mirror of
https://github.com/SigNoz/signoz.git
synced 2026-05-29 21:30:29 +01:00
Compare commits
10 Commits
fix/alert-
...
ns/flamegr
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
1064defd60 | ||
|
|
f9222c9930 | ||
|
|
511dd0add8 | ||
|
|
4ef8500b28 | ||
|
|
168b2eaa9c | ||
|
|
6b613f18a3 | ||
|
|
1b0447181d | ||
|
|
20edff4771 | ||
|
|
2048ef3d2f | ||
|
|
53c551359e |
@@ -434,6 +434,17 @@ tracedetail:
|
||||
max_depth_to_auto_expand: 5
|
||||
# Threshold below which all spans are returned without windowing.
|
||||
max_limit_to_select_all_spans: 10000
|
||||
flamegraph:
|
||||
# Maximum number of BFS depth levels included in a windowed response.
|
||||
max_selected_levels: 50
|
||||
# Maximum spans per level before sampling is applied.
|
||||
max_spans_per_level: 100
|
||||
# Number of highest-latency spans always included when sampling a level.
|
||||
sampling_top_latency_count: 5
|
||||
# Number of timestamp buckets used for uniform sampling within a level.
|
||||
sampling_bucket_count: 50
|
||||
# Threshold below which all spans are returned without windowing or sampling.
|
||||
select_all_spans_limit: 100000
|
||||
|
||||
##################### Authz #################################
|
||||
authz:
|
||||
|
||||
@@ -5663,6 +5663,58 @@ components:
|
||||
- attribute
|
||||
- resource
|
||||
type: string
|
||||
SpantypesFlamegraphSpan:
|
||||
properties:
|
||||
attributes:
|
||||
additionalProperties: {}
|
||||
type: object
|
||||
durationNano:
|
||||
minimum: 0
|
||||
type: integer
|
||||
event:
|
||||
items:
|
||||
$ref: '#/components/schemas/SpantypesEvent'
|
||||
nullable: true
|
||||
type: array
|
||||
hasError:
|
||||
type: boolean
|
||||
level:
|
||||
format: int64
|
||||
type: integer
|
||||
name:
|
||||
type: string
|
||||
parentSpanId:
|
||||
type: string
|
||||
resource:
|
||||
additionalProperties:
|
||||
type: string
|
||||
type: object
|
||||
serviceName:
|
||||
type: string
|
||||
spanId:
|
||||
type: string
|
||||
timestamp:
|
||||
minimum: 0
|
||||
type: integer
|
||||
type: object
|
||||
SpantypesGettableFlamegraphTrace:
|
||||
properties:
|
||||
endTimestampMillis:
|
||||
format: int64
|
||||
type: integer
|
||||
hasMore:
|
||||
type: boolean
|
||||
spans:
|
||||
items:
|
||||
items:
|
||||
$ref: '#/components/schemas/SpantypesFlamegraphSpan'
|
||||
type: array
|
||||
nullable: true
|
||||
type: array
|
||||
startTimestampMillis:
|
||||
format: int64
|
||||
type: integer
|
||||
type: object
|
||||
SpantypesGettableSpanMapperGroups:
|
||||
properties:
|
||||
items:
|
||||
@@ -5710,6 +5762,15 @@ components:
|
||||
nullable: true
|
||||
type: array
|
||||
type: object
|
||||
SpantypesPostableFlamegraph:
|
||||
properties:
|
||||
selectFields:
|
||||
items:
|
||||
$ref: '#/components/schemas/TelemetrytypesTelemetryFieldKey'
|
||||
type: array
|
||||
selectedSpanId:
|
||||
type: string
|
||||
type: object
|
||||
SpantypesPostableSpanMapper:
|
||||
properties:
|
||||
config:
|
||||
@@ -18878,6 +18939,75 @@ paths:
|
||||
summary: Put profile in Zeus for a deployment.
|
||||
tags:
|
||||
- zeus
|
||||
/api/v3/traces/{traceID}/flamegraph:
|
||||
post:
|
||||
deprecated: false
|
||||
description: Returns the flamegraph view of spans for a given trace ID.
|
||||
operationId: GetFlamegraph
|
||||
parameters:
|
||||
- in: path
|
||||
name: traceID
|
||||
required: true
|
||||
schema:
|
||||
type: string
|
||||
requestBody:
|
||||
content:
|
||||
application/json:
|
||||
schema:
|
||||
$ref: '#/components/schemas/SpantypesPostableFlamegraph'
|
||||
responses:
|
||||
"200":
|
||||
content:
|
||||
application/json:
|
||||
schema:
|
||||
properties:
|
||||
data:
|
||||
$ref: '#/components/schemas/SpantypesGettableFlamegraphTrace'
|
||||
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:
|
||||
- VIEWER
|
||||
- tokenizer:
|
||||
- VIEWER
|
||||
summary: Get flamegraph view for a trace
|
||||
tags:
|
||||
- tracedetail
|
||||
/api/v3/traces/{traceID}/waterfall:
|
||||
post:
|
||||
deprecated: false
|
||||
|
||||
@@ -6689,6 +6689,81 @@ export enum SpantypesFieldContextDTO {
|
||||
attribute = 'attribute',
|
||||
resource = 'resource',
|
||||
}
|
||||
export type SpantypesFlamegraphSpanDTOAttributes = { [key: string]: unknown };
|
||||
|
||||
export type SpantypesFlamegraphSpanDTOResource = { [key: string]: string };
|
||||
|
||||
export interface SpantypesFlamegraphSpanDTO {
|
||||
/**
|
||||
* @type object
|
||||
*/
|
||||
attributes?: SpantypesFlamegraphSpanDTOAttributes;
|
||||
/**
|
||||
* @type integer
|
||||
* @minimum 0
|
||||
*/
|
||||
durationNano?: number;
|
||||
/**
|
||||
* @type array,null
|
||||
*/
|
||||
event?: SpantypesEventDTO[] | null;
|
||||
/**
|
||||
* @type boolean
|
||||
*/
|
||||
hasError?: boolean;
|
||||
/**
|
||||
* @type integer
|
||||
* @format int64
|
||||
*/
|
||||
level?: number;
|
||||
/**
|
||||
* @type string
|
||||
*/
|
||||
name?: string;
|
||||
/**
|
||||
* @type string
|
||||
*/
|
||||
parentSpanId?: string;
|
||||
/**
|
||||
* @type object
|
||||
*/
|
||||
resource?: SpantypesFlamegraphSpanDTOResource;
|
||||
/**
|
||||
* @type string
|
||||
*/
|
||||
serviceName?: string;
|
||||
/**
|
||||
* @type string
|
||||
*/
|
||||
spanId?: string;
|
||||
/**
|
||||
* @type integer
|
||||
* @minimum 0
|
||||
*/
|
||||
timestamp?: number;
|
||||
}
|
||||
|
||||
export interface SpantypesGettableFlamegraphTraceDTO {
|
||||
/**
|
||||
* @type integer
|
||||
* @format int64
|
||||
*/
|
||||
endTimestampMillis?: number;
|
||||
/**
|
||||
* @type boolean
|
||||
*/
|
||||
hasMore?: boolean;
|
||||
/**
|
||||
* @type array,null
|
||||
*/
|
||||
spans?: SpantypesFlamegraphSpanDTO[][] | null;
|
||||
/**
|
||||
* @type integer
|
||||
* @format int64
|
||||
*/
|
||||
startTimestampMillis?: number;
|
||||
}
|
||||
|
||||
export type SpantypesSpanMapperGroupConditionDTOAnyOf = {
|
||||
/**
|
||||
* @type array,null
|
||||
@@ -6964,6 +7039,17 @@ export interface SpantypesGettableWaterfallTraceDTO {
|
||||
uncollapsedSpans?: string[] | null;
|
||||
}
|
||||
|
||||
export interface SpantypesPostableFlamegraphDTO {
|
||||
/**
|
||||
* @type array
|
||||
*/
|
||||
selectFields?: TelemetrytypesTelemetryFieldKeyDTO[];
|
||||
/**
|
||||
* @type string
|
||||
*/
|
||||
selectedSpanId?: string;
|
||||
}
|
||||
|
||||
export enum SpantypesSpanMapperOperationDTO {
|
||||
move = 'move',
|
||||
copy = 'copy',
|
||||
@@ -9221,6 +9307,17 @@ export type GetHosts200 = {
|
||||
status: string;
|
||||
};
|
||||
|
||||
export type GetFlamegraphPathParameters = {
|
||||
traceID: string;
|
||||
};
|
||||
export type GetFlamegraph200 = {
|
||||
data: SpantypesGettableFlamegraphTraceDTO;
|
||||
/**
|
||||
* @type string
|
||||
*/
|
||||
status: string;
|
||||
};
|
||||
|
||||
export type GetWaterfallPathParameters = {
|
||||
traceID: string;
|
||||
};
|
||||
|
||||
@@ -12,17 +12,119 @@ import type {
|
||||
} from 'react-query';
|
||||
|
||||
import type {
|
||||
GetFlamegraph200,
|
||||
GetFlamegraphPathParameters,
|
||||
GetWaterfall200,
|
||||
GetWaterfallPathParameters,
|
||||
GetWaterfallV4200,
|
||||
GetWaterfallV4PathParameters,
|
||||
RenderErrorResponseDTO,
|
||||
SpantypesPostableFlamegraphDTO,
|
||||
SpantypesPostableWaterfallDTO,
|
||||
} from '../sigNoz.schemas';
|
||||
|
||||
import { GeneratedAPIInstance } from '../../../generatedAPIInstance';
|
||||
import type { ErrorType, BodyType } from '../../../generatedAPIInstance';
|
||||
|
||||
/**
|
||||
* Returns the flamegraph view of spans for a given trace ID.
|
||||
* @summary Get flamegraph view for a trace
|
||||
*/
|
||||
export const getFlamegraph = (
|
||||
{ traceID }: GetFlamegraphPathParameters,
|
||||
spantypesPostableFlamegraphDTO?: BodyType<SpantypesPostableFlamegraphDTO>,
|
||||
signal?: AbortSignal,
|
||||
) => {
|
||||
return GeneratedAPIInstance<GetFlamegraph200>({
|
||||
url: `/api/v3/traces/${traceID}/flamegraph`,
|
||||
method: 'POST',
|
||||
headers: { 'Content-Type': 'application/json' },
|
||||
data: spantypesPostableFlamegraphDTO,
|
||||
signal,
|
||||
});
|
||||
};
|
||||
|
||||
export const getGetFlamegraphMutationOptions = <
|
||||
TError = ErrorType<RenderErrorResponseDTO>,
|
||||
TContext = unknown,
|
||||
>(options?: {
|
||||
mutation?: UseMutationOptions<
|
||||
Awaited<ReturnType<typeof getFlamegraph>>,
|
||||
TError,
|
||||
{
|
||||
pathParams: GetFlamegraphPathParameters;
|
||||
data?: BodyType<SpantypesPostableFlamegraphDTO>;
|
||||
},
|
||||
TContext
|
||||
>;
|
||||
}): UseMutationOptions<
|
||||
Awaited<ReturnType<typeof getFlamegraph>>,
|
||||
TError,
|
||||
{
|
||||
pathParams: GetFlamegraphPathParameters;
|
||||
data?: BodyType<SpantypesPostableFlamegraphDTO>;
|
||||
},
|
||||
TContext
|
||||
> => {
|
||||
const mutationKey = ['getFlamegraph'];
|
||||
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 getFlamegraph>>,
|
||||
{
|
||||
pathParams: GetFlamegraphPathParameters;
|
||||
data?: BodyType<SpantypesPostableFlamegraphDTO>;
|
||||
}
|
||||
> = (props) => {
|
||||
const { pathParams, data } = props ?? {};
|
||||
|
||||
return getFlamegraph(pathParams, data);
|
||||
};
|
||||
|
||||
return { mutationFn, ...mutationOptions };
|
||||
};
|
||||
|
||||
export type GetFlamegraphMutationResult = NonNullable<
|
||||
Awaited<ReturnType<typeof getFlamegraph>>
|
||||
>;
|
||||
export type GetFlamegraphMutationBody =
|
||||
| BodyType<SpantypesPostableFlamegraphDTO>
|
||||
| undefined;
|
||||
export type GetFlamegraphMutationError = ErrorType<RenderErrorResponseDTO>;
|
||||
|
||||
/**
|
||||
* @summary Get flamegraph view for a trace
|
||||
*/
|
||||
export const useGetFlamegraph = <
|
||||
TError = ErrorType<RenderErrorResponseDTO>,
|
||||
TContext = unknown,
|
||||
>(options?: {
|
||||
mutation?: UseMutationOptions<
|
||||
Awaited<ReturnType<typeof getFlamegraph>>,
|
||||
TError,
|
||||
{
|
||||
pathParams: GetFlamegraphPathParameters;
|
||||
data?: BodyType<SpantypesPostableFlamegraphDTO>;
|
||||
},
|
||||
TContext
|
||||
>;
|
||||
}): UseMutationResult<
|
||||
Awaited<ReturnType<typeof getFlamegraph>>,
|
||||
TError,
|
||||
{
|
||||
pathParams: GetFlamegraphPathParameters;
|
||||
data?: BodyType<SpantypesPostableFlamegraphDTO>;
|
||||
},
|
||||
TContext
|
||||
> => {
|
||||
return useMutation(getGetFlamegraphMutationOptions(options));
|
||||
};
|
||||
/**
|
||||
* Returns the waterfall view of spans for a given trace ID with tree structure, metadata, and windowed pagination
|
||||
* @summary Get waterfall view for a trace
|
||||
|
||||
@@ -48,5 +48,23 @@ func (provider *provider) addTraceDetailRoutes(router *mux.Router) error {
|
||||
return err
|
||||
}
|
||||
|
||||
if err := router.Handle("/api/v3/traces/{traceID}/flamegraph", handler.New(
|
||||
provider.authzMiddleware.ViewAccess(provider.traceDetailHandler.GetFlamegraph),
|
||||
handler.OpenAPIDef{
|
||||
ID: "GetFlamegraph",
|
||||
Tags: []string{"tracedetail"},
|
||||
Summary: "Get flamegraph view for a trace",
|
||||
Description: "Returns the flamegraph view of spans for a given trace ID.",
|
||||
Request: new(spantypes.PostableFlamegraph),
|
||||
RequestContentType: "application/json",
|
||||
Response: new(spantypes.GettableFlamegraphTrace),
|
||||
ResponseContentType: "application/json",
|
||||
SuccessStatusCode: http.StatusOK,
|
||||
ErrorStatusCodes: []int{http.StatusBadRequest, http.StatusNotFound},
|
||||
SecuritySchemes: newSecuritySchemes(types.RoleViewer),
|
||||
},
|
||||
)).Methods(http.MethodPost).GetError(); err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
@@ -6,7 +6,16 @@ import (
|
||||
)
|
||||
|
||||
type Config struct {
|
||||
Waterfall WaterfallConfig `mapstructure:"waterfall"`
|
||||
Waterfall WaterfallConfig `mapstructure:"waterfall"`
|
||||
Flamegraph FlamegraphConfig `mapstructure:"flamegraph"`
|
||||
}
|
||||
|
||||
type FlamegraphConfig struct {
|
||||
MaxSelectedLevels int `mapstructure:"max_selected_levels"`
|
||||
MaxSpansPerLevel int `mapstructure:"max_spans_per_level"`
|
||||
SamplingTopLatencySpansCount int `mapstructure:"sampling_top_latency_count"`
|
||||
SamplingBucketCount int `mapstructure:"sampling_bucket_count"`
|
||||
SelectAllSpansLimit uint `mapstructure:"select_all_spans_limit"`
|
||||
}
|
||||
|
||||
type WaterfallConfig struct {
|
||||
@@ -29,6 +38,13 @@ func newConfig() factory.Config {
|
||||
MaxDepthToAutoExpand: 5,
|
||||
MaxLimitToSelectAllSpans: 10_000,
|
||||
},
|
||||
Flamegraph: FlamegraphConfig{
|
||||
MaxSelectedLevels: 50,
|
||||
MaxSpansPerLevel: 100,
|
||||
SamplingTopLatencySpansCount: 5,
|
||||
SamplingBucketCount: 50,
|
||||
SelectAllSpansLimit: 100_000,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
@@ -45,5 +61,25 @@ func (c Config) Validate() error {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"tracedetail.waterfall.max_limit_to_select_all_spans must be positive")
|
||||
}
|
||||
if c.Flamegraph.MaxSelectedLevels <= 0 {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"tracedetail.flamegraph.level_limit must be positive, got %d", c.Flamegraph.MaxSelectedLevels)
|
||||
}
|
||||
if c.Flamegraph.MaxSpansPerLevel <= 0 {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"tracedetail.flamegraph.spans_per_level must be positive, got %d", c.Flamegraph.MaxSpansPerLevel)
|
||||
}
|
||||
if c.Flamegraph.SamplingTopLatencySpansCount < 0 {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"tracedetail.flamegraph.top_latency_count cannot be negative, got %d", c.Flamegraph.SamplingTopLatencySpansCount)
|
||||
}
|
||||
if c.Flamegraph.SamplingBucketCount <= 0 {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"tracedetail.flamegraph.bucket_count must be positive, got %d", c.Flamegraph.SamplingBucketCount)
|
||||
}
|
||||
if c.Flamegraph.SelectAllSpansLimit == 0 {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"tracedetail.flamegraph.max_limit_to_select_all_spans must be positive")
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
@@ -59,3 +59,19 @@ func (h *handler) GetWaterfallV4(rw http.ResponseWriter, r *http.Request) {
|
||||
|
||||
render.Success(rw, http.StatusOK, result)
|
||||
}
|
||||
|
||||
func (h *handler) GetFlamegraph(rw http.ResponseWriter, r *http.Request) {
|
||||
req := new(spantypes.PostableFlamegraph)
|
||||
if err := binding.JSON.BindBody(r.Body, req); err != nil {
|
||||
render.Error(rw, err)
|
||||
return
|
||||
}
|
||||
|
||||
result, err := h.module.GetFlamegraph(r.Context(), mux.Vars(r)["traceID"], req)
|
||||
if err != nil {
|
||||
render.Error(rw, err)
|
||||
return
|
||||
}
|
||||
|
||||
render.Success(rw, http.StatusOK, result)
|
||||
}
|
||||
|
||||
@@ -46,6 +46,33 @@ func (m *module) GetWaterfall(ctx context.Context, traceID string, req *spantype
|
||||
return spantypes.NewGettableWaterfallTrace(waterfallTrace, selectedSpans, uncollapsedSpans, selectedAllSpans, aggregationResults), nil
|
||||
}
|
||||
|
||||
// GetWaterfallV4 is the OOM-safe V4 waterfall.
|
||||
// For large traces (NumSpans > effectiveLimit) it uses a two-step fetch:
|
||||
// minimal fields for all spans to build the tree, then full fields for the
|
||||
// visible window only. Aggregations are not returned.
|
||||
func (m *module) GetWaterfallV4(ctx context.Context, traceID string, selectedSpanID string, uncollapsedSpans []string, selectAllLimit uint) (*spantypes.GettableWaterfallTrace, error) {
|
||||
summary, err := m.store.GetTraceSummary(ctx, traceID)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
effectiveLimit := min(selectAllLimit, m.config.Waterfall.MaxLimitToSelectAllSpans)
|
||||
if summary.NumSpans > uint64(effectiveLimit) {
|
||||
return m.getWindowedWaterfall(ctx, traceID, selectedSpanID, uncollapsedSpans, summary.Start, summary.End)
|
||||
}
|
||||
return m.getFullWaterfall(ctx, traceID, summary)
|
||||
}
|
||||
|
||||
func (m *module) GetFlamegraph(ctx context.Context, traceID string, req *spantypes.PostableFlamegraph) (*spantypes.GettableFlamegraphTrace, error) {
|
||||
summary, err := m.store.GetTraceSummary(ctx, traceID)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if summary.NumSpans <= uint64(m.config.Flamegraph.SelectAllSpansLimit) {
|
||||
return m.getFullFlamegraph(ctx, traceID, summary)
|
||||
}
|
||||
return m.getWindowedFlamegraph(ctx, traceID, req.SelectedSpanID, summary)
|
||||
}
|
||||
|
||||
// getTraceData fetches all spans for a trace and builds the WaterfallTrace.
|
||||
func (m *module) getTraceData(ctx context.Context, traceID string) (*spantypes.WaterfallTrace, error) {
|
||||
summary, err := m.store.GetTraceSummary(ctx, traceID)
|
||||
@@ -69,22 +96,6 @@ func (m *module) getTraceData(ctx context.Context, traceID string) (*spantypes.W
|
||||
return spantypes.NewWaterfallTraceFromSpans(nodes), nil
|
||||
}
|
||||
|
||||
// GetWaterfallV4 is the OOM-safe V4 waterfall.
|
||||
// For large traces (NumSpans > effectiveLimit) it uses a two-step fetch:
|
||||
// minimal fields for all spans to build the tree, then full fields for the
|
||||
// visible window only. Aggregations are not returned.
|
||||
func (m *module) GetWaterfallV4(ctx context.Context, traceID string, selectedSpanID string, uncollapsedSpans []string, selectAllLimit uint) (*spantypes.GettableWaterfallTrace, error) {
|
||||
summary, err := m.store.GetTraceSummary(ctx, traceID)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
effectiveLimit := min(selectAllLimit, m.config.Waterfall.MaxLimitToSelectAllSpans)
|
||||
if summary.NumSpans > uint64(effectiveLimit) {
|
||||
return m.getWindowedWaterfall(ctx, traceID, selectedSpanID, uncollapsedSpans, summary.Start, summary.End)
|
||||
}
|
||||
return m.getFullWaterfall(ctx, traceID, summary)
|
||||
}
|
||||
|
||||
func (m *module) getFullWaterfall(ctx context.Context, traceID string, summary *spantypes.TraceSummary) (*spantypes.GettableWaterfallTrace, error) {
|
||||
spanItems, err := m.store.GetTraceSpans(ctx, traceID, summary)
|
||||
if err != nil {
|
||||
@@ -145,3 +156,50 @@ func (m *module) getWindowedWaterfall(ctx context.Context, traceID, selectedSpan
|
||||
waterfallTrace, selectedSpans, uncollapsedSpans, false, nil,
|
||||
), nil
|
||||
}
|
||||
|
||||
func (m *module) getFullFlamegraph(ctx context.Context, traceID string, summary *spantypes.TraceSummary) (*spantypes.GettableFlamegraphTrace, error) {
|
||||
fullSpans, err := m.store.GetTraceSpans(ctx, traceID, summary)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if len(fullSpans) == 0 {
|
||||
return nil, spantypes.ErrTraceNotFound
|
||||
}
|
||||
flamegraphTrace := spantypes.NewFlamegraphTraceFromStorable(fullSpans)
|
||||
return spantypes.NewGettableFlamegraphTrace(
|
||||
flamegraphTrace.GetAllLevels(),
|
||||
summary.Start.UnixMilli(), summary.End.UnixMilli(), false,
|
||||
), nil
|
||||
}
|
||||
|
||||
// getWindowedFlamegraph returns a window of a max levels and max sampled spans per level around the selected span.
|
||||
func (m *module) getWindowedFlamegraph(ctx context.Context, traceID, selectedSpanID string, summary *spantypes.TraceSummary) (*spantypes.GettableFlamegraphTrace, error) {
|
||||
minimalSpans, err := m.store.GetMinimalSpans(ctx, traceID, summary.Start, summary.End)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if len(minimalSpans) == 0 {
|
||||
return nil, spantypes.ErrTraceNotFound
|
||||
}
|
||||
|
||||
flamegraphTrace := spantypes.NewFlamegraphTraceFromMinimal(minimalSpans)
|
||||
minimalSpans = nil //nolint:ineffassign,wastedassign // release backing array before further db calls
|
||||
|
||||
cfg := m.config.Flamegraph
|
||||
selectedSpans := flamegraphTrace.GetSelectedLevels(selectedSpanID,
|
||||
cfg.MaxSelectedLevels, cfg.MaxSpansPerLevel, cfg.SamplingTopLatencySpansCount, cfg.SamplingBucketCount)
|
||||
if len(selectedSpans) == 0 {
|
||||
return nil, spantypes.ErrTraceNotFound
|
||||
}
|
||||
|
||||
fullSpans, err := m.store.GetTraceSpansByIDs(ctx, traceID, summary.Start, summary.End,
|
||||
spantypes.FlamegraphWindowSpanIDs(selectedSpans))
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
return spantypes.NewGettableFlamegraphTrace(
|
||||
flamegraphTrace.EnrichSelectedSpans(selectedSpans, fullSpans),
|
||||
summary.Start.UnixMilli(), summary.End.UnixMilli(), true,
|
||||
), nil
|
||||
}
|
||||
|
||||
@@ -11,10 +11,12 @@ import (
|
||||
type Handler interface {
|
||||
GetWaterfall(http.ResponseWriter, *http.Request)
|
||||
GetWaterfallV4(http.ResponseWriter, *http.Request)
|
||||
GetFlamegraph(http.ResponseWriter, *http.Request)
|
||||
}
|
||||
|
||||
// Module defines the business logic for trace detail operations.
|
||||
type Module interface {
|
||||
GetWaterfall(ctx context.Context, traceID string, req *spantypes.PostableWaterfall) (*spantypes.GettableWaterfallTrace, error)
|
||||
GetWaterfallV4(ctx context.Context, traceID string, selectedSpanID string, uncollapsedSpans []string, selectAllLimit uint) (*spantypes.GettableWaterfallTrace, error)
|
||||
GetFlamegraph(ctx context.Context, traceID string, req *spantypes.PostableFlamegraph) (*spantypes.GettableFlamegraphTrace, error)
|
||||
}
|
||||
|
||||
81
pkg/types/spantypes/flamegraph_span.go
Normal file
81
pkg/types/spantypes/flamegraph_span.go
Normal file
@@ -0,0 +1,81 @@
|
||||
package spantypes
|
||||
|
||||
import (
|
||||
"maps"
|
||||
|
||||
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
|
||||
)
|
||||
|
||||
type FlamegraphSpan struct {
|
||||
SpanID string `json:"spanId"`
|
||||
ParentSpanID string `json:"parentSpanId"`
|
||||
Timestamp uint64 `json:"timestamp"`
|
||||
DurationNano uint64 `json:"durationNano"`
|
||||
HasError bool `json:"hasError"`
|
||||
ServiceName string `json:"serviceName"`
|
||||
Name string `json:"name"`
|
||||
Level int64 `json:"level"`
|
||||
Events []Event `json:"event"`
|
||||
Attributes map[string]any `json:"attributes,omitempty"`
|
||||
Resource map[string]string `json:"resource,omitempty"`
|
||||
Children []*FlamegraphSpan `json:"-"` // internal tree use only
|
||||
}
|
||||
|
||||
// FlamegraphLevel groups span IDs at a single level within the selected window.
|
||||
type FlamegraphLevel struct {
|
||||
Level int64
|
||||
SpanIDs []string
|
||||
}
|
||||
|
||||
type PostableFlamegraph struct {
|
||||
SelectedSpanID string `json:"selectedSpanId"`
|
||||
SelectFields []telemetrytypes.TelemetryFieldKey `json:"selectFields,omitempty"`
|
||||
}
|
||||
|
||||
// GettableFlamegraphTrace is the response for the v3 flamegraph API.
|
||||
type GettableFlamegraphTrace struct {
|
||||
Spans [][]*FlamegraphSpan `json:"spans"`
|
||||
StartTimestampMillis int64 `json:"startTimestampMillis"`
|
||||
EndTimestampMillis int64 `json:"endTimestampMillis"`
|
||||
HasMore bool `json:"hasMore"`
|
||||
}
|
||||
|
||||
func NewGettableFlamegraphTrace(spans [][]*FlamegraphSpan, startMs, endMs int64, hasMore bool) *GettableFlamegraphTrace {
|
||||
return &GettableFlamegraphTrace{
|
||||
Spans: spans,
|
||||
StartTimestampMillis: startMs,
|
||||
EndTimestampMillis: endMs,
|
||||
HasMore: hasMore,
|
||||
}
|
||||
}
|
||||
|
||||
func NewFlamegraphSpanFromStorable(s *StorableSpan, level int64) *FlamegraphSpan {
|
||||
resources := make(map[string]string, len(s.ResourcesString))
|
||||
maps.Copy(resources, s.ResourcesString)
|
||||
return &FlamegraphSpan{
|
||||
SpanID: s.SpanID,
|
||||
ParentSpanID: s.ParentSpanID,
|
||||
Timestamp: uint64(s.StartTime.UnixNano()),
|
||||
DurationNano: s.DurationNano,
|
||||
HasError: s.HasError,
|
||||
ServiceName: s.ServiceName,
|
||||
Name: s.Name,
|
||||
Level: level,
|
||||
Events: s.UnmarshalledEvents(),
|
||||
Attributes: s.Attributes(),
|
||||
Resource: resources,
|
||||
}
|
||||
}
|
||||
|
||||
// FlamegraphWindowSpanIDs collects all span IDs from a level window into a flat slice.
|
||||
func FlamegraphWindowSpanIDs(window []FlamegraphLevel) []string {
|
||||
total := 0
|
||||
for _, lvl := range window {
|
||||
total += len(lvl.SpanIDs)
|
||||
}
|
||||
ids := make([]string, 0, total)
|
||||
for _, lvl := range window {
|
||||
ids = append(ids, lvl.SpanIDs...)
|
||||
}
|
||||
return ids
|
||||
}
|
||||
118
pkg/types/spantypes/flamegraph_trace.go
Normal file
118
pkg/types/spantypes/flamegraph_trace.go
Normal file
@@ -0,0 +1,118 @@
|
||||
package spantypes
|
||||
|
||||
import (
|
||||
"sort"
|
||||
)
|
||||
|
||||
// FlamegraphTrace holds the level wise tree built from minimal spans.
|
||||
type FlamegraphTrace struct {
|
||||
roots []*FlamegraphSpan
|
||||
nodeByID map[string]*FlamegraphSpan
|
||||
startTime uint64
|
||||
endTime uint64
|
||||
}
|
||||
|
||||
func NewFlamegraphTraceFromMinimal(spans []MinimalSpan) *FlamegraphTrace {
|
||||
t := &FlamegraphTrace{
|
||||
nodeByID: make(map[string]*FlamegraphSpan, len(spans)),
|
||||
}
|
||||
for i := range spans {
|
||||
node := spans[i].ToFlamegraphSpan()
|
||||
t.updateTimeRange(node.Timestamp, node.DurationNano)
|
||||
t.nodeByID[node.SpanID] = node
|
||||
}
|
||||
t.buildSpanTree()
|
||||
return t
|
||||
}
|
||||
|
||||
func NewFlamegraphTraceFromStorable(spans []StorableSpan) *FlamegraphTrace {
|
||||
t := &FlamegraphTrace{
|
||||
nodeByID: make(map[string]*FlamegraphSpan, len(spans)),
|
||||
}
|
||||
for i := range spans {
|
||||
node := NewFlamegraphSpanFromStorable(&spans[i], 0) // level is set later by BFS
|
||||
t.updateTimeRange(node.Timestamp, node.DurationNano)
|
||||
t.nodeByID[node.SpanID] = node
|
||||
}
|
||||
t.buildSpanTree()
|
||||
return t
|
||||
}
|
||||
|
||||
func (t *FlamegraphTrace) GetAllLevels() [][]*FlamegraphSpan {
|
||||
return nil
|
||||
}
|
||||
|
||||
// GetSelectedLevels returns the window of levels around selectedSpanID with sampling applied to dense levels.
|
||||
func (t *FlamegraphTrace) GetSelectedLevels(
|
||||
selectedSpanID string,
|
||||
levelLimit, spansPerLevel, topLatencyCount, bucketCount int,
|
||||
) []FlamegraphLevel {
|
||||
return nil
|
||||
}
|
||||
|
||||
func (t *FlamegraphTrace) EnrichSelectedSpans(selectedSpans []FlamegraphLevel, fullSpans []StorableSpan) [][]*FlamegraphSpan {
|
||||
fullByID := make(map[string]*StorableSpan, len(fullSpans))
|
||||
for i := range fullSpans {
|
||||
fullByID[fullSpans[i].SpanID] = &fullSpans[i]
|
||||
}
|
||||
|
||||
result := make([][]*FlamegraphSpan, len(selectedSpans))
|
||||
for i, lvl := range selectedSpans {
|
||||
result[i] = make([]*FlamegraphSpan, 0, len(lvl.SpanIDs))
|
||||
for _, spanID := range lvl.SpanIDs {
|
||||
if full, ok := fullByID[spanID]; ok {
|
||||
result[i] = append(result[i], NewFlamegraphSpanFromStorable(full, lvl.Level))
|
||||
} else if lean, ok := t.nodeByID[spanID]; ok {
|
||||
result[i] = append(result[i], lean)
|
||||
}
|
||||
}
|
||||
}
|
||||
return result
|
||||
}
|
||||
|
||||
func (t *FlamegraphTrace) updateTimeRange(timestamp, durationNano uint64) {
|
||||
if t.startTime == 0 || timestamp < t.startTime {
|
||||
t.startTime = timestamp
|
||||
}
|
||||
if end := timestamp + durationNano; end > t.endTime {
|
||||
t.endTime = end
|
||||
}
|
||||
}
|
||||
|
||||
func (t *FlamegraphTrace) buildSpanTree() {
|
||||
for _, node := range t.nodeByID {
|
||||
if node.ParentSpanID != "" {
|
||||
if parent, ok := t.nodeByID[node.ParentSpanID]; ok {
|
||||
parent.Children = append(parent.Children, node)
|
||||
} else {
|
||||
missing := &FlamegraphSpan{
|
||||
SpanID: node.ParentSpanID,
|
||||
Name: "Missing Span",
|
||||
Timestamp: node.Timestamp,
|
||||
DurationNano: node.DurationNano,
|
||||
Children: []*FlamegraphSpan{node},
|
||||
}
|
||||
t.nodeByID[missing.SpanID] = missing
|
||||
t.roots = append(t.roots, missing)
|
||||
}
|
||||
} else if flamegraphSpanIndex(t.roots, node.SpanID) == -1 {
|
||||
t.roots = append(t.roots, node)
|
||||
}
|
||||
}
|
||||
|
||||
sort.Slice(t.roots, func(i, j int) bool {
|
||||
if t.roots[i].Timestamp == t.roots[j].Timestamp {
|
||||
return t.roots[i].SpanID < t.roots[j].SpanID
|
||||
}
|
||||
return t.roots[i].Timestamp < t.roots[j].Timestamp
|
||||
})
|
||||
}
|
||||
|
||||
func flamegraphSpanIndex(spans []*FlamegraphSpan, spanID string) int {
|
||||
for i, s := range spans {
|
||||
if s != nil && s.SpanID == spanID {
|
||||
return i
|
||||
}
|
||||
}
|
||||
return -1
|
||||
}
|
||||
@@ -156,6 +156,18 @@ func (item *MinimalSpan) ToWaterfallSpan(traceID string) *WaterfallSpan {
|
||||
}
|
||||
}
|
||||
|
||||
func (item *MinimalSpan) ToFlamegraphSpan() *FlamegraphSpan {
|
||||
return &FlamegraphSpan{
|
||||
SpanID: item.SpanID,
|
||||
ParentSpanID: item.ParentSpanID,
|
||||
Timestamp: uint64(item.StartTime.UnixNano()),
|
||||
DurationNano: item.DurationNano,
|
||||
HasError: item.HasError,
|
||||
ServiceName: item.ServiceName,
|
||||
Children: make([]*FlamegraphSpan, 0),
|
||||
}
|
||||
}
|
||||
|
||||
// NewMissingWaterfallSpan creates a synthetic placeholder span for a parent that has no recorded data.
|
||||
func NewMissingWaterfallSpan(spanID, traceID string, timeUnixNano, durationNano uint64) *WaterfallSpan {
|
||||
return &WaterfallSpan{
|
||||
|
||||
Reference in New Issue
Block a user