
require.NoError(t, err)
got := receiver.EmailConfigs[0]
assert.Equal(t, "smtp.sendgrid.net:587", got.Smarthost.String())
assert.Equal(t, "operator-secret", string(got.AuthPassword))
assert.Equal(t, "alerts@example.com", got.From)
assert.NotContains(t, cfg.StoreableConfig().Config, "old-secret")
assert.NotContains(t, cfg.StoreableConfig().Config, "amazonaws.com")
assert.NotContains(t, cfg.StoreableConfig().Config, "operator-secret")
}
func TestCreateReceiverDoesNotMutateCaller(t *testing.T) {
cfg := newEmailTestConfig(t)
resolved, err := cfg.Resolved()
require.NoError(t, err)
receiver, err := resolved.GetReceiver("email-receiver")
require.NoError(t, err)
require.Equal(t, "smtp.sendgrid.net:587", receiver.EmailConfigs[0].Smarthost.String())
throwaway, err := cfg.CopyWithReset()
require.NoError(t, err)
require.NoError(t, throwaway.CreateReceiver(receiver))
assert.Equal(t, "smtp.sendgrid.net:587", receiver.EmailConfigs[0].Smarthost.String())
assert.Equal(t, "operator-secret", string(receiver.EmailConfigs[0].AuthPassword))
}
// Round-trip: create → serialize → reload → GetReceiver still has the configs.
func TestConfigPreservesGoogleChatConfigs(t *testing.T) {
webhookURL, err := url.Parse("https://chat.googleapis.com/v1/spaces/test/messages")

View File

@@ -37,6 +37,7 @@ func NewReceiver(input string) (*Receiver, error) {
if err != nil {
return nil, err
}
stripEmailTransport(withDefaults)
receiver.Receiver = withDefaults
// Extend this block when adding another native notifier type.
@@ -53,6 +54,23 @@ func NewReceiver(input string) (*Receiver, error) {
return receiver, nil
}
func stripEmailTransport(base *config.Receiver) {
for _, ec := range base.EmailConfigs {
ec.From = ""
ec.Hello = ""
ec.Smarthost = config.HostPort{}
ec.AuthUsername = ""
ec.AuthPassword = ""
ec.AuthPasswordFile = ""
ec.AuthSecret = ""
ec.AuthSecretFile = ""
ec.AuthIdentity = ""
ec.RequireTLS = nil
ec.TLSConfig = nil
ec.ForceImplicitTLS = nil
}
}
func defaultedBaseReceiver(base *config.Receiver) (*config.Receiver, error) {
bytes, err := yaml.Marshal(base)
if err != nil {
@@ -102,7 +120,12 @@ func TestReceiver(ctx context.Context, receiver *Receiver, receiverIntegrationsF
return err
}
defaultedReceiver, err := testConfig.GetReceiver(receiver.Name)
resolvedConfig, err := testConfig.Resolved()
if err != nil {
return err
}
defaultedReceiver, err := resolvedConfig.GetReceiver(receiver.Name)
if err != nil {
return err
}

View File

@@ -46,6 +46,31 @@ func TestNewReceiver(t *testing.T) {
}
}
func TestNewReceiverStripsEmailTransport(t *testing.T) {
receiver, err := NewReceiver(`{"name":"email","email_configs":[{"to":"team@example.com","from":"attacker@example.com","hello":"example.com","smarthost":"smtp.example.com:587","auth_username":"user","auth_password":"supersecret","auth_secret":"alsosecret","auth_identity":"id","require_tls":false,"headers":{"Subject":"custom"}}]}`)
require.NoError(t, err)
require.Len(t, receiver.EmailConfigs, 1)
got := receiver.EmailConfigs[0]
assert.Equal(t, "team@example.com", got.To)
assert.Equal(t, map[string]string{"Subject": "custom"}, got.Headers)
assert.Empty(t, got.From)
assert.Empty(t, got.Hello)
assert.Empty(t, got.Smarthost.String())
assert.Empty(t, got.AuthUsername)
assert.Empty(t, string(got.AuthPassword))
assert.Empty(t, string(got.AuthSecret))
assert.Empty(t, got.AuthIdentity)
assert.Nil(t, got.RequireTLS)
assert.Nil(t, got.TLSConfig)
bytes, err := json.Marshal(receiver)
require.NoError(t, err)
assert.NotContains(t, string(bytes), "supersecret")
assert.NotContains(t, string(bytes), "smtp.example.com")
}
// Omitted fields fall back to DefaultGoogleChatReceiverConfig.
func TestNewReceiverGoogleChatAppliesDefaults(t *testing.T) {
receiver, err := NewReceiver(`{"name":"googlechat","googlechat_configs":[{"webhook_url":"https://chat.googleapis.com/v1/spaces/test/messages"}]}`)

View File

@@ -179,7 +179,6 @@ func (d *v1Decoder) collectV1QueryEnvelopes(widget map[string]any, panelKind Pan
normalizePreV5GroupBy(q)
normalizePreV5PageSize(q, rowLimitPanel)
normalizeQueryLimit(q)
normalizeQueryOffset(q)
if needsAggregation {
ensureDefaultAggregation(q)
}
@@ -199,7 +198,6 @@ func (d *v1Decoder) collectV1QueryEnvelopes(widget map[string]any, panelKind Pan
assignMissingFormulaNames(formulas)
for _, f := range formulas {
normalizePreV5QueryData(f, widgetType, panelKind)
normalizeQueryLimit(f)
name := d.readString(f, "queryName")
env := qb.WrapInV5Envelope(name, f, string(qb.QueryTypeFormula.StringValue()))
backfillFormulaFields(env, f)
@@ -221,8 +219,6 @@ func (d *v1Decoder) collectV1QueryEnvelopes(widget map[string]any, panelKind Pan
normalizePreV5QueryData(op, widgetType, panelKind)
normalizePreV5GroupBy(op)
normalizeOrderByKeys(op)
normalizeQueryLimit(op)
normalizeQueryOffset(op)
name := d.readString(op, "queryName")
out = append(out, traceOperatorEnvelope(name, expression, op))
}

View File

@@ -618,32 +618,15 @@ func normalizePreV5PageSize(query map[string]any, rowLimitPanel bool) {
}
}
// normalizeQueryLimit coerces limit to the int the v5 decode expects: v1 stored it
// as a string ("5") or float, both of which fail the typed decode. An unparseable
// value or one above the v5 maximum (MaxQueryLimit) is dropped, leaving the query
// unlimited (the field is optional).
// normalizeQueryLimit drops a limit above the v5 maximum (MaxQueryLimit); v1 allowed
// larger/unbounded limits, and an over-max value fails validation. Removing it leaves
// the query unlimited (the field is optional).
func normalizeQueryLimit(query map[string]any) {
if query["limit"] == nil {
return
}
limit, ok := coerceFloat(query["limit"])
if !ok || limit > qb.MaxQueryLimit {
delete(query, "limit")
return
}
query["limit"] = int(limit)
}
// normalizeQueryOffset coerces offset to the int the v5 decode expects; v1 could
// store it as a string. An unparseable value is dropped (offset defaults to 0).
func normalizeQueryOffset(query map[string]any) {
if query["offset"] == nil {
return
}
offset, ok := coerceFloat(query["offset"])
if !ok {
delete(query, "offset")
return
}
query["offset"] = int(offset)
if limit > qb.MaxQueryLimit {
delete(query, "limit")
}
}

View File

@@ -118,6 +118,10 @@ const (
FilterOperatorHasToken
FilterOperatorHasAny
FilterOperatorHasAll
// FilterOperatorSearch backs search('term'): keyless, fanned out by the condition
// builder across every searchable column.
FilterOperatorSearch
)
var operatorInverseMapping = map[FilterOperator]FilterOperator{
@@ -186,7 +190,8 @@ func (f FilterOperator) IsNegativeOperator() bool {
FilterOperatorIn,
FilterOperatorExists,
FilterOperatorRegexp,
FilterOperatorContains:
FilterOperatorContains,
FilterOperatorSearch:
return false
}
return true
@@ -243,10 +248,11 @@ func (f FilterOperator) IsArrayFunctionOperator() bool {
}
// IsFunctionOperator reports whether the operator is a query function
// (has/hasAny/hasAll/hasToken); these apply to the logs body column only.
// (has/hasAny/hasAll/hasToken/search) — logs-only, and skipped by the
// resource-fingerprint builder.
func (f FilterOperator) IsFunctionOperator() bool {
switch f {
case FilterOperatorHas, FilterOperatorHasAny, FilterOperatorHasAll, FilterOperatorHasToken:
case FilterOperatorHas, FilterOperatorHasAny, FilterOperatorHasAll, FilterOperatorHasToken, FilterOperatorSearch:
return true
default:
return false
@@ -265,6 +271,8 @@ func (f FilterOperator) FunctionName() string {
return "hasAll"
case FilterOperatorHasToken:
return "hasToken"
case FilterOperatorSearch:
return "search"
default:
return ""
}

View File

@@ -57,6 +57,12 @@ type Statement struct {
Args []any
Warnings []string
WarningsDocURL string
CostGuard *CostGuard
}
type CostGuard struct {
Warning string
MaxScanRows int64
}
// StatementBuilder builds the query.

View File

@@ -79,6 +79,14 @@ var (
}
)
// FieldContextFromText resolves a context word with the same aliases as key parsing
// ("tag" -> attribute). ok is false for an unknown word, so callers can reject it
// rather than get unspecified.
func FieldContextFromText(text string) (FieldContext, bool) {
fc, ok := fieldContexts[strings.ToLower(strings.TrimSpace(text))]
return fc, ok
}
// UnmarshalJSON implements the json.Unmarshaler interface.
func (f *FieldContext) UnmarshalJSON(data []byte) error {
var str string

View File

@@ -412,10 +412,15 @@ func (v *variableReplacementVisitor) VisitFunctionCall(ctx *grammar.FunctionCall
}
func (v *variableReplacementVisitor) VisitSearchCall(ctx *grammar.SearchCallContext) any {
if ctx.FunctionParamList() == nil {
if ctx.ValueList() == nil {
return "search()"
}
return "search(" + v.Visit(ctx.FunctionParamList()).(string) + ")"
// VisitValueList already parenthesizes the args and propagates the __all__ marker.
result := v.Visit(ctx.ValueList()).(string)
if result == specialSkipMarker {
return specialSkipMarker
}
return "search" + result
}
func (v *variableReplacementVisitor) VisitFunctionParamList(ctx *grammar.FunctionParamListContext) any {

View File

@@ -24,6 +24,7 @@ pytest_plugins = [
"fixtures.keycloak",
"fixtures.idp",
"fixtures.notification_channel",
"fixtures.maildev",
"fixtures.alerts",
"fixtures.cloudintegrations",
"fixtures.jsontypes",

View File

@@ -1,11 +1,13 @@
import base64
import json
import re
import time
import urllib.parse
import uuid
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
from http import HTTPStatus
from urllib.parse import urlparse
import pytest
import requests
@@ -15,6 +17,7 @@ from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.fs import get_testdata_file_path
from fixtures.logger import setup_logger
from fixtures.logs import Logs
from fixtures.maildev import get_all_mails, verify_email_received
from fixtures.metrics import Metrics
from fixtures.traces import Traces
@@ -311,3 +314,145 @@ def update_rule_channel_name(rule_data: dict, channel_name: str):
# loop over all the sepcs and update the channels
for spec in thresholds["spec"]:
spec["channels"] = [channel_name]
def _is_json_subset(subset, superset) -> bool:
"""Check if subset is contained within superset recursively.
- For dicts: all keys in subset must exist in superset with matching values
- For lists: all items in subset must be present in superset
- For scalars: exact equality
"""
if isinstance(subset, dict):
if not isinstance(superset, dict):
return False
return all(key in superset and _is_json_subset(value, superset[key]) for key, value in subset.items())
if isinstance(subset, list):
if not isinstance(superset, list):
return False
return all(any(_is_json_subset(sub_item, sup_item) for sup_item in superset) for sub_item in subset)
if isinstance(subset, re.Pattern):
return isinstance(superset, str) and subset.search(superset) is not None
return subset == superset
def verify_webhook_notification_expectation(
notification_channel: types.TestContainerDocker,
validation_data: dict,
) -> bool:
"""Check if wiremock received a request at the given path
whose JSON body is a superset of the expected json_body."""
path = validation_data["path"]
json_body = validation_data["json_body"]
url = notification_channel.host_configs["8080"].get("__admin/requests/find")
try:
res = requests.post(url, json={"method": "POST", "url": path}, timeout=10)
except requests.exceptions.RequestException:
return False
if res.status_code != HTTPStatus.OK:
return False
for req in res.json()["requests"]:
body = json.loads(base64.b64decode(req["bodyAsBase64"]).decode("utf-8"))
if _is_json_subset(json_body, body):
return True
return False
def _check_notification_validation(
validation: types.NotificationValidation,
notification_channel: types.TestContainerDocker,
maildev: types.TestContainerDocker,
) -> bool:
"""Dispatch a single validation check to the appropriate verifier."""
if validation.destination_type == "webhook":
return verify_webhook_notification_expectation(notification_channel, validation.validation_data)
if validation.destination_type == "email":
return verify_email_received(maildev, validation.validation_data)
raise ValueError(f"Invalid destination type: {validation.destination_type}")
def verify_notification_expectation(
notification_channel: types.TestContainerDocker,
maildev: types.TestContainerDocker,
expected_notification: types.AMNotificationExpectation,
) -> bool:
"""Poll for expected notifications across webhook and email channels."""
time_to_wait = datetime.now() + timedelta(seconds=expected_notification.wait_time_seconds)
while datetime.now() < time_to_wait:
all_found = all(_check_notification_validation(v, notification_channel, maildev) for v in expected_notification.notification_validations)
if expected_notification.should_notify and all_found:
logger.info("All expected notifications found")
return True
time.sleep(1)
# Timeout reached
if not expected_notification.should_notify:
# Verify no notifications were received
for validation in expected_notification.notification_validations:
found = _check_notification_validation(validation, notification_channel, maildev)
assert not found, f"Expected no notification but found one for {validation.destination_type} with data {validation.validation_data}"
logger.info("No notifications found, as expected")
return True
missing = [v for v in expected_notification.notification_validations if not _check_notification_validation(v, notification_channel, maildev)]
assert len(missing) == 0, f"Expected all notifications to be found but missing: {missing}, received: {_received_notifications(notification_channel, maildev, missing)}"
return True
def _received_notifications(
notification_channel: types.TestContainerDocker,
maildev: types.TestContainerDocker,
missing: list[types.NotificationValidation],
) -> dict:
received = {}
if any(v.destination_type == "webhook" for v in missing):
webhook_bodies = []
for validation in missing:
if validation.destination_type != "webhook":
continue
url = notification_channel.host_configs["8080"].get("__admin/requests/find")
try:
res = requests.post(url, json={"method": "POST", "url": validation.validation_data["path"]}, timeout=10)
webhook_bodies.extend(json.loads(base64.b64decode(req["bodyAsBase64"]).decode("utf-8")) for req in res.json()["requests"])
except requests.exceptions.RequestException as exc:
webhook_bodies.append(f"<failed to fetch wiremock journal: {exc}>")
received["webhook"] = webhook_bodies
if any(v.destination_type == "email" for v in missing):
received["email"] = get_all_mails(maildev)
return received
def update_raw_channel_config(
channel_config: dict,
channel_name: str,
notification_channel: types.TestContainerDocker,
) -> dict:
"""
Updates the channel config to point to the given wiremock
notification_channel container to receive notifications.
"""
config = channel_config.copy()
config["name"] = channel_name
url_field_map = {
"slack_configs": "api_url",
"msteamsv2_configs": "webhook_url",
"webhook_configs": "url",
"pagerduty_configs": "url",
"opsgenie_configs": "api_url",
}
for config_key, url_field in url_field_map.items():
if config_key in config:
for entry in config[config_key]:
if url_field in entry:
original_url = entry[url_field]
path = urlparse(original_url).path
entry[url_field] = notification_channel.container_configs["8080"].get(path)
return config

View File

@@ -77,7 +77,7 @@ def register_admin(
timeout=5,
)
assert response.status_code == HTTPStatus.OK
assert response.status_code == HTTPStatus.OK, f"failed to register admin: {response.status_code} {response.text}"
return types.Operation(name="create_user_admin")

View File

@@ -125,13 +125,19 @@ def gateway(
@pytest.fixture(name="make_http_mocks", scope="function")
def make_http_mocks() -> Callable[[types.TestContainerDocker, list[Mapping]], None]:
mocked_containers = []
def _make_http_mocks(container: types.TestContainerDocker, mappings: list[Mapping]) -> None:
Config.base_url = container.host_configs["8080"].get("/__admin")
for mapping in mappings:
Mappings.create_mapping(mapping=mapping)
mocked_containers.append(container)
yield _make_http_mocks
Mappings.delete_all_mappings()
Requests.reset_request_journal()
for container in mocked_containers:
Config.base_url = container.host_configs["8080"].get("/__admin")
Mappings.delete_all_mappings()
Requests.reset_request_journal()

143
tests/fixtures/maildev.py vendored Normal file
View File

@@ -0,0 +1,143 @@
import re
from http import HTTPStatus
import docker
import docker.errors
import pytest
import requests
from testcontainers.core.container import DockerContainer, Network
from fixtures import reuse, types
from fixtures.logger import setup_logger
logger = setup_logger(__name__)
MAILDEV_INCOMING_USER = "apikey"
MAILDEV_INCOMING_PASS = "integration-smtp-secret"
SMTP_TEST_FROM = "alertmanager@integration.test"
OLD_PROVIDER_SMTP_PASS = "old-provider-smtp-secret"
NEW_PROVIDER_SMTP_PASS = "new-provider-smtp-secret"
def signoz_smtp_env(maildev: "types.TestContainerDocker", password: str = MAILDEV_INCOMING_PASS) -> dict:
return {
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_SMTP__SMARTHOST": f"{maildev.container_configs['1025'].address}:{maildev.container_configs['1025'].port}",
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_SMTP__FROM": SMTP_TEST_FROM,
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_SMTP__AUTH__USERNAME": MAILDEV_INCOMING_USER,
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_SMTP__AUTH__PASSWORD": password,
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_SMTP__REQUIRE__TLS": "false",
}
def create_maildev(
network: Network,
request: pytest.FixtureRequest,
pytestconfig: pytest.Config,
cache_key: str = "maildev",
incoming_user: str = MAILDEV_INCOMING_USER,
incoming_pass: str = MAILDEV_INCOMING_PASS,
) -> types.TestContainerDocker:
def create() -> types.TestContainerDocker:
container = DockerContainer(image="maildev/maildev:2.2.1")
container.with_env("MAILDEV_INCOMING_USER", incoming_user)
container.with_env("MAILDEV_INCOMING_PASS", incoming_pass)
container.with_exposed_ports(1025, 1080)
container.with_network(network=network)
container.start()
return types.TestContainerDocker(
id=container.get_wrapped_container().id,
host_configs={
"1025": types.TestContainerUrlConfig(
scheme="smtp",
address=container.get_container_host_ip(),
port=container.get_exposed_port(1025),
),
"1080": types.TestContainerUrlConfig(
scheme="http",
address=container.get_container_host_ip(),
port=container.get_exposed_port(1080),
),
},
container_configs={
"1025": types.TestContainerUrlConfig(
scheme="smtp",
address=container.get_wrapped_container().name,
port=1025,
),
"1080": types.TestContainerUrlConfig(
scheme="http",
address=container.get_wrapped_container().name,
port=1080,
),
},
)
def delete(container: types.TestContainerDocker):
client = docker.from_env()
try:
client.containers.get(container_id=container.id).stop()
client.containers.get(container_id=container.id).remove(v=True)
except docker.errors.NotFound:
logger.info(
"Skipping removal of MailDev, MailDev(%s) not found. Maybe it was manually removed?",
{"id": container.id},
)
def restore(cache: dict) -> types.TestContainerDocker:
return types.TestContainerDocker.from_cache(cache)
return reuse.wrap(
request,
pytestconfig,
cache_key,
lambda: types.TestContainerDocker(id="", host_configs={}, container_configs={}),
create,
delete,
restore,
)
@pytest.fixture(name="maildev", scope="package")
def maildev(network: Network, request: pytest.FixtureRequest, pytestconfig: pytest.Config) -> types.TestContainerDocker:
return create_maildev(network, request, pytestconfig)
def get_all_mails(_maildev: types.TestContainerDocker) -> list[dict]:
url = _maildev.host_configs["1080"].get("/email")
response = requests.get(url, timeout=5)
assert response.status_code == HTTPStatus.OK, f"Failed to fetch emails from MailDev, status code: {response.status_code}, response: {response.text}"
def addresses(entries: list[dict]) -> str:
return ",".join(sorted(entry.get("address", "") for entry in entries))
return [
{
"subject": email.get("subject", ""),
"html": email.get("html", ""),
"text": email.get("text", ""),
"from": addresses(email.get("from", [])),
"to": addresses(email.get("to", [])),
}
for email in response.json()
]
def verify_email_received(_maildev: types.TestContainerDocker, filters: dict) -> bool:
def matches(expected, actual: str) -> bool:
if isinstance(expected, re.Pattern):
return expected.search(actual) is not None
return expected == actual
for email in get_all_mails(_maildev):
if all(key in email and matches(filter_value, email[key]) for key, filter_value in filters.items()):
return True
return False
def delete_all_mails(_maildev: types.TestContainerDocker) -> None:
url = _maildev.host_configs["1080"].get("/email/all")
response = requests.delete(url, timeout=5)
assert response.status_code == HTTPStatus.OK, f"Failed to delete emails from MailDev, status code: {response.status_code}, response: {response.text}"

View File

@@ -1,3 +1,6 @@
# pylint: disable=line-too-long
import json
import time
from collections.abc import Callable
from http import HTTPStatus
@@ -11,10 +14,116 @@ from wiremock.testing.testcontainer import WireMockContainer
from fixtures import reuse, types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.logger import setup_logger
from fixtures.maildev import MAILDEV_INCOMING_PASS, SMTP_TEST_FROM
logger = setup_logger(__name__)
EMAIL_TRANSPORT_KEYS = [
"from",
"hello",
"smarthost",
"auth_username",
"auth_password",
"auth_password_file",
"auth_secret",
"auth_secret_file",
"auth_identity",
"require_tls",
"tls_config",
"force_implicit_tls",
]
def assert_email_channel_payload_clean(payload: str) -> None:
receiver = json.loads(payload)
for email_config in receiver["email_configs"]:
transport_keys = set(email_config.keys()) & set(EMAIL_TRANSPORT_KEYS)
transport_keys -= {"smarthost"} if email_config.get("smarthost", "") == "" else set()
assert not transport_keys, f"email channel payload carries transport keys {transport_keys}: {payload}"
assert MAILDEV_INCOMING_PASS not in payload
assert SMTP_TEST_FROM not in payload
"""
Default notification channel configs shared across alertmanager tests.
"""
slack_default_config = {
# channel name configured on runtime
"slack_configs": [
{
"api_url": "services/TEAM_ID/BOT_ID/TOKEN_ID", # base_url configured on runtime
"title": '[{{ .Status | toUpper }}{{ if eq .Status "firing" }}:{{ .Alerts.Firing | len }}{{ end }}] {{ .CommonLabels.alertname }} for {{ .CommonLabels.job }}\n {{- if gt (len .CommonLabels) (len .GroupLabels) -}}\n {{" "}}(\n {{- with .CommonLabels.Remove .GroupLabels.Names }}\n {{- range $index, $label := .SortedPairs -}}\n {{ if $index }}, {{ end }}\n {{- $label.Name }}="{{ $label.Value -}}"\n {{- end }}\n {{- end -}}\n )\n {{- end }}',
"text": '{{ range .Alerts -}}\r\n *Alert:* {{ .Labels.alertname }}{{ if .Labels.severity }} - {{ .Labels.severity }}{{ end }}\r\n\r\n *Summary:* {{ .Annotations.summary }}\r\n *Description:* {{ .Annotations.description }}\r\n *RelatedLogs:* {{ if gt (len .Annotations.related_logs) 0 -}} View in <{{ .Annotations.related_logs }}|logs explorer> {{- end}}\r\n *RelatedTraces:* {{ if gt (len .Annotations.related_traces) 0 -}} View in <{{ .Annotations.related_traces }}|traces explorer> {{- end}}\r\n\r\n *Details:*\r\n {{ range .Labels.SortedPairs -}}\r\n {{- if ne .Name "ruleId" -}}\r\n \u2022 *{{ .Name }}:* {{ .Value }}\r\n {{ end -}}\r\n {{ end -}}\r\n{{ end }}',
}
],
}
msteams_default_config = {
"msteamsv2_configs": [
{
"webhook_url": "msteams/webhook_url", # base_url configured on runtime
"title": '[{{ .Status | toUpper }}{{ if eq .Status "firing" }}:{{ .Alerts.Firing | len }}{{ end }}] {{ .CommonLabels.alertname }} for {{ .CommonLabels.job }}\n {{- if gt (len .CommonLabels) (len .GroupLabels) -}}\n {{" "}}(\n {{- with .CommonLabels.Remove .GroupLabels.Names }}\n {{- range $index, $label := .SortedPairs -}}\n {{ if $index }}, {{ end }}\n {{- $label.Name }}="{{ $label.Value -}}"\n {{- end }}\n {{- end -}}\n )\n {{- end }}',
"text": '{{ range .Alerts -}}\r\n *Alert:* {{ .Labels.alertname }}{{ if .Labels.severity }} - {{ .Labels.severity }}{{ end }}\r\n\r\n *Summary:* {{ .Annotations.summary }}\r\n *Description:* {{ .Annotations.description }}\r\n *RelatedLogs:* {{ if gt (len .Annotations.related_logs) 0 -}} View in <{{ .Annotations.related_logs }}|logs explorer> {{- end}}\r\n *RelatedTraces:* {{ if gt (len .Annotations.related_traces) 0 -}} View in <{{ .Annotations.related_traces }}|traces explorer> {{- end}}\r\n\r\n *Details:*\r\n {{ range .Labels.SortedPairs -}}\r\n {{- if ne .Name "ruleId" -}}\r\n \u2022 *{{ .Name }}:* {{ .Value }}\r\n {{ end -}}\r\n {{ end -}}\r\n{{ end }}',
}
],
}
pagerduty_default_config = {
"pagerduty_configs": [
{
"routing_key": "PagerDutyRoutingKey",
"url": "v2/enqueue", # base_url configured on runtime
"client": "SigNoz Alert Manager",
"client_url": "https://enter-signoz-host-n-port-here/alerts",
"description": '[{{ .Status | toUpper }}{{ if eq .Status "firing" }}:{{ .Alerts.Firing | len }}{{ end }}] {{ .CommonLabels.alertname }} for {{ .CommonLabels.job }}\n\t{{- if gt (len .CommonLabels) (len .GroupLabels) -}}\n\t {{" "}}(\n\t {{- with .CommonLabels.Remove .GroupLabels.Names }}\n\t\t{{- range $index, $label := .SortedPairs -}}\n\t\t {{ if $index }}, {{ end }}\n\t\t {{- $label.Name }}="{{ $label.Value -}}"\n\t\t{{- end }}\n\t {{- end -}}\n\t )\n\t{{- end }}',
"details": {
"firing": '{{ template "pagerduty.default.instances" .Alerts.Firing }}',
"num_firing": "{{ .Alerts.Firing | len }}",
"num_resolved": "{{ .Alerts.Resolved | len }}",
"resolved": '{{ template "pagerduty.default.instances" .Alerts.Resolved }}',
},
"source": "SigNoz Alert Manager",
"severity": "{{ (index .Alerts 0).Labels.severity }}",
}
],
}
opsgenie_default_config = {
"opsgenie_configs": [
{
"api_key": "OpsGenieAPIKey",
"api_url": "/", # base_url configured on runtime
"description": '{{ if gt (len .Alerts.Firing) 0 -}}\r\n\tAlerts Firing:\r\n\t{{ range .Alerts.Firing }}\r\n\t - Message: {{ .Annotations.description }}\r\n\tLabels:\r\n\t{{ range .Labels.SortedPairs -}}\r\n\t\t{{- if ne .Name "ruleId" }} - {{ .Name }} = {{ .Value }}\r\n\t{{ end -}}\r\n\t{{- end }} Annotations:\r\n\t{{ range .Annotations.SortedPairs }} - {{ .Name }} = {{ .Value }}\r\n\t{{ end }} Source: {{ .GeneratorURL }}\r\n\t{{ end }}\r\n{{- end }}\r\n{{ if gt (len .Alerts.Resolved) 0 -}}\r\n\tAlerts Resolved:\r\n\t{{ range .Alerts.Resolved }}\r\n\t - Message: {{ .Annotations.description }}\r\n\tLabels:\r\n\t{{ range .Labels.SortedPairs -}}\r\n\t\t{{- if ne .Name "ruleId" }} - {{ .Name }} = {{ .Value }}\r\n\t{{ end -}}\r\n\t{{- end }} Annotations:\r\n\t{{ range .Annotations.SortedPairs }} - {{ .Name }} = {{ .Value }}\r\n\t{{ end }} Source: {{ .GeneratorURL }}\r\n\t{{ end }}\r\n{{- end }}',
"priority": '{{ if eq (index .Alerts 0).Labels.severity "critical" }}P1{{ else if eq (index .Alerts 0).Labels.severity "warning" }}P2{{ else if eq (index .Alerts 0).Labels.severity "info" }}P3{{ else }}P4{{ end }}',
"message": "{{ .CommonLabels.alertname }}",
"details": {},
}
],
}
webhook_default_config = {
"webhook_configs": [
{
"url": "webhook/webhook_url", # base_url configured on runtime
}
],
}
email_default_config = {
"email_configs": [
{
"to": "test@example.com",
"html": '<html><body>{{ range .Alerts -}}\r\n *Alert:* {{ .Labels.alertname }}{{ if .Labels.severity }} - {{ .Labels.severity }}{{ end }}\r\n\r\n *Summary:* {{ .Annotations.summary }}\r\n *Description:* {{ .Annotations.description }}\r\n *RelatedLogs:* {{ if gt (len .Annotations.related_logs) 0 -}} View in <{{ .Annotations.related_logs }}|logs explorer> {{- end}}\r\n *RelatedTraces:* {{ if gt (len .Annotations.related_traces) 0 -}} View in <{{ .Annotations.related_traces }}|traces explorer> {{- end}}\r\n\r\n *Details:*\r\n {{ range .Labels.SortedPairs -}}\r\n {{- if ne .Name "ruleId" -}}\r\n \u2022 *{{ .Name }}:* {{ .Value }}\r\n {{ end -}}\r\n {{ end -}}\r\n{{ end }}</body></html>',
"headers": {
"Subject": '[{{ .Status | toUpper }}{{ if eq .Status "firing" }}:{{ .Alerts.Firing | len }}{{ end }}] {{ .CommonLabels.alertname }} for {{ .CommonLabels.job }}\n {{- if gt (len .CommonLabels) (len .GroupLabels) -}}\n {{" "}}(\n {{- with .CommonLabels.Remove .GroupLabels.Names }}\n {{- range $index, $label := .SortedPairs -}}\n {{ if $index }}, {{ end }}\n {{- $label.Name }}="{{ $label.Value -}}"\n {{- end }}\n {{- end -}}\n )\n {{- end }}'
},
}
],
}
@pytest.fixture(name="notification_channel", scope="package")
def notification_channel(
network: Network,
@@ -67,6 +176,40 @@ def notification_channel(
)
@pytest.fixture(name="create_notification_channel", scope="function")
def create_notification_channel(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
) -> Callable[[dict], str]:
admin_token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
channel_ids = []
def _create_notification_channel(channel_config: dict) -> str:
response = requests.post(
signoz.self.host_configs["8080"].get("/api/v1/channels"),
json=channel_config,
headers={"Authorization": f"Bearer {admin_token}"},
timeout=5,
)
assert response.status_code == HTTPStatus.CREATED, f"Failed to create channel, Response: {response.text} Response status: {response.status_code}"
channel_id = response.json()["data"]["id"]
channel_ids.append(channel_id)
return channel_id
yield _create_notification_channel
for channel_id in channel_ids:
response = requests.delete(
signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"),
headers={"Authorization": f"Bearer {admin_token}"},
timeout=5,
)
if response.status_code != HTTPStatus.NO_CONTENT:
logger.error("Failed to delete channel: %s", {"channel_id": channel_id, "status": response.status_code, "response": response.text})
@pytest.fixture(name="create_webhook_notification_channel", scope="function")
def create_webhook_notification_channel(
signoz: types.SigNoz,
@@ -103,3 +246,19 @@ def create_webhook_notification_channel(
return channel_id
return _create_webhook_notification_channel
def send_test_notification(signoz: types.SigNoz, token: str, receiver: dict, wait_seconds: int = 90) -> None:
deadline = time.time() + wait_seconds
last = None
while time.time() < deadline:
last = requests.post(
signoz.self.host_configs["8080"].get("/api/v1/channels/test"),
json=receiver,
headers={"Authorization": f"Bearer {token}"},
timeout=30,
)
if last.status_code == HTTPStatus.NO_CONTENT:
return
time.sleep(2)
raise AssertionError(f"test notification did not succeed within {wait_seconds}s, last response: {last.status_code} {last.text}")

View File

@@ -197,3 +197,40 @@ class AlertTestCase:
alert_data: list[AlertData]
# list of alert expectations for the test case
alert_expectation: AlertExpectation
@dataclass(frozen=True)
class NotificationValidation:
# destination type of the notification, either webhook or email
# slack, msteams, pagerduty, opsgenie, webhook channels send notifications through webhook
# email channels send notifications through email
destination_type: Literal["webhook", "email"]
# validation data for validating the received notification payload
validation_data: dict[str, any]
@dataclass(frozen=True)
class AMNotificationExpectation:
# whether we expect any notifications to be fired or not, false when testing downtime scenarios
# or don't expect any notifications to be fired in given time period
should_notify: bool
# seconds to wait for the notifications to be fired, if no
# notifications are fired in the expected time, the test will fail
wait_time_seconds: int
# list of notifications to expect, as a single rule can trigger multiple notifications
# spanning across different notifiers
notification_validations: list[NotificationValidation]
@dataclass(frozen=True)
class AlertManagerNotificationTestCase:
# name of the test case
name: str
# path to the rule file in testdata directory
rule_path: str
# list of alert data that will be inserted into the database for the rule to be triggered
alert_data: list[AlertData]
# configuration for the notification channel
channel_config: dict[str, any]
# notification expectations for the test case
notification_expectation: AMNotificationExpectation

View File

@@ -39,5 +39,7 @@ def test_teardown(
idp: types.TestContainerIDP, # pylint: disable=unused-argument
create_user_admin: types.Operation, # pylint: disable=unused-argument
migrator: types.Operation, # pylint: disable=unused-argument
maildev: types.TestContainerDocker, # pylint: disable=unused-argument
notification_channel: types.TestContainerDocker, # pylint: disable=unused-argument
) -> None:
pass

View File

@@ -0,0 +1,20 @@
{ "timestamp": "2026-01-29T10:00:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "User login successful", "severity_text": "INFO" }
{ "timestamp": "2026-01-29T10:00:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: gateway timeout", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:01:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: card declined", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:01:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "Database connection established", "severity_text": "INFO" }
{ "timestamp": "2026-01-29T10:02:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: insufficient funds", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:02:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: invalid token", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:03:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "API request received", "severity_text": "INFO" }
{ "timestamp": "2026-01-29T10:03:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: gateway timeout", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:04:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: card declined", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:04:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: invalid token", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:05:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: gateway timeout", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:05:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: card declined", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:06:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: gateway timeout", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:06:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: insufficient funds", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:07:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: card declined", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:07:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: gateway timeout", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:08:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "Response sent to client", "severity_text": "INFO" }
{ "timestamp": "2026-01-29T10:08:30.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: invalid token", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:09:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: card declined", "severity_text": "ERROR" }
{ "timestamp": "2026-01-29T10:10:00.000000Z", "resources": { "service.name": "payment-service" }, "attributes": { "code.file": "payment_handler.py" }, "body": "payment failure: gateway timeout", "severity_text": "ERROR" }

View File

@@ -0,0 +1,69 @@
{
"alert": "content_templating_logs",
"ruleType": "threshold_rule",
"alertType": "LOGS_BASED_ALERT",
"condition": {
"thresholds": {
"kind": "basic",
"spec": [
{
"name": "critical",
"target": 0,
"matchType": "1",
"op": "1",
"channels": [
"test channel"
]
}
]
},
"compositeQuery": {
"queryType": "builder",
"panelType": "graph",
"queries": [
{
"type": "builder_query",
"spec": {
"name": "A",
"signal": "logs",
"filter": {
"expression": "body CONTAINS 'payment failure'"
},
"aggregations": [
{
"expression": "count()"
}
],
"groupBy": [
{"name": "service.name", "fieldContext": "resource"}
]
}
}
]
},
"selectedQueryName": "A"
},
"evaluation": {
"kind": "rolling",
"spec": {
"evalWindow": "5m0s",
"frequency": "15s"
}
},
"labels": {},
"annotations": {
"description": "Payment failure spike detected on $service_name",
"summary": "Payment failures elevated on $service_name"
},
"notificationSettings": {
"groupBy": [],
"usePolicy": false,
"renotify": {
"enabled": false,
"interval": "30m",
"alertStates": []
}
},
"version": "v5",
"schemaVersion": "v2alpha1"
}

View File

@@ -0,0 +1,12 @@
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:01:00+00:00","value":80,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:02:00+00:00","value":95,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:03:00+00:00","value":110,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:04:00+00:00","value":120,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:05:00+00:00","value":125,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:06:00+00:00","value":130,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:07:00+00:00","value":135,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:08:00+00:00","value":140,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:09:00+00:00","value":145,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:10:00+00:00","value":150,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:11:00+00:00","value":155,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}
{"metric_name":"container_memory_bytes_content_templating","labels":{"namespace":"production","pod":"checkout-7d9c8b5f4-x2k9p","container":"checkout","node":"ip-10-0-1-23","severity":"critical","service":"checkout"},"timestamp":"2026-01-29T10:12:00+00:00","value":160,"temporality":"Unspecified","type_":"Gauge","is_monotonic":false,"flags":0,"description":"","unit":"bytes","env":"default","resource_attrs":{},"scope_attrs":{}}

View File

@@ -0,0 +1,72 @@
{
"alert": "content_templating_metrics",
"ruleType": "threshold_rule",
"alertType": "METRIC_BASED_ALERT",
"condition": {
"thresholds": {
"kind": "basic",
"spec": [
{
"name": "critical",
"target": 100,
"matchType": "1",
"op": "1",
"channels": [
"test channel"
]
}
]
},
"compositeQuery": {
"queryType": "builder",
"panelType": "graph",
"queries": [
{
"type": "builder_query",
"spec": {
"name": "A",
"signal": "metrics",
"aggregations": [
{
"metricName": "container_memory_bytes_content_templating",
"timeAggregation": "avg",
"spaceAggregation": "max"
}
],
"groupBy": [
{"name": "namespace", "fieldContext": "attribute", "fieldDataType": "string"},
{"name": "pod", "fieldContext": "attribute", "fieldDataType": "string"},
{"name": "container", "fieldContext": "attribute", "fieldDataType": "string"},
{"name": "node", "fieldContext": "attribute", "fieldDataType": "string"},
{"name": "severity", "fieldContext": "attribute", "fieldDataType": "string"}
]
}
}
]
},
"selectedQueryName": "A"
},
"evaluation": {
"kind": "rolling",
"spec": {
"evalWindow": "5m0s",
"frequency": "15s"
}
},
"labels": {},
"annotations": {
"description": "Container $container in pod $pod ($namespace) exceeded memory threshold",
"summary": "High container memory in $namespace/$pod"
},
"notificationSettings": {
"groupBy": [],
"usePolicy": false,
"renotify": {
"enabled": false,
"interval": "30m",
"alertStates": []
}
},
"version": "v5",
"schemaVersion": "v2alpha1"
}

View File

@@ -0,0 +1,20 @@
{ "timestamp": "2026-01-29T10:00:00.000000Z", "duration": "PT1.2S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a1", "span_id": "c1b2c3d4e5f6a7b8", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:00:30.000000Z", "duration": "PT1.4S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a2", "span_id": "c2b3c4d5e6f7a8b9", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:01:00.000000Z", "duration": "PT1.6S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a3", "span_id": "c3b4c5d6e7f8a9b0", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:01:30.000000Z", "duration": "PT1.8S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a4", "span_id": "c4b5c6d7e8f9a0b1", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:02:00.000000Z", "duration": "PT2.1S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a5", "span_id": "c5b6c7d8e9f0a1b2", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:02:30.000000Z", "duration": "PT2.3S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a6", "span_id": "c6b7c8d9e0f1a2b3", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:03:00.000000Z", "duration": "PT2.5S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a7", "span_id": "c7b8c9d0e1f2a3b4", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:03:30.000000Z", "duration": "PT2.7S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a8", "span_id": "c8b9c0d1e2f3a4b5", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:04:00.000000Z", "duration": "PT2.9S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6a9", "span_id": "c9b0c1d2e3f4a5b6", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:04:30.000000Z", "duration": "PT3.1S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b1", "span_id": "d1c2d3e4f5a6b7c8", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:05:00.000000Z", "duration": "PT3.3S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b2", "span_id": "d2c3d4e5f6a7b8c9", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:05:30.000000Z", "duration": "PT3.5S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b3", "span_id": "d3c4d5e6f7a8b9c0", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:06:00.000000Z", "duration": "PT3.7S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b4", "span_id": "d4c5d6e7f8a9b0c1", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:06:30.000000Z", "duration": "PT3.9S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b5", "span_id": "d5c6d7e8f9a0b1c2", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:07:00.000000Z", "duration": "PT4.1S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b6", "span_id": "d6c7d8e9f0a1b2c3", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:07:30.000000Z", "duration": "PT4.3S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b7", "span_id": "d7c8d9e0f1a2b3c4", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:08:00.000000Z", "duration": "PT4.5S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b8", "span_id": "d8c9d0e1f2a3b4c5", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:08:30.000000Z", "duration": "PT4.7S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6b9", "span_id": "d9c0d1e2f3a4b5c6", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:09:00.000000Z", "duration": "PT4.9S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6c1", "span_id": "e1d2e3f4a5b6c7d8", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }
{ "timestamp": "2026-01-29T10:10:00.000000Z", "duration": "PT5.1S", "trace_id": "591f6d3d6b0a1f9e8a71b2c3d4e5f6c2", "span_id": "e2d3e4f5a6b7c8d9", "parent_span_id": "", "name": "POST /checkout", "kind": 2, "status_code": 1, "status_message": "", "resources": { "deployment.environment": "production", "service.name": "checkout-service", "os.type": "linux", "host.name": "ip-10-0-1-23" }, "attributes": { "net.transport": "IP.TCP", "http.scheme": "http", "http.user_agent": "Integration Test", "http.request.method": "POST", "http.response.status_code": "200", "http.request.path": "/checkout" } }

View File

@@ -0,0 +1,71 @@
{
"alert": "content_templating_traces",
"ruleType": "threshold_rule",
"alertType": "TRACES_BASED_ALERT",
"condition": {
"thresholds": {
"kind": "basic",
"spec": [
{
"name": "critical",
"target": 1,
"matchType": "1",
"op": "1",
"channels": [
"test channel"
],
"targetUnit": "s"
}
]
},
"compositeQuery": {
"queryType": "builder",
"unit": "ns",
"panelType": "graph",
"queries": [
{
"type": "builder_query",
"spec": {
"name": "A",
"signal": "traces",
"filter": {
"expression": "http.request.path = '/checkout'"
},
"aggregations": [
{
"expression": "p90(duration_nano)"
}
],
"groupBy": [
{"name": "service.name", "fieldContext": "resource"}
]
}
}
]
},
"selectedQueryName": "A"
},
"evaluation": {
"kind": "rolling",
"spec": {
"evalWindow": "5m0s",
"frequency": "15s"
}
},
"labels": {},
"annotations": {
"description": "p90 latency high on $service_name",
"summary": "p90 latency exceeded threshold on $service_name"
},
"notificationSettings": {
"groupBy": [],
"usePolicy": false,
"renotify": {
"enabled": false,
"interval": "30m",
"alertStates": []
}
},
"version": "v5",
"schemaVersion": "v2alpha1"
}

View File

@@ -0,0 +1,268 @@
import time
import uuid
from collections.abc import Callable
from http import HTTPStatus
import pytest
import requests
from sqlalchemy import text
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.maildev import (
MAILDEV_INCOMING_PASS,
SMTP_TEST_FROM,
delete_all_mails,
verify_email_received,
)
from fixtures.notification_channel import assert_email_channel_payload_clean, send_test_notification
TIMEOUT = 10
CHANNEL_TYPE_CASES = [
(
"webhook",
lambda sink: {"webhook_configs": [{"url": sink.container_configs["8080"].get("/webhook/crud-original"), "send_resolved": True}]},
lambda sink: {"webhook_configs": [{"url": sink.container_configs["8080"].get("/webhook/crud-updated"), "send_resolved": True}]},
"crud-original",
"crud-updated",
),
(
"slack",
lambda sink: {"slack_configs": [{"api_url": sink.container_configs["8080"].get("/services/T/B/X"), "channel": "#crud-original"}]},
lambda sink: {"slack_configs": [{"api_url": sink.container_configs["8080"].get("/services/T/B/X"), "channel": "#crud-updated"}]},
"#crud-original",
"#crud-updated",
),
(
"pagerduty",
lambda sink: {"pagerduty_configs": [{"routing_key": "crud-original-routing-key"}]},
lambda sink: {"pagerduty_configs": [{"routing_key": "crud-updated-routing-key"}]},
"crud-original-routing-key",
"crud-updated-routing-key",
),
(
"opsgenie",
lambda sink: {"opsgenie_configs": [{"api_key": "crud-original-api-key", "message": "{{ .CommonLabels.alertname }}"}]},
lambda sink: {"opsgenie_configs": [{"api_key": "crud-updated-api-key", "message": "{{ .CommonLabels.alertname }}"}]},
"crud-original-api-key",
"crud-updated-api-key",
),
(
"msteamsv2",
lambda sink: {"msteamsv2_configs": [{"webhook_url": sink.container_configs["8080"].get("/msteams/crud-original")}]},
lambda sink: {"msteamsv2_configs": [{"webhook_url": sink.container_configs["8080"].get("/msteams/crud-updated")}]},
"crud-original",
"crud-updated",
),
(
"email",
lambda sink: {"email_configs": [{"to": "crud-original@integration.test"}]},
lambda sink: {"email_configs": [{"to": "crud-updated@integration.test"}]},
"crud-original@integration.test",
"crud-updated@integration.test",
),
]
@pytest.mark.parametrize(
"channel_type,make_config,make_updated_config,created_marker,updated_marker",
CHANNEL_TYPE_CASES,
ids=[case[0] for case in CHANNEL_TYPE_CASES],
)
def test_channel_crud( # pylint: disable=too-many-arguments,too-many-positional-arguments
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
notification_channel: types.TestContainerDocker,
channel_type: str,
make_config: Callable[[types.TestContainerDocker], dict],
make_updated_config: Callable[[types.TestContainerDocker], dict],
created_marker: str,
updated_marker: str,
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
name = f"crud-{channel_type}-{uuid.uuid4()}"
config = {"name": name, **make_config(notification_channel)}
response = requests.post(signoz.self.host_configs["8080"].get("/api/v1/channels"), json=config, headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.CREATED, response.text
created = response.json()["data"]
channel_id = created["id"]
assert created["name"] == name
assert created["type"] == channel_type
response = requests.get(signoz.self.host_configs["8080"].get("/api/v1/channels"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.OK, response.text
listed = {channel["name"]: channel for channel in response.json()["data"]}
assert name in listed
assert listed[name]["type"] == channel_type
response = requests.get(signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.OK, response.text
assert created_marker in response.json()["data"]["data"]
updated_config = {"name": name, **make_updated_config(notification_channel)}
response = requests.put(signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"), json=updated_config, headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.NO_CONTENT, response.text
response = requests.get(signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.OK, response.text
data = response.json()["data"]["data"]
assert updated_marker in data
assert created_marker not in data
response = requests.delete(signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.NO_CONTENT, response.text
response = requests.get(signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.NOT_FOUND, response.text
def test_create_rejects_duplicate_name(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
create_notification_channel: Callable[[dict], str],
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
name = f"duplicate-{uuid.uuid4()}"
create_notification_channel({"name": name, "email_configs": [{"to": "first@integration.test"}]})
response = requests.post(
signoz.self.host_configs["8080"].get("/api/v1/channels"),
json={"name": name, "email_configs": [{"to": "second@integration.test"}]},
headers={"Authorization": f"Bearer {token}"},
timeout=TIMEOUT,
)
assert response.status_code == HTTPStatus.BAD_REQUEST, response.text
assert "unique" in response.text
def test_create_rejects_channel_without_configs(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = requests.post(
signoz.self.host_configs["8080"].get("/api/v1/channels"),
json={"name": f"empty-{uuid.uuid4()}"},
headers={"Authorization": f"Bearer {token}"},
timeout=TIMEOUT,
)
assert response.status_code == HTTPStatus.BAD_REQUEST, response.text
assert "notification configuration" in response.text
def test_update_rejects_name_change(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
create_notification_channel: Callable[[dict], str],
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
name = f"rename-{uuid.uuid4()}"
channel_id = create_notification_channel({"name": name, "email_configs": [{"to": "rename@integration.test"}]})
response = requests.put(
signoz.self.host_configs["8080"].get(f"/api/v1/channels/{channel_id}"),
json={"name": f"{name}-renamed", "email_configs": [{"to": "rename@integration.test"}]},
headers={"Authorization": f"Bearer {token}"},
timeout=TIMEOUT,
)
assert response.status_code == HTTPStatus.BAD_REQUEST, response.text
assert "cannot update channel name" in response.text
def test_channels_require_authentication(signoz: types.SigNoz) -> None:
response = requests.get(signoz.self.host_configs["8080"].get("/api/v1/channels"), timeout=TIMEOUT)
assert response.status_code == HTTPStatus.UNAUTHORIZED, response.text
def test_email_channel_never_stores_or_serves_smtp_settings(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
hostile_name = f"hostile-email-{uuid.uuid4()}"
hostile_config = {
"name": hostile_name,
"email_configs": [
{
"to": "hostile@integration.test",
"from": "spoofed@integration.test",
"hello": "attacker.test",
"smarthost": "smtp.attacker.test:2525",
"auth_username": "attacker",
"auth_password": "tenant-posted-secret",
"require_tls": False,
"headers": {"Subject": "hostile subject"},
}
],
}
response = requests.post(signoz.self.host_configs["8080"].get("/api/v1/channels"), json=hostile_config, headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.CREATED, response.text
created = response.json()["data"]
assert_email_channel_payload_clean(created["data"])
response = requests.get(signoz.self.host_configs["8080"].get(f"/api/v1/channels/{created['id']}"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.OK, response.text
served = response.json()["data"]["data"]
assert_email_channel_payload_clean(served)
assert "hostile@integration.test" in served
assert "hostile subject" in served
assert "smtp.attacker.test" not in served
assert "tenant-posted-secret" not in served
response = requests.get(signoz.self.host_configs["8080"].get("/api/v1/channels"), headers={"Authorization": f"Bearer {token}"}, timeout=TIMEOUT)
assert response.status_code == HTTPStatus.OK, response.text
assert "tenant-posted-secret" not in response.text
assert MAILDEV_INCOMING_PASS not in response.text
with signoz.sqlstore.conn.connect() as conn:
stored = conn.execute(
text("SELECT data FROM notification_channel WHERE name = :name"),
{"name": hostile_name},
).fetchone()
assert stored is not None
assert_email_channel_payload_clean(stored[0])
assert "tenant-posted-secret" not in stored[0]
configs = conn.execute(text("SELECT config FROM alertmanager_config")).fetchall()
assert len(configs) > 0
for (config_raw,) in configs:
assert MAILDEV_INCOMING_PASS not in config_raw
assert "tenant-posted-secret" not in config_raw
assert '"smtp_auth_password"' not in config_raw
assert '"auth_password"' not in config_raw
def test_email_test_channel_delivers_via_env_transport(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
maildev: types.TestContainerDocker,
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
delete_all_mails(maildev)
recipient = f"delivery-{uuid.uuid4()}@integration.test"
send_test_notification(
signoz,
token,
{"name": f"delivery-{uuid.uuid4()}", "email_configs": [{"to": recipient}]},
)
deadline = time.time() + 30
while time.time() < deadline:
if verify_email_received(maildev, {"to": recipient, "from": SMTP_TEST_FROM}):
return
time.sleep(1)
raise AssertionError(f"no email delivered to {recipient} from {SMTP_TEST_FROM}")

View File

@@ -0,0 +1,360 @@
import json
import re
import time
import uuid
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
import pytest
from wiremock.client import HttpMethods, Mapping, MappingRequest, MappingResponse
from fixtures import types
from fixtures.alerts import (
get_testdata_file_path,
update_raw_channel_config,
update_rule_channel_name,
verify_notification_expectation,
)
from fixtures.logger import setup_logger
from fixtures.maildev import delete_all_mails
from fixtures.notification_channel import (
email_default_config,
msteams_default_config,
opsgenie_default_config,
pagerduty_default_config,
slack_default_config,
webhook_default_config,
)
logger = setup_logger(__name__)
NOTIFIERS_TEST = [
types.AlertManagerNotificationTestCase(
name="slack_notifier_default_templating",
rule_path="alerts/test_scenarios/threshold_above_at_least_once/rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alerts/test_scenarios/threshold_above_at_least_once/alert_data.jsonl",
),
],
channel_config=slack_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
# extra wait for alertmanager server setup
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/services/TEAM_ID/BOT_ID/TOKEN_ID",
"json_body": {
"username": "Alertmanager",
"attachments": [
{
"color": "danger",
"mrkdwn_in": ["fallback", "pretext", "text"],
}
],
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="msteams_notifier_default_templating",
rule_path="alerts/test_scenarios/threshold_above_at_least_once/rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alerts/test_scenarios/threshold_above_at_least_once/alert_data.jsonl",
),
],
channel_config=msteams_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/msteams/webhook_url",
"json_body": {
"type": "message",
"attachments": [
{
"contentType": "application/vnd.microsoft.card.adaptive",
"content": {
"$schema": "http://adaptivecards.io/schemas/adaptive-card.json",
"type": "AdaptiveCard",
"version": "1.2",
"body": [
{
"type": "TextBlock",
"text": "Alerts",
"weight": "Bolder",
"size": "Medium",
"wrap": True,
"color": "Attention",
},
{
"type": "TextBlock",
"text": "Labels",
"weight": "Bolder",
"size": "Medium",
},
{
"type": "FactSet",
"text": "",
"facts": [
{
"title": "threshold.name",
"value": "critical",
}
],
},
{
"type": "TextBlock",
"text": "Annotations",
"weight": "Bolder",
"size": "Medium",
},
{
"type": "FactSet",
"text": "",
"facts": [
{
"title": "description",
"value": "This alert is fired when the defined metric (current value: 15) crosses the threshold (10)",
},
],
},
],
"msteams": {"width": "full"},
"actions": [
{
"type": "Action.OpenUrl",
"title": "View Alert",
}
],
},
}
],
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="pagerduty_notifier_default_templating",
rule_path="alerts/test_scenarios/threshold_above_at_least_once/rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alerts/test_scenarios/threshold_above_at_least_once/alert_data.jsonl",
),
],
channel_config=pagerduty_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/v2/enqueue",
"json_body": {
"routing_key": "PagerDutyRoutingKey",
"event_action": "trigger",
"payload": {
"source": "SigNoz Alert Manager",
"severity": "critical",
"custom_details": {
"firing": {
"Annotations": [
{"description = This alert is fired when the defined metric (current value": "15) crosses the threshold (10)"},
],
"Labels": [
"alertname = threshold_above_at_least_once",
"severity = critical",
"threshold.name = critical",
],
}
},
},
"client": "SigNoz Alert Manager",
"client_url": "https://enter-signoz-host-n-port-here/alerts",
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="opsgenie_notifier_default_templating",
rule_path="alerts/test_scenarios/threshold_above_at_least_once/rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alerts/test_scenarios/threshold_above_at_least_once/alert_data.jsonl",
),
],
channel_config=opsgenie_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/v2/alerts",
"json_body": {
"message": "threshold_above_at_least_once",
"details": {
"alertname": "threshold_above_at_least_once",
"severity": "critical",
"threshold.name": "critical",
},
"priority": "P1",
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="webhook_notifier_default_templating",
rule_path="alerts/test_scenarios/threshold_above_at_least_once/rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alerts/test_scenarios/threshold_above_at_least_once/alert_data.jsonl",
),
],
channel_config=webhook_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/webhook/webhook_url",
"json_body": {
"status": "firing",
"alerts": [
{
"status": "firing",
"labels": {
"alertname": "threshold_above_at_least_once",
"severity": "critical",
"threshold.name": "critical",
},
"annotations": {
"description": "This alert is fired when the defined metric (current value: 15) crosses the threshold (10)",
"summary": "This alert is fired when the defined metric (current value: 15) crosses the threshold (10)",
},
}
],
"commonLabels": {
"alertname": "threshold_above_at_least_once",
"severity": "critical",
"threshold.name": "critical",
},
"commonAnnotations": {
"description": "This alert is fired when the defined metric (current value: 15) crosses the threshold (10)",
"summary": "This alert is fired when the defined metric (current value: 15) crosses the threshold (10)",
},
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="email_notifier_default_templating",
rule_path="alerts/test_scenarios/threshold_above_at_least_once/rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alerts/test_scenarios/threshold_above_at_least_once/alert_data.jsonl",
),
],
channel_config=email_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="email",
validation_data={
"subject": re.compile(r'\[FIRING:1\] threshold_above_at_least_once for \(alertname="threshold_above_at_least_once", ruleSource="http://localhost:8080/alerts/overview\?ruleId=[0-9a-f-]+", severity="critical", threshold\.name="critical"\)'),
},
),
],
),
),
]
@pytest.mark.parametrize(
"notifier_test_case",
NOTIFIERS_TEST,
ids=lambda notifier_test_case: notifier_test_case.name,
)
def test_notifier_templating( # pylint: disable=too-many-arguments,too-many-positional-arguments
notification_channel: types.TestContainerDocker,
make_http_mocks: Callable[[types.TestContainerDocker, list[Mapping]], None],
create_notification_channel: Callable[[dict], str],
create_alert_rule: Callable[[dict], str],
insert_alert_data: Callable[[list[types.AlertData], datetime], None],
maildev: types.TestContainerDocker,
notifier_test_case: types.AlertManagerNotificationTestCase,
):
channel_name = str(uuid.uuid4())
channel_config = update_raw_channel_config(notifier_test_case.channel_config, channel_name, notification_channel)
logger.info("Channel config: %s", {"channel_config": channel_config})
webhook_validations = [v for v in notifier_test_case.notification_expectation.notification_validations if v.destination_type == "webhook"]
if len(webhook_validations) > 0:
mock_mappings = [
Mapping(
request=MappingRequest(method=HttpMethods.POST, url=v.validation_data["path"]),
response=MappingResponse(status=200, json_body={}),
persistent=False,
)
for v in webhook_validations
]
make_http_mocks(notification_channel, mock_mappings)
logger.info("Mock mappings created")
if any(v.destination_type == "email" for v in notifier_test_case.notification_expectation.notification_validations):
delete_all_mails(maildev)
logger.info("Mails deleted")
create_notification_channel(channel_config)
logger.info("Channel created with name: %s", {"channel_name": channel_name})
time.sleep(12)
insert_alert_data(
notifier_test_case.alert_data,
base_time=datetime.now(tz=UTC) - timedelta(minutes=5),
)
rule_path = get_testdata_file_path(notifier_test_case.rule_path)
with open(rule_path, encoding="utf-8") as f:
rule_data = json.loads(f.read())
update_rule_channel_name(rule_data, channel_name)
rule_id = create_alert_rule(rule_data)
logger.info("rule created: %s", {"rule_id": rule_id, "rule_name": rule_data["alert"]})
verify_notification_expectation(
notification_channel,
maildev,
notifier_test_case.notification_expectation,
)

View File

@@ -0,0 +1,332 @@
import json
import re
import time
import uuid
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
import pytest
from wiremock.client import HttpMethods, Mapping, MappingRequest, MappingResponse
from fixtures import types
from fixtures.alerts import (
get_testdata_file_path,
update_raw_channel_config,
update_rule_channel_name,
verify_notification_expectation,
)
from fixtures.logger import setup_logger
from fixtures.maildev import delete_all_mails
from fixtures.notification_channel import (
msteams_default_config,
opsgenie_default_config,
pagerduty_default_config,
slack_default_config,
webhook_default_config,
)
logger = setup_logger(__name__)
CONTENT_TEMPLATING_TEST = [
types.AlertManagerNotificationTestCase(
name="msteams_metrics_default_templating",
rule_path="alertmanager/content_templating/metrics_rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alertmanager/content_templating/metrics_data.jsonl",
),
],
channel_config=msteams_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/msteams/webhook_url",
"json_body": {
"type": "message",
"attachments": [
{
"contentType": "application/vnd.microsoft.card.adaptive",
"content": {
"type": "AdaptiveCard",
"body": [
{
"type": "TextBlock",
"text": re.compile(
r'\[FIRING:1\] content_templating_metrics for \(alertname="content_templating_metrics", container="checkout", namespace="production", node="ip-10-0-1-23", pod="checkout-7d9c8b5f4-x2k9p", ruleSource="http://localhost:8080/alerts/overview\?ruleId=[0-9a-f-]+", severity="critical", threshold\.name="critical"\)'
),
},
],
},
}
],
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="opsgenie_metrics_default_templating",
rule_path="alertmanager/content_templating/metrics_rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alertmanager/content_templating/metrics_data.jsonl",
),
],
channel_config=opsgenie_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/v2/alerts",
"json_body": {
"message": "content_templating_metrics",
"details": {
"alertname": "content_templating_metrics",
"container": "checkout",
"namespace": "production",
"node": "ip-10-0-1-23",
"pod": "checkout-7d9c8b5f4-x2k9p",
"severity": "critical",
"threshold.name": "critical",
},
"priority": "P1",
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="pagerduty_metrics_default_templating",
rule_path="alertmanager/content_templating/metrics_rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alertmanager/content_templating/metrics_data.jsonl",
),
],
channel_config=pagerduty_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/v2/enqueue",
"json_body": {
"routing_key": "PagerDutyRoutingKey",
"payload": {
"severity": "critical",
"custom_details": {
"firing": {
"Labels": [
"alertname = content_templating_metrics",
"container = checkout",
"namespace = production",
"node = ip-10-0-1-23",
"pod = checkout-7d9c8b5f4-x2k9p",
"severity = critical",
"threshold.name = critical",
],
}
},
},
"client": "SigNoz Alert Manager",
"client_url": "https://enter-signoz-host-n-port-here/alerts",
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="slack_logs_default_templating",
rule_path="alertmanager/content_templating/logs_rule.json",
alert_data=[
types.AlertData(
type="logs",
data_path="alertmanager/content_templating/logs_data.jsonl",
),
],
channel_config=slack_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/services/TEAM_ID/BOT_ID/TOKEN_ID",
"json_body": {
"username": "Alertmanager",
"attachments": [
{
"color": "danger",
"mrkdwn_in": ["fallback", "pretext", "text"],
}
],
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="slack_metrics_default_templating",
rule_path="alertmanager/content_templating/metrics_rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alertmanager/content_templating/metrics_data.jsonl",
),
],
channel_config=slack_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/services/TEAM_ID/BOT_ID/TOKEN_ID",
"json_body": {
"username": "Alertmanager",
"attachments": [
{
"color": "danger",
"mrkdwn_in": ["fallback", "pretext", "text"],
}
],
},
},
),
],
),
),
types.AlertManagerNotificationTestCase(
name="webhook_metrics_default_templating",
rule_path="alertmanager/content_templating/metrics_rule.json",
alert_data=[
types.AlertData(
type="metrics",
data_path="alertmanager/content_templating/metrics_data.jsonl",
),
],
channel_config=webhook_default_config,
notification_expectation=types.AMNotificationExpectation(
should_notify=True,
wait_time_seconds=120,
notification_validations=[
types.NotificationValidation(
destination_type="webhook",
validation_data={
"path": "/webhook/webhook_url",
"json_body": {
"status": "firing",
"alerts": [
{
"status": "firing",
"labels": {
"alertname": "content_templating_metrics",
"container": "checkout",
"namespace": "production",
"node": "ip-10-0-1-23",
"pod": "checkout-7d9c8b5f4-x2k9p",
"severity": "critical",
"threshold.name": "critical",
},
"annotations": {
"description": "Container checkout in pod checkout-7d9c8b5f4-x2k9p (production) exceeded memory threshold",
"summary": "High container memory in production/checkout-7d9c8b5f4-x2k9p",
},
}
],
"commonLabels": {
"alertname": "content_templating_metrics",
"container": "checkout",
"namespace": "production",
"node": "ip-10-0-1-23",
"pod": "checkout-7d9c8b5f4-x2k9p",
"severity": "critical",
"threshold.name": "critical",
},
},
},
),
],
),
),
]
@pytest.mark.parametrize(
"content_templating_test_case",
CONTENT_TEMPLATING_TEST,
ids=lambda content_templating_test_case: content_templating_test_case.name,
)
def test_content_templating( # pylint: disable=too-many-arguments,too-many-positional-arguments
notification_channel: types.TestContainerDocker,
make_http_mocks: Callable[[types.TestContainerDocker, list[Mapping]], None],
create_notification_channel: Callable[[dict], str],
create_alert_rule: Callable[[dict], str],
insert_alert_data: Callable[[list[types.AlertData], datetime], None],
maildev: types.TestContainerDocker,
content_templating_test_case: types.AlertManagerNotificationTestCase,
):
channel_name = str(uuid.uuid4())
channel_config = update_raw_channel_config(content_templating_test_case.channel_config, channel_name, notification_channel)
logger.info("Channel config: %s", {"channel_config": channel_config})
webhook_validations = [v for v in content_templating_test_case.notification_expectation.notification_validations if v.destination_type == "webhook"]
if len(webhook_validations) > 0:
mock_mappings = [
Mapping(
request=MappingRequest(method=HttpMethods.POST, url=v.validation_data["path"]),
response=MappingResponse(status=200, json_body={}),
persistent=False,
)
for v in webhook_validations
]
make_http_mocks(notification_channel, mock_mappings)
logger.info("Mock mappings created")
if any(v.destination_type == "email" for v in content_templating_test_case.notification_expectation.notification_validations):
delete_all_mails(maildev)
logger.info("Mails deleted")
create_notification_channel(channel_config)
logger.info("Channel created with name: %s", {"channel_name": channel_name})
time.sleep(12)
insert_alert_data(
content_templating_test_case.alert_data,
base_time=datetime.now(tz=UTC) - timedelta(minutes=10),
)
rule_path = get_testdata_file_path(content_templating_test_case.rule_path)
with open(rule_path, encoding="utf-8") as f:
rule_data = json.loads(f.read())
update_rule_channel_name(rule_data, channel_name)
rule_id = create_alert_rule(rule_data)
logger.info("rule created: %s", {"rule_id": rule_id, "rule_name": rule_data["alert"]})
verify_notification_expectation(
notification_channel,
maildev,
content_templating_test_case.notification_expectation,
)

View File

@@ -0,0 +1,35 @@
import pytest
from testcontainers.core.container import Network
from fixtures import types
from fixtures.maildev import signoz_smtp_env
from fixtures.signoz import create_signoz
@pytest.fixture(name="signoz", scope="package")
def signoz( # pylint: disable=too-many-arguments,too-many-positional-arguments
network: Network,
zeus: types.TestContainerDocker,
gateway: types.TestContainerDocker,
sqlstore: types.TestContainerSQL,
clickhouse: types.TestContainerClickhouse,
request: pytest.FixtureRequest,
pytestconfig: pytest.Config,
maildev: types.TestContainerDocker,
notification_channel: types.TestContainerDocker,
) -> types.SigNoz:
return create_signoz(
network=network,
zeus=zeus,
gateway=gateway,
sqlstore=sqlstore,
clickhouse=clickhouse,
request=request,
pytestconfig=pytestconfig,
cache_key="signoz_alertmanager",
env_overrides={
**signoz_smtp_env(maildev),
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_PAGERDUTY__URL": notification_channel.container_configs["8080"].get("/v2/enqueue"),
"SIGNOZ_ALERTMANAGER_SIGNOZ_GLOBAL_OPSGENIE__API__URL": notification_channel.container_configs["8080"].get("/"),
},
)

View File

@@ -0,0 +1,99 @@
import time
import uuid
from collections.abc import Callable
from http import HTTPStatus
import docker
import pytest
import requests
from testcontainers.core.container import Network
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD, token_getter
from fixtures.logger import setup_logger
from fixtures.maildev import (
NEW_PROVIDER_SMTP_PASS,
SMTP_TEST_FROM,
delete_all_mails,
get_all_mails,
signoz_smtp_env,
verify_email_received,
)
from fixtures.notification_channel import send_test_notification
from fixtures.signoz import create_signoz
logger = setup_logger(__name__)
def wait_for_email(maildev: types.TestContainerDocker, filters: dict, wait_seconds: int = 30) -> None:
deadline = time.time() + wait_seconds
while time.time() < deadline:
if verify_email_received(maildev, filters):
return
time.sleep(1)
raise AssertionError(f"no email matching {filters} within {wait_seconds}s, inbox: {get_all_mails(maildev)}")
def test_smtp_rotation_applies_to_existing_channels( # pylint: disable=too-many-arguments,too-many-positional-arguments,too-many-locals
network: Network,
zeus: types.TestContainerDocker,
gateway: types.TestContainerDocker,
sqlstore: types.TestContainerSQL,
clickhouse: types.TestContainerClickhouse,
request: pytest.FixtureRequest,
pytestconfig: pytest.Config,
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
maildev_old: types.TestContainerDocker,
maildev_new: types.TestContainerDocker,
) -> None:
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
channel_name = f"rotation-{uuid.uuid4()}"
recipient = f"rotation-{uuid.uuid4()}@integration.test"
response = requests.post(
signoz.self.host_configs["8080"].get("/api/v1/channels"),
json={"name": channel_name, "email_configs": [{"to": recipient}]},
headers={"Authorization": f"Bearer {token}"},
timeout=10,
)
assert response.status_code == HTTPStatus.CREATED, response.text
delete_all_mails(maildev_old)
recipient_old_probe = f"probe-old-{uuid.uuid4()}@integration.test"
send_test_notification(signoz, token, {"name": f"probe-{uuid.uuid4()}", "email_configs": [{"to": recipient_old_probe}]})
wait_for_email(maildev_old, {"to": recipient_old_probe, "from": SMTP_TEST_FROM})
logger.info("Delivery through the old provider verified")
docker.from_env().containers.get(signoz.self.id).stop()
logger.info("Stopped signoz running against the old provider")
signoz_new = create_signoz(
network=network,
zeus=zeus,
gateway=gateway,
sqlstore=sqlstore,
clickhouse=clickhouse,
request=request,
pytestconfig=pytestconfig,
cache_key="signoz_smtp_rotation_new",
env_overrides=signoz_smtp_env(maildev_new, password=NEW_PROVIDER_SMTP_PASS),
)
token_new = token_getter(signoz_new)(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = requests.get(
signoz_new.self.host_configs["8080"].get("/api/v1/channels"),
headers={"Authorization": f"Bearer {token_new}"},
timeout=10,
)
assert response.status_code == HTTPStatus.OK, response.text
listed = {channel["name"]: channel for channel in response.json()["data"]}
assert channel_name in listed
delete_all_mails(maildev_new)
mails_at_old_provider = len(get_all_mails(maildev_old))
recipient_new_probe = f"probe-new-{uuid.uuid4()}@integration.test"
send_test_notification(signoz_new, token_new, {"name": f"probe-{uuid.uuid4()}", "email_configs": [{"to": recipient_new_probe}]})
wait_for_email(maildev_new, {"to": recipient_new_probe, "from": SMTP_TEST_FROM})
assert len(get_all_mails(maildev_old)) == mails_at_old_provider, "old provider must receive nothing after rotation"

View File

@@ -0,0 +1,40 @@
import pytest
from testcontainers.core.container import Network
from fixtures import types
from fixtures.maildev import NEW_PROVIDER_SMTP_PASS, OLD_PROVIDER_SMTP_PASS, create_maildev, signoz_smtp_env
from fixtures.signoz import create_signoz
@pytest.fixture(name="maildev_old", scope="package")
def maildev_old(network: Network, request: pytest.FixtureRequest, pytestconfig: pytest.Config) -> types.TestContainerDocker:
return create_maildev(network, request, pytestconfig, cache_key="maildev_smtp_old", incoming_pass=OLD_PROVIDER_SMTP_PASS)
@pytest.fixture(name="maildev_new", scope="package")
def maildev_new(network: Network, request: pytest.FixtureRequest, pytestconfig: pytest.Config) -> types.TestContainerDocker:
return create_maildev(network, request, pytestconfig, cache_key="maildev_smtp_new", incoming_pass=NEW_PROVIDER_SMTP_PASS)
@pytest.fixture(name="signoz", scope="package")
def signoz( # pylint: disable=too-many-arguments,too-many-positional-arguments
network: Network,
zeus: types.TestContainerDocker,
gateway: types.TestContainerDocker,
sqlstore: types.TestContainerSQL,
clickhouse: types.TestContainerClickhouse,
request: pytest.FixtureRequest,
pytestconfig: pytest.Config,
maildev_old: types.TestContainerDocker,
) -> types.SigNoz:
return create_signoz(
network=network,
zeus=zeus,
gateway=gateway,
sqlstore=sqlstore,
clickhouse=clickhouse,
request=request,
pytestconfig=pytestconfig,
cache_key="signoz_smtp_rotation",
env_overrides=signoz_smtp_env(maildev_old, password=OLD_PROVIDER_SMTP_PASS),
)

View File

@@ -0,0 +1,253 @@
import json
from collections import namedtuple
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
from http import HTTPStatus
import pytest
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.logs import Logs
from fixtures.querier import build_order_by, build_raw_query, get_rows, make_query_request
# querierlogs/15_search.py with use_json_body on (see conftest.py): body matches run against
# body_v2, the map/log fan-out is unchanged, and the response `body` comes back parsed — a
# plain-string body is {"message": <body>}.
Bodies = namedtuple("Bodies", ["a", "b", "c", "d"])
@pytest.mark.parametrize(
"expression,expected",
[
# ── keyless: fans across every field ────────────────────────────────
pytest.param("search('login')", lambda b: {b.a}, id="keyless_body"),
pytest.param("search('checkout')", lambda b: {b.a, b.d}, id="keyless_body_and_resource"),
pytest.param("search('useast')", lambda b: {b.a, b.c}, id="keyless_resource_value"),
pytest.param("search('acme')", lambda b: {b.a, b.c}, id="keyless_attribute_value"),
pytest.param("search('tenant')", lambda b: {b.a, b.b, b.c, b.d}, id="keyless_attribute_key"),
pytest.param("search('error')", lambda b: {b.b, b.d}, id="keyless_severity_case_insensitive"),
pytest.param("search('CHECKOUT')", lambda b: {b.a, b.d}, id="keyless_term_case_insensitive"),
# ── scoped: narrows to one context ──────────────────────────────────
pytest.param("search('login', body)", lambda b: {b.a}, id="scope_body"),
pytest.param("search('login', 'body')", lambda b: {b.a}, id="scope_body_quoted"),
pytest.param("search('checkout', body)", lambda b: {b.a}, id="scope_body_excludes_resource"),
pytest.param("search('checkout', resource)", lambda b: {b.a, b.d}, id="scope_resource"),
pytest.param("search('acme', attribute)", lambda b: {b.a, b.c}, id="scope_attribute"),
pytest.param("search('error', log)", lambda b: {b.b, b.d}, id="scope_log_severity"),
pytest.param("search('acme', body)", lambda b: set(), id="scope_body_no_match"),
pytest.param("search('checkout', attribute)", lambda b: set(), id="scope_attribute_no_match"),
# ── multiple scopes: union of the named contexts ────────────────────
pytest.param("search('login', body, resource)", lambda b: {b.a}, id="scopes_body_resource_body_only"),
pytest.param("search('checkout', body, resource)", lambda b: {b.a, b.d}, id="scopes_body_resource_union"),
# ── composition with boolean / field filters ────────────────────────
pytest.param("NOT search('login')", lambda b: {b.b, b.c, b.d}, id="negated"),
pytest.param("search('useast') AND severity_text = 'INFO'", lambda b: {b.a}, id="and_field_filter"),
],
)
def test_search(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
expression: str,
expected: Callable[[Bodies], set[str]],
) -> None:
"""Four self-naming logs, each with a token planted in a distinct place (body,
resource, attribute, severity), assert search() reaches exactly the right ones."""
body = Bodies(
a="alpha checkout login ok", # service checkout / region useast / tenant acme / INFO
b="bravo declined", # service payment / region euwest / tenant globex / ERROR
c="charlie miss", # service cart / region useast / tenant acme / WARN
d="delta slow", # service checkout / region apac / tenant initech / ERROR
)
# (body, resources, attributes, severity_text)
specs = [
(body.a, {"service.name": "checkout", "region": "useast"}, {"tenant": "acme"}, "INFO"),
(body.b, {"service.name": "payment", "region": "euwest"}, {"tenant": "globex"}, "ERROR"),
(body.c, {"service.name": "cart", "region": "useast"}, {"tenant": "acme"}, "WARN"),
(body.d, {"service.name": "checkout", "region": "apac"}, {"tenant": "initech"}, "ERROR"),
]
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
logs = [Logs(timestamp=now - timedelta(seconds=i + 1), resources=res, attributes=attrs, body=b, severity_text=sev) for i, (b, res, attrs, sev) in enumerate(specs)]
insert_logs(logs)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = make_query_request(
signoz,
token,
start_ms=int((now - timedelta(minutes=5)).timestamp() * 1000),
end_ms=int(now.timestamp() * 1000),
request_type="raw",
queries=[
build_raw_query(
"A",
"logs",
filter_expression=expression,
order=[build_order_by("timestamp", "desc"), build_order_by("id", "desc")],
limit=100,
)
],
)
assert response.status_code == HTTPStatus.OK, response.text
assert response.json()["status"] == "success"
# body_v2 comes back parsed; a plain-string body is {"message": <body>}.
assert {row["data"]["body"]["message"] for row in get_rows(response)} == expected(body)
@pytest.mark.parametrize(
"term",
[
pytest.param("eve@acme.io", id="nested_string_value"),
pytest.param("503", id="nested_numeric_value"),
pytest.param("status", id="nested_key"),
],
)
def test_search_body_reaches_nested_json(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
term: str,
) -> None:
"""A body-scoped search matches values and keys nested inside the body_v2 JSON."""
# searchable content lives only in nested fields, not a top-level message
nested_body = json.dumps({"user": {"email": "eve@acme.io"}, "http": {"status": 503}}, separators=(",", ":"))
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
insert_logs([Logs(timestamp=now - timedelta(seconds=1), resources={"service.name": "api"}, body=nested_body, severity_text="INFO")])
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = make_query_request(
signoz,
token,
start_ms=int((now - timedelta(minutes=5)).timestamp() * 1000),
end_ms=int(now.timestamp() * 1000),
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression=f"search('{term}', body)", order=[build_order_by("timestamp", "desc")], limit=100)],
)
assert response.status_code == HTTPStatus.OK, response.text
rows = get_rows(response)
assert len(rows) == 1
assert rows[0]["data"]["body"]["user"]["email"] == "eve@acme.io"
@pytest.mark.parametrize(
"expression",
[
pytest.param("search('login', bogus)", id="unknown_scope_word"),
pytest.param("search('login', body.message)", id="qualified_field_not_a_scope"),
],
)
def test_search_invalid_scope_rejected(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
expression: str,
) -> None:
"""A scope that is not a field context (an unknown word, or a qualified
`context.field`) is rejected at build time with a 400 — even when, as with
`body.message`, it names a real body path."""
now = datetime.now(tz=UTC)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = make_query_request(
signoz,
token,
start_ms=int((now - timedelta(minutes=5)).timestamp() * 1000),
end_ms=int(now.timestamp() * 1000),
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression=expression, limit=100)],
)
assert response.status_code == HTTPStatus.BAD_REQUEST, response.text
assert "invalid search scope" in response.text
# The querier gates search() on EXPLAIN ESTIMATE against search_max_scan_rows_json_body on
# this path (50000 here, see conftest.py). Both recoveries the advisory suggests — a
# selective filter, a narrower range — are exercised below.
def test_search_cost_guard_trips_then_passes_with_filter(
signoz_search_scan_budget: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
) -> None:
"""A broad search() over ~61000 logs exceeds the 50000-row budget and is rejected;
the same search narrowed to the 1000-log 'checkout' service scans under budget and
succeeds — the advisory's "add a more selective filter" made real."""
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
# 60000 'catalog' logs now + 1000 'checkout' logs ~45m back, in an earlier ts_bucket, so
# the checkout fingerprint owns its own marks and a resource filter on it prunes the scan.
logs = [Logs(timestamp=now - timedelta(seconds=1 + i % 30), resources={"service.name": "catalog"}, body="log line") for i in range(60000)]
logs += [Logs(timestamp=now - timedelta(minutes=45, seconds=i % 30), resources={"service.name": "checkout"}, body="log line") for i in range(1000)]
insert_logs(logs)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
start_ms = int((now - timedelta(minutes=60)).timestamp() * 1000)
end_ms = int((now + timedelta(minutes=1)).timestamp() * 1000)
def run(expression: str):
return make_query_request(
signoz_search_scan_budget,
token,
start_ms=start_ms,
end_ms=end_ms,
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression=expression, order=[build_order_by("timestamp", "desc")], limit=100)],
)
# Broad search over the whole range is over budget -> rejected before executing.
over_budget = run("search('log')")
assert over_budget.status_code == HTTPStatus.BAD_REQUEST, over_budget.text
assert "over the per-shard limit" in over_budget.text
# The advisory leads the suggestions, then how to get under budget.
assert "runs across all fields" in over_budget.text
assert "Narrow the time range or add a more selective filter." in over_budget.text
# Adding a selective resource filter prunes the scan under budget -> runs.
within_budget = run("search('log') AND resource.service.name = 'checkout'")
assert within_budget.status_code == HTTPStatus.OK, within_budget.text
assert len(get_rows(within_budget)) > 0
def test_search_cost_guard_passes_with_narrower_time_range(
signoz_search_scan_budget: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
) -> None:
"""A steady stream of ~80000 logs (~4/sec over the last ~5.5h). A search() over the
whole window is over the 50000-row budget and rejected; the same search over the last
15 minutes scans only a few thousand rows and runs — the advisory's "narrow the time
range" made real."""
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
# ~4 logs/sec, oldest ~5.5h back, newest ~now.
logs = [Logs(timestamp=now - timedelta(seconds=i // 4), resources={"service.name": "app"}, body="log line") for i in range(80000)]
insert_logs(logs)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
def run(lookback_minutes: int):
return make_query_request(
signoz_search_scan_budget,
token,
start_ms=int((now - timedelta(minutes=lookback_minutes)).timestamp() * 1000),
end_ms=int((now + timedelta(minutes=1)).timestamp() * 1000),
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression="search('log')", order=[build_order_by("timestamp", "desc")], limit=100)],
)
# The last 6 hours cover the whole stream (~80000) -> over budget -> rejected.
wide = run(360)
assert wide.status_code == HTTPStatus.BAD_REQUEST, wide.text
assert "over the per-shard limit" in wide.text
# The last 15 minutes hold only ~3600 logs -> under budget -> runs, returning rows.
narrow = run(15)
assert narrow.status_code == HTTPStatus.OK, narrow.text
assert len(get_rows(narrow)) > 0

View File

@@ -55,3 +55,34 @@ def signoz_json_body(
"SIGNOZ_FLAGGER_CONFIG_BOOLEAN_USE__JSON__BODY": True,
},
)
@pytest.fixture(name="signoz_search_scan_budget", scope="package")
def signoz_search_scan_budget(
network: Network,
migrator: types.Operation, # pylint: disable=unused-argument
zeus: types.TestContainerDocker,
gateway: types.TestContainerDocker,
sqlstore: types.TestContainerSQL,
clickhouse: types.TestContainerClickhouse,
request: pytest.FixtureRequest,
pytestconfig: pytest.Config,
) -> types.SigNoz:
"""The querierlogs budget instance over body_v2: same 50000 rows, but on
search_max_scan_rows_json_body. search_max_scan_rows keeps its 60M default, so only the
body_v2 budget can trip here. Shares the default instance's sqlstore + clickhouse, so
the same admin token and seeded logs work against it."""
return create_signoz(
network=network,
zeus=zeus,
gateway=gateway,
sqlstore=sqlstore,
clickhouse=clickhouse,
request=request,
pytestconfig=pytestconfig,
cache_key="signoz-json-body-search-scan-budget-50k",
env_overrides={
"SIGNOZ_FLAGGER_CONFIG_BOOLEAN_USE__JSON__BODY": True,
"SIGNOZ_QUERIER_SEARCH__MAX__SCAN__ROWS__JSON__BODY": 50000,
},
)

View File

@@ -0,0 +1,212 @@
from collections import namedtuple
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
from http import HTTPStatus
import pytest
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.logs import Logs
from fixtures.querier import build_order_by, build_raw_query, get_column_data_from_response, get_rows, make_query_request
# search(): keyless fans across every field; scoped search('term', <ctx>...) narrows to
# the named contexts (body/attribute/resource/log). Flag off here, so body matches the
# `body` String column (querier_json_body mirrors this over body_v2).
Bodies = namedtuple("Bodies", ["a", "b", "c", "d"])
@pytest.mark.parametrize(
"expression,expected",
[
# ── keyless: fans across every field ────────────────────────────────
pytest.param("search('login')", lambda b: {b.a}, id="keyless_body"),
pytest.param("search('checkout')", lambda b: {b.a, b.d}, id="keyless_body_and_resource"),
pytest.param("search('useast')", lambda b: {b.a, b.c}, id="keyless_resource_value"),
pytest.param("search('acme')", lambda b: {b.a, b.c}, id="keyless_attribute_value"),
pytest.param("search('tenant')", lambda b: {b.a, b.b, b.c, b.d}, id="keyless_attribute_key"),
pytest.param("search('error')", lambda b: {b.b, b.d}, id="keyless_severity_case_insensitive"),
pytest.param("search('CHECKOUT')", lambda b: {b.a, b.d}, id="keyless_term_case_insensitive"),
# ── scoped: narrows to one context ──────────────────────────────────
pytest.param("search('login', body)", lambda b: {b.a}, id="scope_body"),
pytest.param("search('login', 'body')", lambda b: {b.a}, id="scope_body_quoted"),
pytest.param("search('checkout', body)", lambda b: {b.a}, id="scope_body_excludes_resource"),
pytest.param("search('checkout', resource)", lambda b: {b.a, b.d}, id="scope_resource"),
pytest.param("search('acme', attribute)", lambda b: {b.a, b.c}, id="scope_attribute"),
pytest.param("search('error', log)", lambda b: {b.b, b.d}, id="scope_log_severity"),
pytest.param("search('acme', body)", lambda b: set(), id="scope_body_no_match"),
pytest.param("search('checkout', attribute)", lambda b: set(), id="scope_attribute_no_match"),
# ── multiple scopes: union of the named contexts ────────────────────
pytest.param("search('login', body, resource)", lambda b: {b.a}, id="scopes_body_resource_body_only"),
pytest.param("search('checkout', body, resource)", lambda b: {b.a, b.d}, id="scopes_body_resource_union"),
# ── composition with boolean / field filters ────────────────────────
pytest.param("NOT search('login')", lambda b: {b.b, b.c, b.d}, id="negated"),
pytest.param("search('useast') AND severity_text = 'INFO'", lambda b: {b.a}, id="and_field_filter"),
],
)
def test_search(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
expression: str,
expected: Callable[[Bodies], set[str]],
) -> None:
"""Four self-naming logs, each with a token planted in a distinct place (body,
resource, attribute, severity), assert search() reaches exactly the right ones."""
body = Bodies(
a="alpha checkout login ok", # service checkout / region useast / tenant acme / INFO
b="bravo declined", # service payment / region euwest / tenant globex / ERROR
c="charlie miss", # service cart / region useast / tenant acme / WARN
d="delta slow", # service checkout / region apac / tenant initech / ERROR
)
# (body, resources, attributes, severity_text)
specs = [
(body.a, {"service.name": "checkout", "region": "useast"}, {"tenant": "acme"}, "INFO"),
(body.b, {"service.name": "payment", "region": "euwest"}, {"tenant": "globex"}, "ERROR"),
(body.c, {"service.name": "cart", "region": "useast"}, {"tenant": "acme"}, "WARN"),
(body.d, {"service.name": "checkout", "region": "apac"}, {"tenant": "initech"}, "ERROR"),
]
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
logs = [Logs(timestamp=now - timedelta(seconds=i + 1), resources=res, attributes=attrs, body=b, severity_text=sev) for i, (b, res, attrs, sev) in enumerate(specs)]
insert_logs(logs)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = make_query_request(
signoz,
token,
start_ms=int((now - timedelta(minutes=5)).timestamp() * 1000),
end_ms=int(now.timestamp() * 1000),
request_type="raw",
queries=[
build_raw_query(
"A",
"logs",
filter_expression=expression,
order=[build_order_by("timestamp", "desc"), build_order_by("id", "desc")],
limit=100,
)
],
)
assert response.status_code == HTTPStatus.OK, response.text
assert response.json()["status"] == "success"
assert set(get_column_data_from_response(response.json(), "body")) == expected(body)
@pytest.mark.parametrize(
"expression",
[
pytest.param("search('login', bogus)", id="unknown_scope_word"),
pytest.param("search('login', body.message)", id="qualified_field_not_a_scope"),
],
)
def test_search_invalid_scope_rejected(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
expression: str,
) -> None:
"""A scope that is not a field context (an unknown word, or a qualified
`context.field`) is rejected at build time with a 400."""
now = datetime.now(tz=UTC)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
response = make_query_request(
signoz,
token,
start_ms=int((now - timedelta(minutes=5)).timestamp() * 1000),
end_ms=int(now.timestamp() * 1000),
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression=expression, limit=100)],
)
assert response.status_code == HTTPStatus.BAD_REQUEST, response.text
assert "invalid search scope" in response.text
# The querier gates search() on EXPLAIN ESTIMATE against search_max_scan_rows (50000 here,
# see conftest.py). Both recoveries the advisory suggests — a selective filter, a narrower
# range — are exercised below.
def test_search_cost_guard_trips_then_passes_with_filter(
signoz_search_scan_budget: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
) -> None:
"""A broad search() over ~61000 logs exceeds the 50000-row budget and is rejected;
the same search narrowed to the 1000-log 'checkout' service scans under budget and
succeeds — the advisory's "add a more selective filter" made real."""
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
# 60000 'catalog' logs now + 1000 'checkout' logs ~45m back, in an earlier ts_bucket, so
# the checkout fingerprint owns its own marks and a resource filter on it prunes the scan.
logs = [Logs(timestamp=now - timedelta(seconds=1 + i % 30), resources={"service.name": "catalog"}, body="log line") for i in range(60000)]
logs += [Logs(timestamp=now - timedelta(minutes=45, seconds=i % 30), resources={"service.name": "checkout"}, body="log line") for i in range(1000)]
insert_logs(logs)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
start_ms = int((now - timedelta(minutes=60)).timestamp() * 1000)
end_ms = int((now + timedelta(minutes=1)).timestamp() * 1000)
def run(expression: str):
return make_query_request(
signoz_search_scan_budget,
token,
start_ms=start_ms,
end_ms=end_ms,
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression=expression, order=[build_order_by("timestamp", "desc")], limit=100)],
)
# Broad search over the whole range is over budget -> rejected before executing.
over_budget = run("search('log')")
assert over_budget.status_code == HTTPStatus.BAD_REQUEST, over_budget.text
assert "over the per-shard limit" in over_budget.text
# The advisory leads the suggestions, then how to get under budget.
assert "runs across all fields" in over_budget.text
assert "Narrow the time range or add a more selective filter." in over_budget.text
# Adding a selective resource filter prunes the scan under budget -> runs.
within_budget = run("search('log') AND resource.service.name = 'checkout'")
assert within_budget.status_code == HTTPStatus.OK, within_budget.text
assert len(get_rows(within_budget)) > 0
def test_search_cost_guard_passes_with_narrower_time_range(
signoz_search_scan_budget: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
) -> None:
"""A steady stream of ~80000 logs (~4/sec over the last ~5.5h). A search() over the
whole window is over the 50000-row budget and rejected; the same search over the last
15 minutes scans only a few thousand rows and runs — the advisory's "narrow the time
range" made real."""
now = datetime.now(tz=UTC).replace(second=0, microsecond=0)
# ~4 logs/sec, oldest ~5.5h back, newest ~now.
logs = [Logs(timestamp=now - timedelta(seconds=i // 4), resources={"service.name": "app"}, body="log line") for i in range(80000)]
insert_logs(logs)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
def run(lookback_minutes: int):
return make_query_request(
signoz_search_scan_budget,
token,
start_ms=int((now - timedelta(minutes=lookback_minutes)).timestamp() * 1000),
end_ms=int((now + timedelta(minutes=1)).timestamp() * 1000),
request_type="raw",
queries=[build_raw_query("A", "logs", filter_expression="search('log')", order=[build_order_by("timestamp", "desc")], limit=100)],
)
# The last 6 hours cover the whole stream (~80000) -> over budget -> rejected.
wide = run(360)
assert wide.status_code == HTTPStatus.BAD_REQUEST, wide.text
assert "over the per-shard limit" in wide.text
# The last 15 minutes hold only ~3600 logs -> under budget -> runs, returning rows.
narrow = run(15)
assert narrow.status_code == HTTPStatus.OK, narrow.text
assert len(get_rows(narrow)) > 0

View File

@@ -0,0 +1,34 @@
import pytest
from testcontainers.core.container import Network
from fixtures import types
from fixtures.signoz import create_signoz
@pytest.fixture(name="signoz_search_scan_budget", scope="package")
def signoz_search_scan_budget(
network: Network,
migrator: types.Operation, # pylint: disable=unused-argument
zeus: types.TestContainerDocker,
gateway: types.TestContainerDocker,
sqlstore: types.TestContainerSQL,
clickhouse: types.TestContainerClickhouse,
request: pytest.FixtureRequest,
pytestconfig: pytest.Config,
) -> types.SigNoz:
"""SigNoz with a low search_max_scan_rows (50000) so a broad search() trips the cost
guard while a selective one stays under it. Shares the default instance's sqlstore +
clickhouse, so the same admin token and seeded logs work against it."""
return create_signoz(
network=network,
zeus=zeus,
gateway=gateway,
sqlstore=sqlstore,
clickhouse=clickhouse,
request=request,
pytestconfig=pytestconfig,
cache_key="signoz-search-scan-budget-50k",
env_overrides={
"SIGNOZ_QUERIER_SEARCH__MAX__SCAN__ROWS": 50000,
},
)

View File

@@ -75,3 +75,7 @@ ignore = [
[tool.ruff.format]
# Defaults align with black (double quotes, 4-space indent).
[tool.ruff.lint.per-file-ignores]
"fixtures/notification_channel.py" = ["E501"]
"integration/tests/alertmanager/*" = ["E501"]