mirror of
https://github.com/SigNoz/signoz.git
synced 2026-10-17 15:40:42 +01:00
Compare commits
3 Commits
ci/e2e-pul
...
nv/alerts-
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
d6744c1d8e | ||
|
|
6476be26bb | ||
|
|
06c54c6483 |
@@ -9510,35 +9510,34 @@ components:
|
||||
type: string
|
||||
disabled:
|
||||
type: boolean
|
||||
evalWindow:
|
||||
type: string
|
||||
evaluation:
|
||||
$ref: '#/components/schemas/RuletypesEvaluationEnvelope'
|
||||
frequency:
|
||||
type: string
|
||||
labels:
|
||||
additionalProperties:
|
||||
type: string
|
||||
type: object
|
||||
notificationSettings:
|
||||
$ref: '#/components/schemas/RuletypesNotificationSettings'
|
||||
preferredChannels:
|
||||
items:
|
||||
type: string
|
||||
type: array
|
||||
ruleType:
|
||||
$ref: '#/components/schemas/RuletypesRuleType'
|
||||
schemaVersion:
|
||||
enum:
|
||||
- v2alpha1
|
||||
type: string
|
||||
source:
|
||||
deprecated: true
|
||||
type: string
|
||||
version:
|
||||
deprecated: true
|
||||
type: string
|
||||
required:
|
||||
- alert
|
||||
- alertType
|
||||
- ruleType
|
||||
- condition
|
||||
- evaluation
|
||||
- schemaVersion
|
||||
- notificationSettings
|
||||
type: object
|
||||
RuletypesPostableRuleView:
|
||||
properties:
|
||||
@@ -9599,12 +9598,8 @@ components:
|
||||
type: string
|
||||
disabled:
|
||||
type: boolean
|
||||
evalWindow:
|
||||
type: string
|
||||
evaluation:
|
||||
$ref: '#/components/schemas/RuletypesEvaluationEnvelope'
|
||||
frequency:
|
||||
type: string
|
||||
id:
|
||||
type: string
|
||||
labels:
|
||||
@@ -9613,15 +9608,14 @@ components:
|
||||
type: object
|
||||
notificationSettings:
|
||||
$ref: '#/components/schemas/RuletypesNotificationSettings'
|
||||
preferredChannels:
|
||||
items:
|
||||
type: string
|
||||
type: array
|
||||
ruleType:
|
||||
$ref: '#/components/schemas/RuletypesRuleType'
|
||||
schemaVersion:
|
||||
enum:
|
||||
- v2alpha1
|
||||
type: string
|
||||
source:
|
||||
deprecated: true
|
||||
type: string
|
||||
state:
|
||||
$ref: '#/components/schemas/RuletypesAlertState'
|
||||
@@ -9631,6 +9625,7 @@ components:
|
||||
updatedBy:
|
||||
type: string
|
||||
version:
|
||||
deprecated: true
|
||||
type: string
|
||||
required:
|
||||
- id
|
||||
@@ -9639,6 +9634,9 @@ components:
|
||||
- alertType
|
||||
- ruleType
|
||||
- condition
|
||||
- evaluation
|
||||
- schemaVersion
|
||||
- notificationSettings
|
||||
type: object
|
||||
RuletypesRuleCondition:
|
||||
properties:
|
||||
@@ -9651,10 +9649,6 @@ components:
|
||||
type: string
|
||||
compositeQuery:
|
||||
$ref: '#/components/schemas/RuletypesAlertCompositeQuery'
|
||||
matchType:
|
||||
$ref: '#/components/schemas/RuletypesMatchType'
|
||||
op:
|
||||
$ref: '#/components/schemas/RuletypesCompareOperator'
|
||||
requireMinPoints:
|
||||
type: boolean
|
||||
requiredNumPoints:
|
||||
@@ -9663,16 +9657,11 @@ components:
|
||||
$ref: '#/components/schemas/RuletypesSeasonality'
|
||||
selectedQueryName:
|
||||
type: string
|
||||
target:
|
||||
format: double
|
||||
nullable: true
|
||||
type: number
|
||||
targetUnit:
|
||||
type: string
|
||||
thresholds:
|
||||
$ref: '#/components/schemas/RuletypesRuleThresholdData'
|
||||
required:
|
||||
- compositeQuery
|
||||
- thresholds
|
||||
type: object
|
||||
RuletypesRuleThresholdData:
|
||||
discriminator:
|
||||
@@ -22232,56 +22221,6 @@ paths:
|
||||
- rules
|
||||
x-signoz-stability: alpha
|
||||
/api/v2/rules:
|
||||
get:
|
||||
deprecated: true
|
||||
description: 'This endpoint lists all alert rules with their current evaluation
|
||||
state. Deprecated: use ListRulesV3, which supports filtering, sorting and
|
||||
pagination.'
|
||||
operationId: ListRules
|
||||
responses:
|
||||
"200":
|
||||
content:
|
||||
application/json:
|
||||
schema:
|
||||
properties:
|
||||
data:
|
||||
items:
|
||||
$ref: '#/components/schemas/RuletypesRule'
|
||||
type: array
|
||||
status:
|
||||
type: string
|
||||
required:
|
||||
- status
|
||||
- data
|
||||
type: object
|
||||
description: OK
|
||||
"401":
|
||||
content:
|
||||
application/json:
|
||||
schema:
|
||||
$ref: '#/components/schemas/RenderErrorResponse'
|
||||
description: Unauthorized
|
||||
"403":
|
||||
content:
|
||||
application/json:
|
||||
schema:
|
||||
$ref: '#/components/schemas/RenderErrorResponse'
|
||||
description: Forbidden
|
||||
"500":
|
||||
content:
|
||||
application/json:
|
||||
schema:
|
||||
$ref: '#/components/schemas/RenderErrorResponse'
|
||||
description: Internal Server Error
|
||||
security:
|
||||
- api_key:
|
||||
- VIEWER
|
||||
- tokenizer:
|
||||
- VIEWER
|
||||
summary: List alert rules
|
||||
tags:
|
||||
- rules
|
||||
x-signoz-stability: alpha
|
||||
post:
|
||||
deprecated: false
|
||||
description: This endpoint creates a new alert rule
|
||||
@@ -22440,13 +22379,13 @@ paths:
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_anomaly:
|
||||
description: Anomaly rules are not yet supported under schemaVersion
|
||||
v2alpha1, so this example uses the v1 shape. Wraps a builder query
|
||||
in the `anomaly` function with daily seasonality SigNoz compares
|
||||
each point against the forecast for that time of day. Fires when
|
||||
the anomaly score stays below the threshold for the entire window;
|
||||
`requireMinPoints` guards against noisy intervals.
|
||||
summary: Metric anomaly rule (v1 only)
|
||||
description: Wraps a builder query in the `anomaly` function with
|
||||
daily seasonality, so SigNoz compares each point against the forecast
|
||||
for that time of day. The threshold is compared against the signed
|
||||
anomaly score, so `below` with a target of -2 fires when an abrupt
|
||||
drop holds for the entire window; `requireMinPoints` guards against
|
||||
noisy intervals.
|
||||
summary: Metric anomaly rule
|
||||
value:
|
||||
alert: Anomalous drop in ingested spans
|
||||
alertType: METRIC_BASED_ALERT
|
||||
@@ -22481,22 +22420,36 @@ paths:
|
||||
stepInterval: 21600
|
||||
type: builder_query
|
||||
queryType: builder
|
||||
matchType: all_the_times
|
||||
op: below
|
||||
requireMinPoints: true
|
||||
requiredNumPoints: 3
|
||||
seasonality: daily
|
||||
selectedQueryName: A
|
||||
target: 2
|
||||
thresholds:
|
||||
kind: basic
|
||||
spec:
|
||||
- channels:
|
||||
- slack-ingestion
|
||||
matchType: all_the_times
|
||||
name: warning
|
||||
op: below
|
||||
target: -2
|
||||
description: Detect an abrupt drop in span ingestion using a z-score
|
||||
anomaly function
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
evaluation:
|
||||
kind: rolling
|
||||
spec:
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
labels:
|
||||
severity: warning
|
||||
preferredChannels:
|
||||
- slack-ingestion
|
||||
notificationSettings:
|
||||
renotify:
|
||||
alertStates:
|
||||
- firing
|
||||
enabled: true
|
||||
interval: 4h
|
||||
ruleType: anomaly_rule
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_promql:
|
||||
description: PromQL expression instead of the builder. Dotted OTEL
|
||||
@@ -23392,13 +23345,13 @@ paths:
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_anomaly:
|
||||
description: Anomaly rules are not yet supported under schemaVersion
|
||||
v2alpha1, so this example uses the v1 shape. Wraps a builder query
|
||||
in the `anomaly` function with daily seasonality SigNoz compares
|
||||
each point against the forecast for that time of day. Fires when
|
||||
the anomaly score stays below the threshold for the entire window;
|
||||
`requireMinPoints` guards against noisy intervals.
|
||||
summary: Metric anomaly rule (v1 only)
|
||||
description: Wraps a builder query in the `anomaly` function with
|
||||
daily seasonality, so SigNoz compares each point against the forecast
|
||||
for that time of day. The threshold is compared against the signed
|
||||
anomaly score, so `below` with a target of -2 fires when an abrupt
|
||||
drop holds for the entire window; `requireMinPoints` guards against
|
||||
noisy intervals.
|
||||
summary: Metric anomaly rule
|
||||
value:
|
||||
alert: Anomalous drop in ingested spans
|
||||
alertType: METRIC_BASED_ALERT
|
||||
@@ -23433,22 +23386,36 @@ paths:
|
||||
stepInterval: 21600
|
||||
type: builder_query
|
||||
queryType: builder
|
||||
matchType: all_the_times
|
||||
op: below
|
||||
requireMinPoints: true
|
||||
requiredNumPoints: 3
|
||||
seasonality: daily
|
||||
selectedQueryName: A
|
||||
target: 2
|
||||
thresholds:
|
||||
kind: basic
|
||||
spec:
|
||||
- channels:
|
||||
- slack-ingestion
|
||||
matchType: all_the_times
|
||||
name: warning
|
||||
op: below
|
||||
target: -2
|
||||
description: Detect an abrupt drop in span ingestion using a z-score
|
||||
anomaly function
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
evaluation:
|
||||
kind: rolling
|
||||
spec:
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
labels:
|
||||
severity: warning
|
||||
preferredChannels:
|
||||
- slack-ingestion
|
||||
notificationSettings:
|
||||
renotify:
|
||||
alertStates:
|
||||
- firing
|
||||
enabled: true
|
||||
interval: 4h
|
||||
ruleType: anomaly_rule
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_promql:
|
||||
description: PromQL expression instead of the builder. Dotted OTEL
|
||||
@@ -24245,13 +24212,13 @@ paths:
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_anomaly:
|
||||
description: Anomaly rules are not yet supported under schemaVersion
|
||||
v2alpha1, so this example uses the v1 shape. Wraps a builder query
|
||||
in the `anomaly` function with daily seasonality SigNoz compares
|
||||
each point against the forecast for that time of day. Fires when
|
||||
the anomaly score stays below the threshold for the entire window;
|
||||
`requireMinPoints` guards against noisy intervals.
|
||||
summary: Metric anomaly rule (v1 only)
|
||||
description: Wraps a builder query in the `anomaly` function with
|
||||
daily seasonality, so SigNoz compares each point against the forecast
|
||||
for that time of day. The threshold is compared against the signed
|
||||
anomaly score, so `below` with a target of -2 fires when an abrupt
|
||||
drop holds for the entire window; `requireMinPoints` guards against
|
||||
noisy intervals.
|
||||
summary: Metric anomaly rule
|
||||
value:
|
||||
alert: Anomalous drop in ingested spans
|
||||
alertType: METRIC_BASED_ALERT
|
||||
@@ -24286,22 +24253,36 @@ paths:
|
||||
stepInterval: 21600
|
||||
type: builder_query
|
||||
queryType: builder
|
||||
matchType: all_the_times
|
||||
op: below
|
||||
requireMinPoints: true
|
||||
requiredNumPoints: 3
|
||||
seasonality: daily
|
||||
selectedQueryName: A
|
||||
target: 2
|
||||
thresholds:
|
||||
kind: basic
|
||||
spec:
|
||||
- channels:
|
||||
- slack-ingestion
|
||||
matchType: all_the_times
|
||||
name: warning
|
||||
op: below
|
||||
target: -2
|
||||
description: Detect an abrupt drop in span ingestion using a z-score
|
||||
anomaly function
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
evaluation:
|
||||
kind: rolling
|
||||
spec:
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
labels:
|
||||
severity: warning
|
||||
preferredChannels:
|
||||
- slack-ingestion
|
||||
notificationSettings:
|
||||
renotify:
|
||||
alertStates:
|
||||
- firing
|
||||
enabled: true
|
||||
interval: 4h
|
||||
ruleType: anomaly_rule
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_promql:
|
||||
description: PromQL expression instead of the builder. Dotted OTEL
|
||||
@@ -25607,13 +25588,13 @@ paths:
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_anomaly:
|
||||
description: Anomaly rules are not yet supported under schemaVersion
|
||||
v2alpha1, so this example uses the v1 shape. Wraps a builder query
|
||||
in the `anomaly` function with daily seasonality SigNoz compares
|
||||
each point against the forecast for that time of day. Fires when
|
||||
the anomaly score stays below the threshold for the entire window;
|
||||
`requireMinPoints` guards against noisy intervals.
|
||||
summary: Metric anomaly rule (v1 only)
|
||||
description: Wraps a builder query in the `anomaly` function with
|
||||
daily seasonality, so SigNoz compares each point against the forecast
|
||||
for that time of day. The threshold is compared against the signed
|
||||
anomaly score, so `below` with a target of -2 fires when an abrupt
|
||||
drop holds for the entire window; `requireMinPoints` guards against
|
||||
noisy intervals.
|
||||
summary: Metric anomaly rule
|
||||
value:
|
||||
alert: Anomalous drop in ingested spans
|
||||
alertType: METRIC_BASED_ALERT
|
||||
@@ -25648,22 +25629,36 @@ paths:
|
||||
stepInterval: 21600
|
||||
type: builder_query
|
||||
queryType: builder
|
||||
matchType: all_the_times
|
||||
op: below
|
||||
requireMinPoints: true
|
||||
requiredNumPoints: 3
|
||||
seasonality: daily
|
||||
selectedQueryName: A
|
||||
target: 2
|
||||
thresholds:
|
||||
kind: basic
|
||||
spec:
|
||||
- channels:
|
||||
- slack-ingestion
|
||||
matchType: all_the_times
|
||||
name: warning
|
||||
op: below
|
||||
target: -2
|
||||
description: Detect an abrupt drop in span ingestion using a z-score
|
||||
anomaly function
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
evaluation:
|
||||
kind: rolling
|
||||
spec:
|
||||
evalWindow: 24h
|
||||
frequency: 3h
|
||||
labels:
|
||||
severity: warning
|
||||
preferredChannels:
|
||||
- slack-ingestion
|
||||
notificationSettings:
|
||||
renotify:
|
||||
alertStates:
|
||||
- firing
|
||||
enabled: true
|
||||
interval: 4h
|
||||
ruleType: anomaly_rule
|
||||
schemaVersion: v2alpha1
|
||||
version: v5
|
||||
metric_promql:
|
||||
description: PromQL expression instead of the builder. Dotted OTEL
|
||||
|
||||
@@ -105,15 +105,15 @@ func (r *AnomalyRule) Type() ruletypes.RuleType {
|
||||
}
|
||||
|
||||
func (r *AnomalyRule) prepareQueryRange(ctx context.Context, ts time.Time) *qbtypes.QueryRangeRequest {
|
||||
r.logger.InfoContext(
|
||||
ctx, "prepare query range request", slog.Int64("ts", ts.UnixMilli()),
|
||||
slog.Int64("eval.window_ms", r.EvalWindow().Milliseconds()),
|
||||
slog.Int64("eval.delay_ms", r.EvalDelay().Milliseconds()),
|
||||
)
|
||||
|
||||
startTs, endTs := r.Timestamps(ts)
|
||||
start, end := startTs.UnixMilli(), endTs.UnixMilli()
|
||||
|
||||
r.logger.InfoContext(
|
||||
ctx, "prepare query range request", slog.Int64("ts", ts.UnixMilli()),
|
||||
slog.Int64("eval.window_ms", end-start),
|
||||
slog.Int64("eval.delay_ms", r.EvalDelay().Milliseconds()),
|
||||
)
|
||||
|
||||
req := &qbtypes.QueryRangeRequest{
|
||||
Start: uint64(start),
|
||||
End: uint64(end),
|
||||
@@ -423,12 +423,10 @@ func (r *AnomalyRule) Eval(ctx context.Context, ts time.Time) (int, error) {
|
||||
func (r *AnomalyRule) String() string {
|
||||
|
||||
ar := ruletypes.PostableRule{
|
||||
AlertName: r.Name(),
|
||||
RuleCondition: r.Condition(),
|
||||
EvalWindow: r.EvalWindow(),
|
||||
Labels: r.Labels().Map(),
|
||||
Annotations: r.Annotations().Map(),
|
||||
PreferredChannels: r.PreferredChannels(),
|
||||
AlertName: r.Name(),
|
||||
RuleCondition: r.Condition(),
|
||||
Labels: r.Labels().Map(),
|
||||
Annotations: r.Annotations().Map(),
|
||||
}
|
||||
|
||||
byt, err := json.Marshal(ar)
|
||||
|
||||
@@ -54,9 +54,6 @@ func TestAnomalyRule_NoData_AlertOnAbsent(t *testing.T) {
|
||||
Frequency: valuer.MustParseTextDuration("1m"),
|
||||
}},
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
Target: &target,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypeBuilder,
|
||||
Queries: []qbtypes.QueryEnvelope{{
|
||||
@@ -157,10 +154,7 @@ func TestAnomalyRule_NoData_AbsentFor(t *testing.T) {
|
||||
Frequency: valuer.MustParseTextDuration("1m"),
|
||||
}},
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
AlertOnAbsent: true,
|
||||
Target: &target,
|
||||
AlertOnAbsent: true,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypeBuilder,
|
||||
Queries: []qbtypes.QueryEnvelope{{
|
||||
|
||||
@@ -9,8 +9,6 @@ import (
|
||||
"github.com/SigNoz/signoz/pkg/sqlstore"
|
||||
"github.com/SigNoz/signoz/pkg/types/alertmanagertypes"
|
||||
"github.com/SigNoz/signoz/pkg/valuer"
|
||||
"github.com/tidwall/gjson"
|
||||
"github.com/uptrace/bun"
|
||||
)
|
||||
|
||||
type config struct {
|
||||
@@ -239,37 +237,6 @@ func (store *config) ListAllChannels(ctx context.Context) ([]*alertmanagertypes.
|
||||
return channels, nil
|
||||
}
|
||||
|
||||
func (store *config) GetMatchers(ctx context.Context, orgID string) (map[string][]string, error) {
|
||||
type matcher struct {
|
||||
bun.BaseModel `bun:"table:rule"`
|
||||
ID valuer.UUID `bun:"id,pk"`
|
||||
Data string `bun:"data"`
|
||||
}
|
||||
|
||||
matchers := []matcher{}
|
||||
|
||||
err := store.
|
||||
sqlstore.
|
||||
BunDB().
|
||||
NewSelect().
|
||||
Column("id", "data").
|
||||
Model(&matchers).
|
||||
Scan(ctx)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
matchersMap := make(map[string][]string)
|
||||
for _, matcher := range matchers {
|
||||
receivers := gjson.Get(matcher.Data, "preferredChannels").Array()
|
||||
for _, receiver := range receivers {
|
||||
matchersMap[matcher.ID.StringValue()] = append(matchersMap[matcher.ID.StringValue()], receiver.String())
|
||||
}
|
||||
}
|
||||
|
||||
return matchersMap, nil
|
||||
}
|
||||
|
||||
func (store *config) wrap(ctx context.Context, fn func(ctx context.Context) error, opts ...alertmanagertypes.StoreOption) error {
|
||||
storeOpts := alertmanagertypes.NewStoreOptions(opts...)
|
||||
|
||||
|
||||
@@ -245,25 +245,13 @@ func (service *Service) compareAndSelectConfig(ctx context.Context, incomingConf
|
||||
return nil, err
|
||||
}
|
||||
|
||||
matchers, err := service.configStore.GetMatchers(ctx, incomingConfig.StoreableConfig().OrgID)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
config, err := alertmanagertypes.NewConfigFromChannels(service.config.Global, service.config.Route, channels, incomingConfig.StoreableConfig().OrgID)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
for ruleID, receivers := range matchers {
|
||||
err = config.CreateRuleIDMatcher(ruleID, receivers)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
}
|
||||
|
||||
if incomingConfig.StoreableConfig().Hash != config.StoreableConfig().Hash {
|
||||
service.settings.Logger().InfoContext(ctx, "mismatch found, updating config to match channels and matchers")
|
||||
service.settings.Logger().InfoContext(ctx, "mismatch found, updating config to match channels")
|
||||
return config, nil
|
||||
}
|
||||
|
||||
|
||||
@@ -11,20 +11,6 @@ import (
|
||||
)
|
||||
|
||||
func (provider *provider) addRulerRoutes(router *mux.Router) error {
|
||||
if err := router.Handle("/api/v2/rules", handler.New(provider.authzMiddleware.ViewAccess(provider.rulerHandler.ListRules), handler.OpenAPIDef{
|
||||
ID: "ListRules",
|
||||
Tags: []string{"rules"},
|
||||
Summary: "List alert rules",
|
||||
Description: "This endpoint lists all alert rules with their current evaluation state. Deprecated: use ListRulesV3, which supports filtering, sorting and pagination.",
|
||||
Response: make([]*ruletypes.Rule, 0),
|
||||
ResponseContentType: "application/json",
|
||||
SuccessStatusCode: http.StatusOK,
|
||||
Deprecated: true,
|
||||
SecuritySchemes: newSecuritySchemes(types.RoleViewer),
|
||||
})).Methods(http.MethodGet).GetError(); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
if err := router.Handle("/api/v3/rules", handler.New(provider.authzMiddleware.ViewAccess(provider.rulerHandler.ListRulesV3), handler.OpenAPIDef{
|
||||
ID: "ListRulesV3",
|
||||
Tags: []string{"rules"},
|
||||
|
||||
@@ -221,16 +221,15 @@ func postableRuleExamples() []handler.OpenAPIExample {
|
||||
},
|
||||
{
|
||||
Name: "metric_anomaly",
|
||||
Summary: "Metric anomaly rule (v1 only)",
|
||||
Description: "Anomaly rules are not yet supported under schemaVersion v2alpha1, so this example uses the v1 shape. Wraps a builder query in the `anomaly` function with daily seasonality SigNoz compares each point against the forecast for that time of day. Fires when the anomaly score stays below the threshold for the entire window; `requireMinPoints` guards against noisy intervals.",
|
||||
Summary: "Metric anomaly rule",
|
||||
Description: "Wraps a builder query in the `anomaly` function with daily seasonality, so SigNoz compares each point against the forecast for that time of day. The threshold is compared against the signed anomaly score, so `below` with a target of -2 fires when an abrupt drop holds for the entire window; `requireMinPoints` guards against noisy intervals.",
|
||||
Value: map[string]any{
|
||||
"alert": "Anomalous drop in ingested spans",
|
||||
"alertType": "METRIC_BASED_ALERT",
|
||||
"description": "Detect an abrupt drop in span ingestion using a z-score anomaly function",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"evalWindow": "24h",
|
||||
"frequency": "3h",
|
||||
"alert": "Anomalous drop in ingested spans",
|
||||
"alertType": "METRIC_BASED_ALERT",
|
||||
"description": "Detect an abrupt drop in span ingestion using a z-score anomaly function",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"schemaVersion": "v2alpha1",
|
||||
"condition": map[string]any{
|
||||
"compositeQuery": map[string]any{
|
||||
"queryType": "builder",
|
||||
@@ -256,17 +255,29 @@ func postableRuleExamples() []handler.OpenAPIExample {
|
||||
},
|
||||
},
|
||||
},
|
||||
"op": "below",
|
||||
"matchType": "all_the_times",
|
||||
"target": 2,
|
||||
"algorithm": "standard",
|
||||
"seasonality": "daily",
|
||||
"selectedQueryName": "A",
|
||||
"requireMinPoints": true,
|
||||
"requiredNumPoints": 3,
|
||||
"thresholds": map[string]any{
|
||||
"kind": "basic",
|
||||
"spec": []any{
|
||||
map[string]any{
|
||||
"name": "warning",
|
||||
"op": "below",
|
||||
"matchType": "all_the_times",
|
||||
"target": -2,
|
||||
"channels": []any{"slack-ingestion"},
|
||||
},
|
||||
},
|
||||
},
|
||||
},
|
||||
"labels": map[string]any{"severity": "warning"},
|
||||
"preferredChannels": []any{"slack-ingestion"},
|
||||
"evaluation": rolling("24h", "3h"),
|
||||
"notificationSettings": map[string]any{
|
||||
"renotify": renotify("4h", "firing"),
|
||||
},
|
||||
"labels": map[string]any{"severity": "warning"},
|
||||
"annotations": map[string]any{
|
||||
"description": "Ingestion rate for tenant {{$tenant_id}} is anomalously low (z-score {{$value}}).",
|
||||
"summary": "Span ingestion anomaly",
|
||||
|
||||
@@ -49,19 +49,14 @@ func (m *module) relatedLinkBuilderForRule(ctx context.Context, orgID valuer.UUI
|
||||
return nil
|
||||
}
|
||||
|
||||
builder := &relatedLinkBuilder{alertType: rule.AlertType}
|
||||
if rule.Evaluation != nil {
|
||||
if evaluation, err := rule.Evaluation.GetEvaluation(); err == nil {
|
||||
builder.evaluation = evaluation
|
||||
}
|
||||
if rule.Evaluation == nil {
|
||||
return nil
|
||||
}
|
||||
if builder.evaluation == nil {
|
||||
evalWindow := rule.EvalWindow
|
||||
if evalWindow.IsZero() {
|
||||
evalWindow = valuer.MustParseTextDuration("5m")
|
||||
}
|
||||
builder.evaluation = ruletypes.RollingWindow{EvalWindow: evalWindow}
|
||||
evaluation, err := rule.Evaluation.GetEvaluation()
|
||||
if err != nil {
|
||||
return nil
|
||||
}
|
||||
builder := &relatedLinkBuilder{alertType: rule.AlertType, evaluation: evaluation}
|
||||
|
||||
// links are still built from the labels alone when the rule has no builder
|
||||
// query for the signal (e.g. ClickHouse SQL alerts)
|
||||
|
||||
@@ -3,10 +3,8 @@ package app
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"database/sql"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"io"
|
||||
|
||||
"github.com/SigNoz/signoz/pkg/clickhousesql"
|
||||
"github.com/SigNoz/signoz/pkg/errors"
|
||||
@@ -360,14 +358,6 @@ func (aH *APIHandler) RegisterRoutes(router *mux.Router, am *middleware.AuthZ) {
|
||||
router.HandleFunc("/api/v1/query_range", am.ViewAccess(aH.queryRangeMetrics)).Methods(http.MethodGet)
|
||||
router.HandleFunc("/api/v1/query", am.ViewAccess(aH.queryMetrics)).Methods(http.MethodGet)
|
||||
|
||||
router.HandleFunc("/api/v1/rules", am.ViewAccess(aH.listRules)).Methods(http.MethodGet)
|
||||
router.HandleFunc("/api/v1/rules/{id}", am.ViewAccess(aH.getRule)).Methods(http.MethodGet)
|
||||
router.HandleFunc("/api/v1/rules", am.EditAccess(aH.createRule)).Methods(http.MethodPost)
|
||||
router.HandleFunc("/api/v1/rules/{id}", am.EditAccess(aH.editRule)).Methods(http.MethodPut)
|
||||
router.HandleFunc("/api/v1/rules/{id}", am.EditAccess(aH.deleteRule)).Methods(http.MethodDelete)
|
||||
router.HandleFunc("/api/v1/rules/{id}", am.EditAccess(aH.patchRule)).Methods(http.MethodPatch)
|
||||
router.HandleFunc("/api/v1/testRule", am.EditAccess(aH.testRule)).Methods(http.MethodPost)
|
||||
|
||||
router.HandleFunc("/api/v2/variables/query", am.ViewAccess(aH.queryDashboardVarsV2)).Methods(http.MethodPost)
|
||||
|
||||
router.HandleFunc("/api/v1/explorer/views", am.ViewAccess(aH.Signoz.Handlers.SavedView.List)).Methods(http.MethodGet)
|
||||
@@ -496,166 +486,6 @@ func (aH *APIHandler) PopulateTemporality(ctx context.Context, orgID valuer.UUID
|
||||
return nil
|
||||
}
|
||||
|
||||
func (aH *APIHandler) listRules(w http.ResponseWriter, r *http.Request) {
|
||||
|
||||
rules, err := aH.ruleManager.ListRuleStates(r.Context())
|
||||
if err != nil {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorInternal, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
// todo(amol): need to add sorter
|
||||
|
||||
aH.Respond(w, rules)
|
||||
}
|
||||
|
||||
func (aH *APIHandler) getRule(w http.ResponseWriter, r *http.Request) {
|
||||
idStr := mux.Vars(r)["id"]
|
||||
id, err := valuer.NewUUID(idStr)
|
||||
if err != nil {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
ruleResponse, err := aH.ruleManager.GetRule(r.Context(), id)
|
||||
if err != nil {
|
||||
if errors.Is(err, sql.ErrNoRows) {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorNotFound, Err: fmt.Errorf("rule not found")}, nil)
|
||||
return
|
||||
}
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorInternal, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
aH.Respond(w, ruleResponse)
|
||||
}
|
||||
|
||||
func (aH *APIHandler) createRule(w http.ResponseWriter, r *http.Request) {
|
||||
defer r.Body.Close()
|
||||
body, err := io.ReadAll(r.Body)
|
||||
if err != nil {
|
||||
aH.logger.ErrorContext(r.Context(), "error reading request body for create rule", errors.Attr(err))
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
rule, err := aH.ruleManager.CreateRule(r.Context(), string(body))
|
||||
if err != nil {
|
||||
RespondError(w, toApiError(err), nil)
|
||||
return
|
||||
}
|
||||
|
||||
aH.Respond(w, rule)
|
||||
}
|
||||
|
||||
func (aH *APIHandler) editRule(w http.ResponseWriter, r *http.Request) {
|
||||
idStr := mux.Vars(r)["id"]
|
||||
id, err := valuer.NewUUID(idStr)
|
||||
if err != nil {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
defer r.Body.Close()
|
||||
body, err := io.ReadAll(r.Body)
|
||||
if err != nil {
|
||||
aH.logger.ErrorContext(r.Context(), "error reading request body for edit rule", errors.Attr(err))
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
err = aH.ruleManager.EditRule(r.Context(), string(body), id)
|
||||
if err != nil {
|
||||
if errors.Is(err, sql.ErrNoRows) {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorNotFound, Err: fmt.Errorf("rule not found")}, nil)
|
||||
return
|
||||
}
|
||||
RespondError(w, toApiError(err), nil)
|
||||
return
|
||||
}
|
||||
|
||||
aH.Respond(w, "rule successfully edited")
|
||||
}
|
||||
|
||||
func (aH *APIHandler) deleteRule(w http.ResponseWriter, r *http.Request) {
|
||||
id := mux.Vars(r)["id"]
|
||||
|
||||
err := aH.ruleManager.DeleteRule(r.Context(), id)
|
||||
if err != nil {
|
||||
if errors.Is(err, sql.ErrNoRows) {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorNotFound, Err: fmt.Errorf("rule not found")}, nil)
|
||||
return
|
||||
}
|
||||
RespondError(w, toApiError(err), nil)
|
||||
return
|
||||
}
|
||||
|
||||
aH.Respond(w, "rule successfully deleted")
|
||||
}
|
||||
|
||||
func (aH *APIHandler) patchRule(w http.ResponseWriter, r *http.Request) {
|
||||
idStr := mux.Vars(r)["id"]
|
||||
id, err := valuer.NewUUID(idStr)
|
||||
if err != nil {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
defer r.Body.Close()
|
||||
body, err := io.ReadAll(r.Body)
|
||||
if err != nil {
|
||||
aH.logger.ErrorContext(r.Context(), "error reading request body for patch rule", errors.Attr(err))
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
gettableRule, err := aH.ruleManager.PatchRule(r.Context(), string(body), id)
|
||||
if err != nil {
|
||||
if errors.Is(err, sql.ErrNoRows) {
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorNotFound, Err: fmt.Errorf("rule not found")}, nil)
|
||||
return
|
||||
}
|
||||
RespondError(w, toApiError(err), nil)
|
||||
return
|
||||
}
|
||||
|
||||
aH.Respond(w, gettableRule)
|
||||
}
|
||||
|
||||
func (aH *APIHandler) testRule(w http.ResponseWriter, r *http.Request) {
|
||||
claims, err := authtypes.ClaimsFromContext(r.Context())
|
||||
if err != nil {
|
||||
render.Error(w, err)
|
||||
return
|
||||
}
|
||||
orgID, err := valuer.NewUUID(claims.OrgID)
|
||||
if err != nil {
|
||||
render.Error(w, err)
|
||||
return
|
||||
}
|
||||
defer r.Body.Close()
|
||||
body, err := io.ReadAll(r.Body)
|
||||
if err != nil {
|
||||
aH.logger.ErrorContext(r.Context(), "error reading request body for test rule", errors.Attr(err))
|
||||
RespondError(w, &model.ApiError{Typ: model.ErrorBadData, Err: err}, nil)
|
||||
return
|
||||
}
|
||||
|
||||
ctx, cancel := context.WithTimeout(context.Background(), 1*time.Minute)
|
||||
defer cancel()
|
||||
|
||||
alertCount, err := aH.ruleManager.TestNotification(ctx, orgID, string(body))
|
||||
if err != nil {
|
||||
RespondError(w, toApiError(err), nil)
|
||||
return
|
||||
}
|
||||
|
||||
response := map[string]interface{}{
|
||||
"alertCount": alertCount,
|
||||
"message": "notification sent",
|
||||
}
|
||||
aH.Respond(w, response)
|
||||
}
|
||||
|
||||
func prepareQuery(r *http.Request) (string, error) {
|
||||
var postData *model.DashboardVars
|
||||
|
||||
|
||||
@@ -33,10 +33,6 @@ type BaseRule struct {
|
||||
ruleCondition *ruletypes.RuleCondition
|
||||
|
||||
Threshold ruletypes.RuleThreshold
|
||||
// evalWindow is the time window used for evaluating the rule
|
||||
// i.e. each time we lookback from the current time, we look at data for the last
|
||||
// evalWindow duration
|
||||
evalWindow valuer.TextDuration
|
||||
// holdDuration is the duration for which the alert waits before firing
|
||||
holdDuration valuer.TextDuration
|
||||
|
||||
@@ -48,10 +44,7 @@ type BaseRule struct {
|
||||
// these are the same for all alerts created for this rule
|
||||
labels ruletypes.Labels
|
||||
annotations ruletypes.Labels
|
||||
// preferredChannels is the list of channels to send the alert to
|
||||
// if the rule is triggered
|
||||
preferredChannels []string
|
||||
mtx sync.Mutex
|
||||
mtx sync.Mutex
|
||||
// the time it took to evaluate the rule (most recent evaluation)
|
||||
evaluationDuration time.Duration
|
||||
// the timestamp of the last evaluation
|
||||
@@ -156,20 +149,18 @@ func NewBaseRule(
|
||||
}
|
||||
|
||||
baseRule := &BaseRule{
|
||||
id: id,
|
||||
orgID: orgID,
|
||||
externalURL: externalURL,
|
||||
name: p.AlertName,
|
||||
typ: p.AlertType,
|
||||
ruleCondition: p.RuleCondition,
|
||||
evalWindow: p.EvalWindow,
|
||||
labels: ruletypes.FromMap(p.Labels),
|
||||
annotations: ruletypes.FromMap(p.Annotations),
|
||||
preferredChannels: p.PreferredChannels,
|
||||
health: ruletypes.HealthUnknown,
|
||||
Active: map[uint64]*ruletypes.Alert{},
|
||||
Threshold: threshold,
|
||||
evaluation: evaluation,
|
||||
id: id,
|
||||
orgID: orgID,
|
||||
externalURL: externalURL,
|
||||
name: p.AlertName,
|
||||
typ: p.AlertType,
|
||||
ruleCondition: p.RuleCondition,
|
||||
labels: ruletypes.FromMap(p.Labels),
|
||||
annotations: ruletypes.FromMap(p.Annotations),
|
||||
health: ruletypes.HealthUnknown,
|
||||
Active: map[uint64]*ruletypes.Alert{},
|
||||
Threshold: threshold,
|
||||
evaluation: evaluation,
|
||||
}
|
||||
|
||||
// Store newGroupEvalDelay and groupBy keys from NotificationSettings
|
||||
@@ -177,10 +168,6 @@ func NewBaseRule(
|
||||
baseRule.newGroupEvalDelay = p.NotificationSettings.NewGroupEvalDelay
|
||||
}
|
||||
|
||||
if baseRule.evalWindow.IsZero() {
|
||||
baseRule.evalWindow = valuer.MustParseTextDuration("5m")
|
||||
}
|
||||
|
||||
for _, opt := range opts {
|
||||
opt(baseRule)
|
||||
}
|
||||
@@ -231,10 +218,6 @@ func (r *BaseRule) EvalDelay() valuer.TextDuration {
|
||||
return r.evalDelay
|
||||
}
|
||||
|
||||
func (r *BaseRule) EvalWindow() valuer.TextDuration {
|
||||
return r.evalWindow
|
||||
}
|
||||
|
||||
func (r *BaseRule) HoldDuration() valuer.TextDuration {
|
||||
return r.holdDuration
|
||||
}
|
||||
@@ -245,7 +228,6 @@ func (r *BaseRule) Name() string { return r.name }
|
||||
func (r *BaseRule) Condition() *ruletypes.RuleCondition { return r.ruleCondition }
|
||||
func (r *BaseRule) Labels() ruletypes.Labels { return r.labels }
|
||||
func (r *BaseRule) Annotations() ruletypes.Labels { return r.annotations }
|
||||
func (r *BaseRule) PreferredChannels() []string { return r.preferredChannels }
|
||||
|
||||
func (r *BaseRule) GeneratorURL() string {
|
||||
params := url.Values{}
|
||||
|
||||
@@ -49,11 +49,9 @@ func ThresholdRuleAtLeastOnceValueAbove(target float64, recovery *float64) rulet
|
||||
Annotations: map[string]string{
|
||||
"value": "{{$value}}",
|
||||
},
|
||||
Version: "v5",
|
||||
Version: "v5",
|
||||
SchemaVersion: ruletypes.SchemaVersionV2Alpha1,
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
Target: &target,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypeBuilder,
|
||||
Queries: []qbtypes.QueryEnvelope{
|
||||
@@ -108,12 +106,10 @@ func BuildPromAtLeastOnceValueAbove(target float64, recovery *float64) ruletypes
|
||||
Annotations: map[string]string{
|
||||
"value": "{{$value}}",
|
||||
},
|
||||
Version: "v5",
|
||||
Version: "v5",
|
||||
SchemaVersion: ruletypes.SchemaVersionV2Alpha1,
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
SelectedQuery: "A",
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
Target: &target,
|
||||
SelectedQuery: "A",
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypePromQL,
|
||||
PanelType: ruletypes.PanelTypeGraph,
|
||||
@@ -138,7 +134,6 @@ func BuildPromAtLeastOnceValueAbove(target float64, recovery *float64) ruletypes
|
||||
RecoveryTarget: recovery,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
Channels: []string{"slack"},
|
||||
},
|
||||
},
|
||||
},
|
||||
|
||||
@@ -359,12 +359,10 @@ func (r *PromRule) Eval(ctx context.Context, ts time.Time) (int, error) {
|
||||
|
||||
func (r *PromRule) String() string {
|
||||
ar := ruletypes.PostableRule{
|
||||
AlertName: r.name,
|
||||
RuleCondition: r.ruleCondition,
|
||||
EvalWindow: r.evalWindow,
|
||||
Labels: r.labels.Map(),
|
||||
Annotations: r.annotations.Map(),
|
||||
PreferredChannels: r.preferredChannels,
|
||||
AlertName: r.name,
|
||||
RuleCondition: r.ruleCondition,
|
||||
Labels: r.labels.Map(),
|
||||
Annotations: r.annotations.Map(),
|
||||
}
|
||||
|
||||
byt, err := json.Marshal(ar)
|
||||
|
||||
@@ -690,9 +690,6 @@ func TestPromRuleEval(t *testing.T) {
|
||||
logger := instrumentationtest.New().Logger()
|
||||
|
||||
for idx, c := range cases {
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -905,11 +902,7 @@ func TestPromRuleUnitCombinations(t *testing.T) {
|
||||
|
||||
promProvider := prometheustest.New(context.Background(), instrumentationtest.New().ToProviderSettings(), prometheus.Config{Timeout: 2 * time.Minute}, telemetryStore)
|
||||
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = ruletypes.MatchType(c.matchType)
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -1198,11 +1191,7 @@ func TestMultipleThresholdPromRule(t *testing.T) {
|
||||
|
||||
promProvider := prometheustest.New(context.Background(), instrumentationtest.New().ToProviderSettings(), prometheus.Config{Timeout: 2 * time.Minute}, telemetryStore)
|
||||
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -1269,8 +1258,6 @@ func TestPromRule_NoData(t *testing.T) {
|
||||
Frequency: valuer.MustParseTextDuration("1m"),
|
||||
}},
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypePromQL,
|
||||
Queries: []qbtypes.QueryEnvelope{
|
||||
@@ -1365,10 +1352,7 @@ func TestPromRule_NoData_AbsentFor(t *testing.T) {
|
||||
Frequency: valuer.MustParseTextDuration("1m"),
|
||||
}},
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
AlertOnAbsent: true,
|
||||
Target: &target,
|
||||
AlertOnAbsent: true,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypePromQL,
|
||||
Queries: []qbtypes.QueryEnvelope{
|
||||
@@ -1491,8 +1475,6 @@ func TestPromRuleEval_RequireMinPoints(t *testing.T) {
|
||||
Frequency: valuer.MustParseTextDuration("1m"),
|
||||
}},
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypePromQL,
|
||||
Queries: []qbtypes.QueryEnvelope{
|
||||
@@ -1562,9 +1544,9 @@ func TestPromRuleEval_RequireMinPoints(t *testing.T) {
|
||||
for _, c := range cases {
|
||||
rc := postableRule.RuleCondition
|
||||
|
||||
rc.Target = &targetForNoAlert
|
||||
target := &targetForNoAlert
|
||||
if c.alertCondition {
|
||||
rc.Target = &targetForAlert
|
||||
target = &targetForAlert
|
||||
}
|
||||
rc.RequireMinPoints = c.requireMinPoints
|
||||
rc.RequiredNumPoints = c.requiredNumPoints
|
||||
@@ -1573,9 +1555,9 @@ func TestPromRuleEval_RequireMinPoints(t *testing.T) {
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
{
|
||||
Name: postableRule.AlertName,
|
||||
TargetValue: rc.Target,
|
||||
MatchType: rc.MatchType,
|
||||
CompareOperator: rc.CompareOperator,
|
||||
TargetValue: target,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
@@ -20,15 +20,12 @@ type Rule interface {
|
||||
Annotations() ruletypes.Labels
|
||||
Condition() *ruletypes.RuleCondition
|
||||
EvalDelay() valuer.TextDuration
|
||||
EvalWindow() valuer.TextDuration
|
||||
HoldDuration() valuer.TextDuration
|
||||
State() ruletypes.AlertState
|
||||
ActiveAlerts() []*ruletypes.Alert
|
||||
// ActiveAlertsLabelFP returns a map of active alert labels fingerprint
|
||||
ActiveAlertsLabelFP() map[uint64]struct{}
|
||||
|
||||
PreferredChannels() []string
|
||||
|
||||
// Eval evaluates the rule at the given timestamp and returns the number of active alerts.
|
||||
Eval(context.Context, time.Time) (int, error)
|
||||
String() string
|
||||
|
||||
@@ -61,16 +61,16 @@ func (r *ThresholdRule) Type() ruletypes.RuleType {
|
||||
}
|
||||
|
||||
func (r *ThresholdRule) prepareQueryRange(ctx context.Context, ts time.Time) (*qbtypes.QueryRangeRequest, error) {
|
||||
startTs, endTs := r.Timestamps(ts)
|
||||
start, end := startTs.UnixMilli(), endTs.UnixMilli()
|
||||
|
||||
r.logger.InfoContext(
|
||||
ctx, "prepare query range request v5",
|
||||
slog.Int64("ts", ts.UnixMilli()),
|
||||
slog.Int64("eval_window", r.evalWindow.Milliseconds()),
|
||||
slog.Int64("eval_window", end-start),
|
||||
slog.Int64("eval_delay", r.evalDelay.Milliseconds()),
|
||||
)
|
||||
|
||||
startTs, endTs := r.Timestamps(ts)
|
||||
start, end := startTs.UnixMilli(), endTs.UnixMilli()
|
||||
|
||||
req := &qbtypes.QueryRangeRequest{
|
||||
Start: uint64(start),
|
||||
End: uint64(end),
|
||||
@@ -468,12 +468,10 @@ func (r *ThresholdRule) Eval(ctx context.Context, ts time.Time) (int, error) {
|
||||
|
||||
func (r *ThresholdRule) String() string {
|
||||
ar := ruletypes.PostableRule{
|
||||
AlertName: r.name,
|
||||
RuleCondition: r.ruleCondition,
|
||||
EvalWindow: r.evalWindow,
|
||||
Labels: r.labels.Map(),
|
||||
Annotations: r.annotations.Map(),
|
||||
PreferredChannels: r.preferredChannels,
|
||||
AlertName: r.name,
|
||||
RuleCondition: r.ruleCondition,
|
||||
Labels: r.labels.Map(),
|
||||
Annotations: r.annotations.Map(),
|
||||
}
|
||||
|
||||
byt, err := json.Marshal(ar)
|
||||
|
||||
@@ -171,10 +171,7 @@ func TestPrepareParamsForLogs(t *testing.T) {
|
||||
},
|
||||
},
|
||||
},
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
Target: &[]float64{0.0}[0],
|
||||
SelectedQuery: "A",
|
||||
SelectedQuery: "A",
|
||||
},
|
||||
}
|
||||
|
||||
@@ -183,9 +180,9 @@ func TestPrepareParamsForLogs(t *testing.T) {
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
{
|
||||
TargetValue: postableRule.RuleCondition.Target,
|
||||
MatchType: postableRule.RuleCondition.MatchType,
|
||||
CompareOperator: postableRule.RuleCondition.CompareOperator,
|
||||
TargetValue: &[]float64{0.0}[0],
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
},
|
||||
},
|
||||
}
|
||||
@@ -239,10 +236,7 @@ func TestPrepareParamsForLogsFilterExpression(t *testing.T) {
|
||||
},
|
||||
},
|
||||
},
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
Target: &[]float64{0.0}[0],
|
||||
SelectedQuery: "A",
|
||||
SelectedQuery: "A",
|
||||
},
|
||||
Version: "v5",
|
||||
}
|
||||
@@ -252,9 +246,9 @@ func TestPrepareParamsForLogsFilterExpression(t *testing.T) {
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
{
|
||||
TargetValue: postableRule.RuleCondition.Target,
|
||||
MatchType: postableRule.RuleCondition.MatchType,
|
||||
CompareOperator: postableRule.RuleCondition.CompareOperator,
|
||||
TargetValue: &[]float64{0.0}[0],
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
},
|
||||
},
|
||||
}
|
||||
@@ -304,10 +298,7 @@ func TestPrepareParamsForTracesFilterExpression(t *testing.T) {
|
||||
},
|
||||
},
|
||||
},
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
Target: &[]float64{0.0}[0],
|
||||
SelectedQuery: "A",
|
||||
SelectedQuery: "A",
|
||||
},
|
||||
Version: "v5",
|
||||
}
|
||||
@@ -317,9 +308,9 @@ func TestPrepareParamsForTracesFilterExpression(t *testing.T) {
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
{
|
||||
TargetValue: postableRule.RuleCondition.Target,
|
||||
MatchType: postableRule.RuleCondition.MatchType,
|
||||
CompareOperator: postableRule.RuleCondition.CompareOperator,
|
||||
TargetValue: &[]float64{0.0}[0],
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
},
|
||||
},
|
||||
}
|
||||
@@ -366,10 +357,7 @@ func TestPrepareParamsForTraces(t *testing.T) {
|
||||
},
|
||||
},
|
||||
},
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
Target: &[]float64{0.0}[0],
|
||||
SelectedQuery: "A",
|
||||
SelectedQuery: "A",
|
||||
},
|
||||
}
|
||||
|
||||
@@ -378,9 +366,9 @@ func TestPrepareParamsForTraces(t *testing.T) {
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
{
|
||||
TargetValue: postableRule.RuleCondition.Target,
|
||||
MatchType: postableRule.RuleCondition.MatchType,
|
||||
CompareOperator: postableRule.RuleCondition.CompareOperator,
|
||||
TargetValue: &[]float64{0.0}[0],
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsNotEq,
|
||||
},
|
||||
},
|
||||
}
|
||||
@@ -653,11 +641,7 @@ func TestThresholdRuleUnitCombinations(t *testing.T) {
|
||||
ExpectQuery(queryString).
|
||||
WithArgs(nil, nil, nil, nil, nil, nil, nil, nil).
|
||||
WillReturnRows(rows)
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -867,11 +851,7 @@ func TestThresholdRuleTracesLink(t *testing.T) {
|
||||
|
||||
querier := prepareQuerierForTraces(t, telemetryStore, keysMap)
|
||||
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -975,11 +955,7 @@ func TestThresholdRuleAITracesLink(t *testing.T) {
|
||||
|
||||
querier := prepareQuerierForAITraces(t, telemetryStore, keysMap)
|
||||
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -1083,11 +1059,7 @@ func TestThresholdRuleLogsLink(t *testing.T) {
|
||||
|
||||
querier := prepareQuerierForLogs(t, telemetryStore, keysMap)
|
||||
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -1256,11 +1228,7 @@ func TestMultipleThresholdRule(t *testing.T) {
|
||||
querier, mockMetadataStore := prepareQuerierForMetrics(t, telemetryStore)
|
||||
mockMetadataStore.TypeMap["signoz_calls_total"] = metrictypes.SumType
|
||||
|
||||
postableRule.RuleCondition.CompareOperator = c.compareOperator
|
||||
postableRule.RuleCondition.MatchType = c.matchType
|
||||
postableRule.RuleCondition.Target = &c.target
|
||||
postableRule.RuleCondition.CompositeQuery.Unit = c.yAxisUnit
|
||||
postableRule.RuleCondition.TargetUnit = c.targetUnit
|
||||
postableRule.RuleCondition.Thresholds = &ruletypes.RuleThresholdData{
|
||||
Kind: ruletypes.BasicThresholdKind,
|
||||
Spec: ruletypes.BasicRuleThresholds{
|
||||
@@ -1930,8 +1898,6 @@ func TestThresholdEval_RequireMinPoints(t *testing.T) {
|
||||
Frequency: valuer.MustParseTextDuration("1m"),
|
||||
}},
|
||||
RuleCondition: &ruletypes.RuleCondition{
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompositeQuery: &ruletypes.AlertCompositeQuery{
|
||||
QueryType: ruletypes.QueryTypeBuilder,
|
||||
Queries: []qbtypes.QueryEnvelope{{
|
||||
@@ -2050,7 +2016,6 @@ func TestThresholdEval_RequireMinPoints(t *testing.T) {
|
||||
mockMetadataStore.TypeMap["signoz_calls_total"] = metrictypes.SumType
|
||||
|
||||
rc := postableRule.RuleCondition
|
||||
rc.Target = &c.target
|
||||
rc.RequireMinPoints = c.requireMinPoints
|
||||
rc.RequiredNumPoints = c.requiredNumPoints
|
||||
rc.Thresholds = &ruletypes.RuleThresholdData{
|
||||
@@ -2059,8 +2024,8 @@ func TestThresholdEval_RequireMinPoints(t *testing.T) {
|
||||
{
|
||||
Name: postableRule.AlertName,
|
||||
TargetValue: &c.target,
|
||||
MatchType: rc.MatchType,
|
||||
CompareOperator: rc.CompareOperator,
|
||||
MatchType: ruletypes.AtleastOnce,
|
||||
CompareOperator: ruletypes.ValueIsAbove,
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
@@ -3,7 +3,6 @@ package ruler
|
||||
import "net/http"
|
||||
|
||||
type Handler interface {
|
||||
ListRules(http.ResponseWriter, *http.Request)
|
||||
ListRulesV3(http.ResponseWriter, *http.Request)
|
||||
GetRuleByID(http.ResponseWriter, *http.Request)
|
||||
CreateRule(http.ResponseWriter, *http.Request)
|
||||
|
||||
@@ -25,24 +25,6 @@ func NewHandler(ruler ruler.Ruler) ruler.Handler {
|
||||
return &handler{ruler: ruler}
|
||||
}
|
||||
|
||||
func (handler *handler) ListRules(rw http.ResponseWriter, req *http.Request) {
|
||||
ctx, cancel := context.WithTimeout(req.Context(), 30*time.Second)
|
||||
defer cancel()
|
||||
|
||||
rules, err := handler.ruler.ListRuleStates(ctx)
|
||||
if err != nil {
|
||||
render.Error(rw, err)
|
||||
return
|
||||
}
|
||||
|
||||
view := make([]*ruletypes.Rule, 0, len(rules.Rules))
|
||||
for _, rule := range rules.Rules {
|
||||
view = append(view, ruletypes.NewRule(rule))
|
||||
}
|
||||
|
||||
render.Success(rw, http.StatusOK, view)
|
||||
}
|
||||
|
||||
func (handler *handler) ListRulesV3(rw http.ResponseWriter, req *http.Request) {
|
||||
ctx, cancel := context.WithTimeout(req.Context(), 30*time.Second)
|
||||
defer cancel()
|
||||
|
||||
@@ -262,6 +262,7 @@ func NewSQLMigrationProviderFactories(
|
||||
sqlmigration.NewAddRuleViewFactory(sqlstore, sqlschema),
|
||||
sqlmigration.NewAddFieldTuplesFactory(sqlstore),
|
||||
sqlmigration.NewKeepRelevantLLMPricingRulesFactory(),
|
||||
sqlmigration.NewMigrateRulesV1ToV2Alpha1Factory(sqlstore),
|
||||
)
|
||||
}
|
||||
|
||||
|
||||
476
pkg/sqlmigration/136_migrate_rules_v1_to_v2alpha1.go
Normal file
476
pkg/sqlmigration/136_migrate_rules_v1_to_v2alpha1.go
Normal file
@@ -0,0 +1,476 @@
|
||||
package sqlmigration
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"log/slog"
|
||||
"slices"
|
||||
"time"
|
||||
|
||||
"github.com/uptrace/bun"
|
||||
"github.com/uptrace/bun/migrate"
|
||||
|
||||
"github.com/SigNoz/signoz/pkg/errors"
|
||||
"github.com/SigNoz/signoz/pkg/factory"
|
||||
"github.com/SigNoz/signoz/pkg/sqlstore"
|
||||
)
|
||||
|
||||
const (
|
||||
schemaVersionV2Alpha1 = "v2alpha1"
|
||||
ruleTypeThreshold = "threshold_rule"
|
||||
ruleTypeProm = "promql_rule"
|
||||
ruleTypeAnomaly = "anomaly_rule"
|
||||
queryTypeBuilder = "builder"
|
||||
queryTypePromQL = "promql"
|
||||
compareOperatorBelow = "2"
|
||||
criticalThresholdName = "critical"
|
||||
labelSeverity = "severity"
|
||||
)
|
||||
|
||||
// The values the threshold decoder accepts; a stored value outside them fails
|
||||
// every later read of the migrated row.
|
||||
var (
|
||||
supportedCompareOperators = []string{"1", "above", ">", "2", "below", "<", "3", "equal", "eq", "=", "4", "not_equal", "not_eq", "!=", "5", "above_or_equal", "above_or_eq", ">=", "6", "below_or_equal", "below_or_eq", "<=", "7", "outside_bounds"}
|
||||
supportedMatchTypes = []string{"1", "at_least_once", "2", "all_the_times", "3", "on_average", "avg", "4", "in_total", "sum", "5", "last"}
|
||||
)
|
||||
|
||||
type migrateRulesV1ToV2Alpha1 struct {
|
||||
sqlstore sqlstore.SQLStore
|
||||
logger *slog.Logger
|
||||
}
|
||||
|
||||
func NewMigrateRulesV1ToV2Alpha1Factory(sqlstore sqlstore.SQLStore) factory.ProviderFactory[SQLMigration, Config] {
|
||||
return factory.NewProviderFactory(factory.MustNewName("migrate_rules_v1_to_v2alpha1"), func(ctx context.Context, ps factory.ProviderSettings, c Config) (SQLMigration, error) {
|
||||
return &migrateRulesV1ToV2Alpha1{sqlstore: sqlstore, logger: ps.Logger}, nil
|
||||
})
|
||||
}
|
||||
|
||||
func (migration *migrateRulesV1ToV2Alpha1) Register(migrations *migrate.Migrations) error {
|
||||
return migrations.Register(migration.Up, migration.Down)
|
||||
}
|
||||
|
||||
// Up rewrites every v1 rule in every org as v2alpha1 in place. A row that does
|
||||
// not decode, or a v1 rule that cannot be rewritten, is logged and left as is;
|
||||
// a failed write aborts the run and rolls back every org.
|
||||
func (migration *migrateRulesV1ToV2Alpha1) Up(ctx context.Context, _ *bun.DB) error {
|
||||
return migration.sqlstore.RunInTxCtx(ctx, nil, func(ctx context.Context) error {
|
||||
var orgIDs []string
|
||||
if err := migration.sqlstore.BunDBCtx(ctx).NewSelect().Table("organizations").Column("id").Scan(ctx, &orgIDs); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
for _, orgID := range orgIDs {
|
||||
if err := migration.convertOrg(ctx, orgID); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
})
|
||||
}
|
||||
|
||||
func (migration *migrateRulesV1ToV2Alpha1) convertOrg(ctx context.Context, orgID string) error {
|
||||
db := migration.sqlstore.BunDBCtx(ctx)
|
||||
|
||||
rules := make([]*ruleRow, 0)
|
||||
if err := db.NewSelect().Model(&rules).Where("org_id = ?", orgID).Scan(ctx); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
routes := make([]*routePolicyRow, 0)
|
||||
if err := db.NewSelect().Model(&routes).Where("org_id = ?", orgID).Scan(ctx); err != nil {
|
||||
return err
|
||||
}
|
||||
routesByName := make(map[string][]*routePolicyRow)
|
||||
for _, route := range routes {
|
||||
routesByName[route.Name] = append(routesByName[route.Name], route)
|
||||
}
|
||||
|
||||
var migrated, skipped, undecodable, unmigratable int
|
||||
for _, stored := range rules {
|
||||
var header struct {
|
||||
SchemaVersion string `json:"schemaVersion"`
|
||||
}
|
||||
if err := json.Unmarshal([]byte(stored.Data), &header); err != nil {
|
||||
undecodable++
|
||||
migration.logger.WarnContext(ctx, "failed to decode stored rule", slog.String("org_id", orgID), slog.String("rule_id", stored.ID), errors.Attr(err))
|
||||
continue
|
||||
}
|
||||
if header.SchemaVersion == schemaVersionV2Alpha1 {
|
||||
skipped++
|
||||
continue
|
||||
}
|
||||
|
||||
rule := new(postableRule)
|
||||
if err := json.Unmarshal([]byte(stored.Data), rule); err != nil {
|
||||
undecodable++
|
||||
migration.logger.WarnContext(ctx, "failed to decode stored rule", slog.String("org_id", orgID), slog.String("rule_id", stored.ID), errors.Attr(err))
|
||||
continue
|
||||
}
|
||||
|
||||
route, err := selectRoute(routesByName[stored.ID], rule.PreferredChannels)
|
||||
if err != nil {
|
||||
unmigratable++
|
||||
migration.logger.WarnContext(ctx, "leaving rule at v1", slog.String("org_id", orgID), slog.String("rule_id", stored.ID), errors.Attr(err))
|
||||
continue
|
||||
}
|
||||
expression := ""
|
||||
if route != nil {
|
||||
expression = route.Expression
|
||||
}
|
||||
|
||||
if err := convertRule(stored.ID, rule, route); err != nil {
|
||||
unmigratable++
|
||||
migration.logger.WarnContext(ctx, "leaving rule at v1", slog.String("org_id", orgID), slog.String("rule_id", stored.ID), errors.Attr(err))
|
||||
continue
|
||||
}
|
||||
|
||||
data, err := marshalUnescaped(rule)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
stored.Data = string(data)
|
||||
if _, err := db.NewUpdate().Model(stored).Column("data").WherePK().Exec(ctx); err != nil {
|
||||
return err
|
||||
}
|
||||
if route != nil && route.Expression != expression {
|
||||
if _, err := db.NewUpdate().Model(route).Column("expression").WherePK().Exec(ctx); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
migrated++
|
||||
}
|
||||
|
||||
migration.logger.InfoContext(ctx, "migrated rules from v1 to v2alpha1", slog.String("org_id", orgID), slog.Int("total", len(rules)), slog.Int("migrated", migrated), slog.Int("skipped", skipped), slog.Int("undecodable", undecodable), slog.Int("unmigratable", unmigratable))
|
||||
return nil
|
||||
}
|
||||
|
||||
// convertRule rewrites rule in place, in the order the v1 decode derived the
|
||||
// same values. Version and source are left as stored. A nil route is a rule
|
||||
// that never had channels.
|
||||
func convertRule(ruleID string, rule *postableRule, route *routePolicyRow) error {
|
||||
if rule.RuleCondition == nil || rule.RuleCondition.CompositeQuery == nil {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "condition or condition.compositeQuery is null")
|
||||
}
|
||||
if err := rule.validateThreshold(); err != nil {
|
||||
return err
|
||||
}
|
||||
channels := []string{}
|
||||
if route != nil {
|
||||
channels = route.Channels
|
||||
}
|
||||
|
||||
rule.fillDurationDefaults()
|
||||
if err := rule.validateDurations(); err != nil {
|
||||
return err
|
||||
}
|
||||
rule.deriveRuleType()
|
||||
rule.negateAnomalyBelowTarget()
|
||||
rule.repairEmptySeverity(ruleID, route)
|
||||
rule.relocateThreshold(channels)
|
||||
rule.relocateEvaluation()
|
||||
rule.synthesizeNotificationSettings()
|
||||
rule.SchemaVersion = schemaVersionV2Alpha1
|
||||
rule.clearV1Fields()
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// selectRoute picks the rule's route_policy row, named after the rule. A rule
|
||||
// that never had channels got no row from migration 049 and migrates with
|
||||
// none; a rule with preferred channels but no row would lose them, so it is
|
||||
// left at v1.
|
||||
func selectRoute(routes []*routePolicyRow, preferredChannels []string) (*routePolicyRow, error) {
|
||||
switch {
|
||||
case len(routes) == 1:
|
||||
return routes[0], nil
|
||||
case len(routes) > 1:
|
||||
return nil, errors.NewInvalidInputf(errors.CodeInvalidInput, "expected one route_policy row, found %d", len(routes))
|
||||
case len(preferredChannels) > 0:
|
||||
return nil, errors.NewInvalidInputf(errors.CodeInvalidInput, "no route_policy row for a rule with %d preferred channels", len(preferredChannels))
|
||||
default:
|
||||
return nil, nil
|
||||
}
|
||||
}
|
||||
|
||||
type ruleRow struct {
|
||||
bun.BaseModel `bun:"table:rule"`
|
||||
|
||||
ID string `bun:"id,pk"`
|
||||
OrgID string `bun:"org_id"`
|
||||
Data string `bun:"data"`
|
||||
}
|
||||
|
||||
type routePolicyRow struct {
|
||||
bun.BaseModel `bun:"table:route_policy"`
|
||||
|
||||
ID string `bun:"id,pk"`
|
||||
OrgID string `bun:"org_id"`
|
||||
Name string `bun:"name"`
|
||||
Expression string `bun:"expression"`
|
||||
Channels []string `bun:"channels,type:jsonb"`
|
||||
}
|
||||
|
||||
// postableRule mirrors PostableRule key for key, in its order. Keys the
|
||||
// crosswalk reads or writes are typed; the rest pass through as stored.
|
||||
type postableRule struct {
|
||||
AlertName json.RawMessage `json:"alert,omitempty"`
|
||||
AlertType json.RawMessage `json:"alertType,omitempty"`
|
||||
Description json.RawMessage `json:"description,omitempty"`
|
||||
RuleType string `json:"ruleType"`
|
||||
EvalWindow textDuration `json:"evalWindow,omitzero"`
|
||||
Frequency textDuration `json:"frequency,omitzero"`
|
||||
RuleCondition *ruleCondition `json:"condition"`
|
||||
Labels map[string]string `json:"labels,omitempty"`
|
||||
Annotations json.RawMessage `json:"annotations,omitempty"`
|
||||
Disabled json.RawMessage `json:"disabled,omitempty"`
|
||||
Source json.RawMessage `json:"source,omitempty"`
|
||||
PreferredChannels []string `json:"preferredChannels,omitempty"`
|
||||
Version json.RawMessage `json:"version,omitempty"`
|
||||
Evaluation *evaluationEnvelope `json:"evaluation,omitempty"`
|
||||
SchemaVersion string `json:"schemaVersion,omitempty"`
|
||||
NotificationSettings *notificationSettings `json:"notificationSettings,omitempty"`
|
||||
}
|
||||
|
||||
type ruleCondition struct {
|
||||
CompositeQuery *compositeQuery `json:"compositeQuery"`
|
||||
CompareOperator string `json:"op,omitempty"`
|
||||
Target *float64 `json:"target,omitempty"`
|
||||
AlertOnAbsent bool `json:"alertOnAbsent,omitempty"`
|
||||
AbsentFor json.RawMessage `json:"absentFor,omitempty"`
|
||||
MatchType string `json:"matchType,omitempty"`
|
||||
TargetUnit string `json:"targetUnit,omitempty"`
|
||||
Algorithm json.RawMessage `json:"algorithm,omitempty"`
|
||||
Seasonality json.RawMessage `json:"seasonality,omitempty"`
|
||||
SelectedQuery json.RawMessage `json:"selectedQueryName,omitempty"`
|
||||
RequireMinPoints json.RawMessage `json:"requireMinPoints,omitempty"`
|
||||
RequiredNumPoints json.RawMessage `json:"requiredNumPoints,omitempty"`
|
||||
Thresholds *ruleThresholds `json:"thresholds,omitempty"`
|
||||
}
|
||||
|
||||
type compositeQuery struct {
|
||||
Queries json.RawMessage `json:"queries,omitempty"`
|
||||
PanelType json.RawMessage `json:"panelType,omitempty"`
|
||||
QueryType string `json:"queryType"`
|
||||
Unit json.RawMessage `json:"unit,omitempty"`
|
||||
}
|
||||
|
||||
type ruleThresholds struct {
|
||||
Kind string `json:"kind"`
|
||||
Spec []basicThreshold `json:"spec"`
|
||||
}
|
||||
|
||||
type basicThreshold struct {
|
||||
Name string `json:"name"`
|
||||
TargetValue *float64 `json:"target"`
|
||||
TargetUnit string `json:"targetUnit"`
|
||||
RecoveryTarget *float64 `json:"recoveryTarget"`
|
||||
MatchType string `json:"matchType"`
|
||||
CompareOperator string `json:"op"`
|
||||
Channels []string `json:"channels"`
|
||||
}
|
||||
|
||||
type evaluationEnvelope struct {
|
||||
Kind string `json:"kind"`
|
||||
Spec rollingWindow `json:"spec"`
|
||||
}
|
||||
|
||||
type rollingWindow struct {
|
||||
EvalWindow textDuration `json:"evalWindow"`
|
||||
Frequency textDuration `json:"frequency"`
|
||||
}
|
||||
|
||||
type notificationSettings struct {
|
||||
Renotify *renotify `json:"renotify,omitempty"`
|
||||
UsePolicy bool `json:"usePolicy"`
|
||||
}
|
||||
|
||||
type renotify struct {
|
||||
Enabled bool `json:"enabled"`
|
||||
ReNotifyInterval textDuration `json:"interval,omitzero"`
|
||||
AlertStates []string `json:"alertStates,omitzero"`
|
||||
}
|
||||
|
||||
// validateThreshold rejects what the v1 write path once let through: a
|
||||
// missing target, or an op or matchType the threshold decoder does not
|
||||
// accept. Each fails every later read of the migrated row.
|
||||
func (rule *postableRule) validateThreshold() error {
|
||||
condition := rule.RuleCondition
|
||||
if condition.Target == nil {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "condition.target is null")
|
||||
}
|
||||
if !slices.Contains(supportedCompareOperators, condition.CompareOperator) {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "condition.op: unsupported value %q", condition.CompareOperator)
|
||||
}
|
||||
if !slices.Contains(supportedMatchTypes, condition.MatchType) {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "condition.matchType: unsupported value %q", condition.MatchType)
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (rule *postableRule) fillDurationDefaults() {
|
||||
if rule.EvalWindow.IsZero() {
|
||||
rule.EvalWindow = mustParseTextDuration("5m")
|
||||
}
|
||||
if rule.Frequency.IsZero() {
|
||||
rule.Frequency = mustParseTextDuration("1m")
|
||||
}
|
||||
}
|
||||
|
||||
// validateDurations rejects what the v1 write path never checked: a negative
|
||||
// evalWindow or frequency, which no read of a rolling window accepts.
|
||||
func (rule *postableRule) validateDurations() error {
|
||||
if !rule.EvalWindow.IsPositive() || !rule.Frequency.IsPositive() {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "evalWindow %s or frequency %s is negative", rule.EvalWindow, rule.Frequency)
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (rule *postableRule) deriveRuleType() {
|
||||
switch rule.RuleCondition.CompositeQuery.QueryType {
|
||||
case queryTypeBuilder:
|
||||
if rule.RuleType == "" {
|
||||
rule.RuleType = ruleTypeThreshold
|
||||
}
|
||||
case queryTypePromQL:
|
||||
rule.RuleType = ruleTypeProm
|
||||
}
|
||||
}
|
||||
|
||||
// negateAnomalyBelowTarget stores "below 3" as -3, since anomaly scores are
|
||||
// signed. Only the raw "2" negates; "below" and "<" keep the sign the rule
|
||||
// evaluates with today.
|
||||
func (rule *postableRule) negateAnomalyBelowTarget() {
|
||||
condition := rule.RuleCondition
|
||||
if rule.RuleType != ruleTypeAnomaly || condition.CompareOperator != compareOperatorBelow || condition.Target == nil {
|
||||
return
|
||||
}
|
||||
negated := -*condition.Target
|
||||
condition.Target = &negated
|
||||
}
|
||||
|
||||
// repairEmptySeverity rewrites a severity of "" as critical in labels and in
|
||||
// the route expression that matches alerts on it, since a threshold named ""
|
||||
// fails every later read.
|
||||
func (rule *postableRule) repairEmptySeverity(ruleID string, route *routePolicyRow) {
|
||||
if severity, ok := rule.Labels[labelSeverity]; !ok || severity != "" {
|
||||
return
|
||||
}
|
||||
rule.Labels[labelSeverity] = criticalThresholdName
|
||||
if route != nil {
|
||||
route.Expression = fmt.Sprintf(`threshold.name == "%s" && ruleId == "%s"`, criticalThresholdName, ruleID)
|
||||
}
|
||||
}
|
||||
|
||||
func (rule *postableRule) relocateThreshold(channels []string) {
|
||||
thresholdName := criticalThresholdName
|
||||
if severity, ok := rule.Labels[labelSeverity]; ok {
|
||||
thresholdName = severity
|
||||
}
|
||||
|
||||
condition := rule.RuleCondition
|
||||
condition.Thresholds = &ruleThresholds{
|
||||
Kind: "basic",
|
||||
Spec: []basicThreshold{{
|
||||
Name: thresholdName,
|
||||
TargetValue: condition.Target,
|
||||
TargetUnit: condition.TargetUnit,
|
||||
MatchType: condition.MatchType,
|
||||
CompareOperator: condition.CompareOperator,
|
||||
Channels: channels,
|
||||
}},
|
||||
}
|
||||
}
|
||||
|
||||
func (rule *postableRule) relocateEvaluation() {
|
||||
rule.Evaluation = &evaluationEnvelope{
|
||||
Kind: "rolling",
|
||||
Spec: rollingWindow{EvalWindow: rule.EvalWindow, Frequency: rule.Frequency},
|
||||
}
|
||||
}
|
||||
|
||||
func (rule *postableRule) synthesizeNotificationSettings() {
|
||||
alertStates := []string{"firing"}
|
||||
if rule.RuleCondition.AlertOnAbsent {
|
||||
alertStates = append(alertStates, "nodata")
|
||||
}
|
||||
rule.NotificationSettings = ¬ificationSettings{
|
||||
Renotify: &renotify{
|
||||
Enabled: true,
|
||||
ReNotifyInterval: mustParseTextDuration("4h"),
|
||||
AlertStates: alertStates,
|
||||
},
|
||||
UsePolicy: false,
|
||||
}
|
||||
}
|
||||
|
||||
func (rule *postableRule) clearV1Fields() {
|
||||
rule.EvalWindow = textDuration{}
|
||||
rule.Frequency = textDuration{}
|
||||
rule.PreferredChannels = nil
|
||||
rule.RuleCondition.Target = nil
|
||||
rule.RuleCondition.CompareOperator = ""
|
||||
rule.RuleCondition.MatchType = ""
|
||||
rule.RuleCondition.TargetUnit = ""
|
||||
}
|
||||
|
||||
// textDuration decodes a duration the way the app's TextDuration does, a
|
||||
// string or a number of nanoseconds, and writes a string back in its stored
|
||||
// spelling.
|
||||
type textDuration struct {
|
||||
text string
|
||||
value time.Duration
|
||||
}
|
||||
|
||||
func mustParseTextDuration(text string) textDuration {
|
||||
value, err := time.ParseDuration(text)
|
||||
if err != nil {
|
||||
panic(err)
|
||||
}
|
||||
return textDuration{text: text, value: value}
|
||||
}
|
||||
|
||||
func (d textDuration) IsZero() bool {
|
||||
return d.value == 0
|
||||
}
|
||||
|
||||
func (d textDuration) IsPositive() bool {
|
||||
return d.value > 0
|
||||
}
|
||||
|
||||
func (d textDuration) String() string {
|
||||
if d.text != "" {
|
||||
return d.text
|
||||
}
|
||||
return d.value.String()
|
||||
}
|
||||
|
||||
func (d textDuration) MarshalJSON() ([]byte, error) {
|
||||
return json.Marshal(d.String())
|
||||
}
|
||||
|
||||
func (d *textDuration) UnmarshalJSON(data []byte) error {
|
||||
var raw any
|
||||
if err := json.Unmarshal(data, &raw); err != nil {
|
||||
return err
|
||||
}
|
||||
switch value := raw.(type) {
|
||||
case float64:
|
||||
*d = textDuration{value: time.Duration(value)}
|
||||
return nil
|
||||
case string:
|
||||
parsed, err := time.ParseDuration(value)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
*d = textDuration{text: value, value: parsed}
|
||||
return nil
|
||||
default:
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "invalid duration %s", data)
|
||||
}
|
||||
}
|
||||
|
||||
func (migration *migrateRulesV1ToV2Alpha1) Down(context.Context, *bun.DB) error {
|
||||
return nil
|
||||
}
|
||||
@@ -610,10 +610,6 @@ type ConfigStore interface {
|
||||
|
||||
// ListAllChannels returns the list of channels for all organizations.
|
||||
ListAllChannels(context.Context) ([]*Channel, error)
|
||||
|
||||
// GetMatchers gets a list of matchers per organization.
|
||||
// Matchers is an array of ruleId to receiver names.
|
||||
GetMatchers(context.Context, string) (map[string][]string, error)
|
||||
}
|
||||
|
||||
// MarshalSecretValue if set to true will expose Secret type
|
||||
|
||||
@@ -111,18 +111,14 @@ type AlertCompositeQuery struct {
|
||||
|
||||
type RuleCondition struct {
|
||||
CompositeQuery *AlertCompositeQuery `json:"compositeQuery" required:"true"`
|
||||
CompareOperator CompareOperator `json:"op,omitzero"`
|
||||
Target *float64 `json:"target,omitempty" format:"double"`
|
||||
AlertOnAbsent bool `json:"alertOnAbsent,omitempty"`
|
||||
AbsentFor uint64 `json:"absentFor,omitempty"`
|
||||
MatchType MatchType `json:"matchType,omitzero"`
|
||||
TargetUnit string `json:"targetUnit,omitempty"`
|
||||
Algorithm string `json:"algorithm,omitempty"`
|
||||
Seasonality Seasonality `json:"seasonality,omitzero"`
|
||||
SelectedQuery string `json:"selectedQueryName,omitempty"`
|
||||
RequireMinPoints bool `json:"requireMinPoints,omitempty"`
|
||||
RequiredNumPoints int `json:"requiredNumPoints,omitempty"`
|
||||
Thresholds *RuleThresholdData `json:"thresholds,omitempty"`
|
||||
Thresholds *RuleThresholdData `json:"thresholds" required:"true"`
|
||||
}
|
||||
|
||||
func (rc *RuleCondition) SelectedQueryName() string {
|
||||
|
||||
@@ -9,6 +9,7 @@ import (
|
||||
"unicode/utf8"
|
||||
|
||||
"github.com/prometheus/alertmanager/config"
|
||||
"github.com/swaggest/jsonschema-go"
|
||||
|
||||
"github.com/SigNoz/signoz/pkg/errors"
|
||||
"github.com/SigNoz/signoz/pkg/types"
|
||||
@@ -48,7 +49,6 @@ func (t AlertType) BuilderQueryType() qbtypes.QueryType {
|
||||
}
|
||||
|
||||
const (
|
||||
DefaultSchemaVersion = "v1"
|
||||
SchemaVersionV2Alpha1 = "v2alpha1"
|
||||
)
|
||||
|
||||
@@ -60,12 +60,10 @@ const (
|
||||
|
||||
// PostableRule is used to create alerting rule from HTTP api.
|
||||
type PostableRule struct {
|
||||
AlertName string `json:"alert" required:"true"`
|
||||
AlertType AlertType `json:"alertType" required:"true"`
|
||||
Description string `json:"description,omitempty"`
|
||||
RuleType RuleType `json:"ruleType" required:"true"`
|
||||
EvalWindow valuer.TextDuration `json:"evalWindow,omitzero"`
|
||||
Frequency valuer.TextDuration `json:"frequency,omitzero"`
|
||||
AlertName string `json:"alert" required:"true"`
|
||||
AlertType AlertType `json:"alertType" required:"true"`
|
||||
Description string `json:"description,omitempty"`
|
||||
RuleType RuleType `json:"ruleType" required:"true"`
|
||||
|
||||
RuleCondition *RuleCondition `json:"condition" required:"true"`
|
||||
Labels map[string]string `json:"labels,omitempty"`
|
||||
@@ -76,14 +74,12 @@ type PostableRule struct {
|
||||
// Source captures the source url where rule has been created
|
||||
Source string `json:"source,omitempty"`
|
||||
|
||||
PreferredChannels []string `json:"preferredChannels,omitempty"`
|
||||
|
||||
Version string `json:"version"`
|
||||
|
||||
Evaluation *EvaluationEnvelope `json:"evaluation,omitempty"`
|
||||
SchemaVersion string `json:"schemaVersion,omitempty"`
|
||||
Evaluation *EvaluationEnvelope `json:"evaluation" required:"true"`
|
||||
SchemaVersion string `json:"schemaVersion" required:"true" enum:"v2alpha1"`
|
||||
|
||||
NotificationSettings *NotificationSettings `json:"notificationSettings,omitempty"`
|
||||
NotificationSettings *NotificationSettings `json:"notificationSettings" required:"true"`
|
||||
}
|
||||
|
||||
type NotificationSettings struct {
|
||||
@@ -229,27 +225,11 @@ func (ns *NotificationSettings) UnmarshalJSON(data []byte) error {
|
||||
// processRuleDefaults applies the default values
|
||||
// for the rule options that are blank or unset.
|
||||
func (r *PostableRule) processRuleDefaults() {
|
||||
if r.SchemaVersion == "" {
|
||||
r.SchemaVersion = DefaultSchemaVersion
|
||||
}
|
||||
|
||||
// TODO(srikanthccv): remove as this is now a legacy field
|
||||
if r.Version == "" {
|
||||
r.Version = "v5"
|
||||
}
|
||||
|
||||
// v2alpha1 uses the Evaluation envelope for window/frequency;
|
||||
// only default top-level fields for v1.
|
||||
if r.SchemaVersion != SchemaVersionV2Alpha1 {
|
||||
if r.EvalWindow.IsZero() {
|
||||
r.EvalWindow = valuer.MustParseTextDuration("5m")
|
||||
}
|
||||
|
||||
if r.Frequency.IsZero() {
|
||||
r.Frequency = valuer.MustParseTextDuration("1m")
|
||||
}
|
||||
}
|
||||
|
||||
if r.RuleCondition != nil && r.RuleCondition.CompositeQuery != nil {
|
||||
switch r.RuleCondition.CompositeQuery.QueryType {
|
||||
case QueryTypeBuilder:
|
||||
@@ -259,71 +239,6 @@ func (r *PostableRule) processRuleDefaults() {
|
||||
case QueryTypePromQL:
|
||||
r.RuleType = RuleTypeProm
|
||||
}
|
||||
|
||||
if r.SchemaVersion == DefaultSchemaVersion {
|
||||
thresholdName := CriticalThresholdName
|
||||
if r.Labels != nil {
|
||||
if severity, ok := r.Labels["severity"]; ok {
|
||||
thresholdName = severity
|
||||
}
|
||||
}
|
||||
|
||||
// For anomaly detection with ValueIsBelow, negate the target
|
||||
targetValue := r.RuleCondition.Target
|
||||
if r.RuleType == RuleTypeAnomaly && r.RuleCondition.CompareOperator == ValueIsBelow && targetValue != nil {
|
||||
negated := -1 * *targetValue
|
||||
targetValue = &negated
|
||||
}
|
||||
|
||||
thresholdData := RuleThresholdData{
|
||||
Kind: BasicThresholdKind,
|
||||
Spec: BasicRuleThresholds{{
|
||||
Name: thresholdName,
|
||||
TargetUnit: r.RuleCondition.TargetUnit,
|
||||
TargetValue: targetValue,
|
||||
MatchType: r.RuleCondition.MatchType,
|
||||
CompareOperator: r.RuleCondition.CompareOperator,
|
||||
Channels: r.PreferredChannels,
|
||||
}},
|
||||
}
|
||||
r.RuleCondition.Thresholds = &thresholdData
|
||||
r.Evaluation = &EvaluationEnvelope{RollingEvaluation, RollingWindow{EvalWindow: r.EvalWindow, Frequency: r.Frequency}}
|
||||
r.NotificationSettings = &NotificationSettings{
|
||||
Renotify: &Renotify{
|
||||
Enabled: true,
|
||||
ReNotifyInterval: valuer.MustParseTextDuration("4h"),
|
||||
AlertStates: []AlertState{StateFiring},
|
||||
},
|
||||
}
|
||||
if r.RuleCondition.AlertOnAbsent {
|
||||
r.NotificationSettings.Renotify.AlertStates = append(r.NotificationSettings.Renotify.AlertStates, StateNoData)
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (r *PostableRule) MarshalJSON() ([]byte, error) {
|
||||
type Alias PostableRule
|
||||
|
||||
switch r.SchemaVersion {
|
||||
case DefaultSchemaVersion:
|
||||
copyStruct := *r
|
||||
aux := Alias(copyStruct)
|
||||
if aux.RuleCondition != nil {
|
||||
aux.RuleCondition.Thresholds = nil
|
||||
}
|
||||
aux.Evaluation = nil
|
||||
aux.SchemaVersion = ""
|
||||
aux.NotificationSettings = nil
|
||||
return json.Marshal(aux)
|
||||
case SchemaVersionV2Alpha1:
|
||||
copyStruct := *r
|
||||
aux := Alias(copyStruct)
|
||||
return json.Marshal(aux)
|
||||
default:
|
||||
copyStruct := *r
|
||||
aux := Alias(copyStruct)
|
||||
return json.Marshal(aux)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -337,6 +252,19 @@ func (r *PostableRule) UnmarshalJSON(bytes []byte) error {
|
||||
return r.validate()
|
||||
}
|
||||
|
||||
var _ jsonschema.Preparer = PostableRule{}
|
||||
|
||||
// PrepareJSONSchema marks version and source deprecated: nothing reads them,
|
||||
// they stay only so existing clients keep a stable schema.
|
||||
func (PostableRule) PrepareJSONSchema(schema *jsonschema.Schema) error {
|
||||
for _, name := range []string{"version", "source"} {
|
||||
if property, ok := schema.Properties[name]; ok && property.TypeObject != nil {
|
||||
property.TypeObject.WithDeprecated(true)
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func isValidLabelName(ln string) bool {
|
||||
if len(ln) == 0 {
|
||||
return false
|
||||
@@ -364,10 +292,6 @@ func (r *PostableRule) validate() error {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "condition: field is required")
|
||||
}
|
||||
|
||||
if r.Version != "v5" {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput, "version: only v5 is supported, got %q", r.Version))
|
||||
}
|
||||
|
||||
for k, v := range r.Labels {
|
||||
if !isValidLabelName(k) {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput, "invalid label name: %s", k))
|
||||
@@ -411,10 +335,6 @@ func (r *PostableRule) Validate() error {
|
||||
return errors.NewInvalidInputf(errors.CodeInvalidInput, "condition: field is required")
|
||||
}
|
||||
|
||||
if r.Version != "v5" {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput, "version: only v5 is supported, got %q", r.Version))
|
||||
}
|
||||
|
||||
if r.AlertType != "" {
|
||||
switch r.AlertType {
|
||||
case AlertTypeMetric, AlertTypeTraces, AlertTypeLogs, AlertTypeExceptions, AlertTypeAITraces:
|
||||
@@ -502,65 +422,18 @@ func (r *PostableRule) Validate() error {
|
||||
|
||||
func (r *PostableRule) validateSchemaVersion() []error {
|
||||
switch r.SchemaVersion {
|
||||
case DefaultSchemaVersion:
|
||||
return r.validateV1()
|
||||
case SchemaVersionV2Alpha1:
|
||||
return r.validateV2Alpha1()
|
||||
default:
|
||||
return []error{errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"schemaVersion: unsupported value %q; must be one of %q, %q",
|
||||
r.SchemaVersion, DefaultSchemaVersion, SchemaVersionV2Alpha1)}
|
||||
"schemaVersion: unsupported value %q; must be %q",
|
||||
r.SchemaVersion, SchemaVersionV2Alpha1)}
|
||||
}
|
||||
}
|
||||
|
||||
func (r *PostableRule) validateV1() []error {
|
||||
var errs []error
|
||||
|
||||
if r.RuleCondition.Target == nil {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"condition.target: field is required for schemaVersion %q", DefaultSchemaVersion))
|
||||
}
|
||||
if r.RuleCondition.CompareOperator.IsZero() {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"condition.op: field is required for schemaVersion %q", DefaultSchemaVersion))
|
||||
} else if err := r.RuleCondition.CompareOperator.Validate(); err != nil {
|
||||
errs = append(errs, err)
|
||||
}
|
||||
if r.RuleCondition.MatchType.IsZero() {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
"condition.matchType: field is required for schemaVersion %q", DefaultSchemaVersion))
|
||||
} else if err := r.RuleCondition.MatchType.Validate(); err != nil {
|
||||
errs = append(errs, err)
|
||||
}
|
||||
|
||||
return errs
|
||||
}
|
||||
|
||||
func (r *PostableRule) validateV2Alpha1() []error {
|
||||
var errs []error
|
||||
|
||||
// TODO(srikanthccv): reject v1-only fields?
|
||||
// if r.RuleCondition.Target != nil {
|
||||
// errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
// "condition.target: field is not used in schemaVersion %q; set target in condition.thresholds entries instead",
|
||||
// SchemaVersionV2Alpha1))
|
||||
// }
|
||||
// if !r.RuleCondition.CompareOperator.IsZero() {
|
||||
// errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
// "condition.op: field is not used in schemaVersion %q; set op in condition.thresholds entries instead",
|
||||
// SchemaVersionV2Alpha1))
|
||||
// }
|
||||
// if !r.RuleCondition.MatchType.IsZero() {
|
||||
// errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
// "condition.matchType: field is not used in schemaVersion %q; set matchType in condition.thresholds entries instead",
|
||||
// SchemaVersionV2Alpha1))
|
||||
// }
|
||||
// if len(r.PreferredChannels) > 0 {
|
||||
// errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
// "preferredChannels: field is not used in schemaVersion %q; set channels in condition.thresholds entries instead",
|
||||
// SchemaVersionV2Alpha1))
|
||||
// }
|
||||
|
||||
// Require v2alpha1-specific fields
|
||||
if r.RuleCondition.Thresholds == nil {
|
||||
errs = append(errs, errors.NewInvalidInputf(errors.CodeInvalidInput,
|
||||
@@ -644,31 +517,6 @@ type GettableRule struct {
|
||||
UpdatedBy *string `json:"updateBy" nullable:"true"`
|
||||
}
|
||||
|
||||
func (g *GettableRule) MarshalJSON() ([]byte, error) {
|
||||
type Alias GettableRule
|
||||
|
||||
switch g.SchemaVersion {
|
||||
case DefaultSchemaVersion:
|
||||
copyStruct := *g
|
||||
aux := Alias(copyStruct)
|
||||
if aux.RuleCondition != nil {
|
||||
aux.RuleCondition.Thresholds = nil
|
||||
}
|
||||
aux.Evaluation = nil
|
||||
aux.SchemaVersion = ""
|
||||
aux.NotificationSettings = nil
|
||||
return json.Marshal(aux)
|
||||
case SchemaVersionV2Alpha1:
|
||||
copyStruct := *g
|
||||
aux := Alias(copyStruct)
|
||||
return json.Marshal(aux)
|
||||
default:
|
||||
copyStruct := *g
|
||||
aux := Alias(copyStruct)
|
||||
return json.Marshal(aux)
|
||||
}
|
||||
}
|
||||
|
||||
// Rule is the v2 API read model for an alerting rule. It aligns audit fields
|
||||
// with the canonical types.TimeAuditable / types.UserAuditable shape used by
|
||||
// PlannedMaintenance and other entities. v1 handlers keep serializing
|
||||
@@ -697,28 +545,3 @@ func NewRule(g *GettableRule) *Rule {
|
||||
}
|
||||
return r
|
||||
}
|
||||
|
||||
func (r *Rule) MarshalJSON() ([]byte, error) {
|
||||
type Alias Rule
|
||||
|
||||
switch r.SchemaVersion {
|
||||
case DefaultSchemaVersion:
|
||||
copyStruct := *r
|
||||
aux := Alias(copyStruct)
|
||||
if aux.RuleCondition != nil {
|
||||
aux.RuleCondition.Thresholds = nil
|
||||
}
|
||||
aux.Evaluation = nil
|
||||
aux.SchemaVersion = ""
|
||||
aux.NotificationSettings = nil
|
||||
return json.Marshal(aux)
|
||||
case SchemaVersionV2Alpha1:
|
||||
copyStruct := *r
|
||||
aux := Alias(copyStruct)
|
||||
return json.Marshal(aux)
|
||||
default:
|
||||
copyStruct := *r
|
||||
aux := Alias(copyStruct)
|
||||
return json.Marshal(aux)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -3,7 +3,6 @@ package ruletypes
|
||||
import (
|
||||
"encoding/json"
|
||||
"testing"
|
||||
"time"
|
||||
|
||||
"github.com/stretchr/testify/assert"
|
||||
"github.com/stretchr/testify/require"
|
||||
@@ -29,8 +28,7 @@ func TestParseIntoRule(t *testing.T) {
|
||||
"alertType": "METRIC_BASED_ALERT",
|
||||
"description": "Test description",
|
||||
"ruleType": "threshold_rule",
|
||||
"evalWindow": "5m",
|
||||
"frequency": "1m",
|
||||
"schemaVersion": "v2alpha1",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
@@ -45,11 +43,11 @@ func TestParseIntoRule(t *testing.T) {
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 10.0,
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
"thresholds": {"kind": "basic", "spec": [{"name": "critical", "target": 10.0, "matchType": "1", "op": "1"}]},
|
||||
"selectedQueryName": "A"
|
||||
},
|
||||
"evaluation": {"kind": "rolling", "spec": {"evalWindow": "5m", "frequency": "1m"}},
|
||||
"notificationSettings": {"renotify": {"enabled": true, "interval": "4h", "alertStates": ["firing"]}}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
@@ -63,46 +61,6 @@ func TestParseIntoRule(t *testing.T) {
|
||||
if rule.RuleCondition.Thresholds.Kind.IsZero() {
|
||||
t.Error("Expected thresholds to be populated")
|
||||
}
|
||||
if rule.RuleCondition.Target == nil {
|
||||
t.Error("Expected target to be populated")
|
||||
}
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "rule with default values applied",
|
||||
initRule: PostableRule{},
|
||||
content: []byte(`{
|
||||
"alert": "DefaultsRule",
|
||||
"ruleType": "threshold_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test_metric", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 5.0,
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
validate: func(t *testing.T, rule *PostableRule) {
|
||||
if rule.EvalWindow.Duration() != 5*time.Minute {
|
||||
t.Errorf("Expected default eval window '5m', got '%v'", rule.EvalWindow)
|
||||
}
|
||||
if rule.Frequency.Duration() != time.Minute {
|
||||
t.Errorf("Expected default frequency '1m', got '%v'", rule.Frequency)
|
||||
}
|
||||
},
|
||||
},
|
||||
{
|
||||
@@ -110,6 +68,7 @@ func TestParseIntoRule(t *testing.T) {
|
||||
initRule: PostableRule{},
|
||||
content: []byte(`{
|
||||
"alert": "PromQLRule",
|
||||
"schemaVersion": "v2alpha1",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
@@ -123,11 +82,11 @@ func TestParseIntoRule(t *testing.T) {
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 10.0,
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
"thresholds": {"kind": "basic", "spec": [{"name": "critical", "target": 10.0, "matchType": "1", "op": "1"}]},
|
||||
"selectedQueryName": "A"
|
||||
},
|
||||
"evaluation": {"kind": "rolling", "spec": {"evalWindow": "5m", "frequency": "1m"}},
|
||||
"notificationSettings": {"renotify": {"enabled": true, "interval": "4h", "alertStates": ["firing"]}}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
@@ -138,9 +97,6 @@ func TestParseIntoRule(t *testing.T) {
|
||||
if rule.RuleCondition.Thresholds.Kind.IsZero() {
|
||||
t.Error("Expected thresholds to be populated")
|
||||
}
|
||||
if rule.RuleCondition.Target == nil {
|
||||
t.Error("Expected target to be populated")
|
||||
}
|
||||
},
|
||||
},
|
||||
}
|
||||
@@ -177,233 +133,6 @@ func TestParseIntoRuleSchemaVersioning(t *testing.T) {
|
||||
expectValidateError bool // Validate() error (write path only)
|
||||
validate func(*testing.T, *PostableRule)
|
||||
}{
|
||||
{
|
||||
name: "schema v1 - threshold name from severity label",
|
||||
initRule: PostableRule{},
|
||||
content: []byte(`{
|
||||
"alert": "SeverityLabelTest",
|
||||
"schemaVersion": "v1",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "cpu_usage", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}],
|
||||
"unit": "percent"
|
||||
},
|
||||
"target": 85.0,
|
||||
"targetUnit": "%",
|
||||
"matchType": "1",
|
||||
"op": "1"
|
||||
},
|
||||
"labels": {
|
||||
"severity": "warning",
|
||||
"team": "platform"
|
||||
}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
validate: func(t *testing.T, rule *PostableRule) {
|
||||
if rule.RuleCondition.Thresholds == nil {
|
||||
t.Fatal("Expected Thresholds to be populated for v1")
|
||||
}
|
||||
|
||||
threshold := rule.RuleCondition.Thresholds
|
||||
if threshold.Kind != BasicThresholdKind {
|
||||
t.Errorf("Expected BasicThresholdKind, got %s", threshold.Kind)
|
||||
}
|
||||
|
||||
specs, ok := threshold.Spec.(BasicRuleThresholds)
|
||||
if !ok {
|
||||
t.Fatalf("Expected BasicRuleThresholds, got %T", threshold.Spec)
|
||||
}
|
||||
|
||||
if len(specs) != 1 {
|
||||
t.Fatalf("Expected 1 threshold spec, got %d", len(specs))
|
||||
}
|
||||
|
||||
spec := specs[0]
|
||||
if spec.Name != "warning" {
|
||||
t.Errorf("Expected threshold name 'warning' from severity label, got '%s'", spec.Name)
|
||||
}
|
||||
|
||||
if spec.TargetUnit != "%" {
|
||||
t.Errorf("Expected TargetUnit '%%', got '%s'", spec.TargetUnit)
|
||||
}
|
||||
if *spec.TargetValue != 85.0 {
|
||||
t.Errorf("Expected TargetValue 85.0, got %v", *spec.TargetValue)
|
||||
}
|
||||
if spec.MatchType != rule.RuleCondition.MatchType {
|
||||
t.Error("Expected MatchType to be copied from RuleCondition")
|
||||
}
|
||||
if spec.CompareOperator != rule.RuleCondition.CompareOperator {
|
||||
t.Error("Expected CompareOp to be copied from RuleCondition")
|
||||
}
|
||||
|
||||
// Verify evaluation envelope is populated
|
||||
if rule.Evaluation == nil {
|
||||
t.Fatal("Expected Evaluation to be populated for v1")
|
||||
}
|
||||
if rule.Evaluation.Kind != RollingEvaluation {
|
||||
t.Errorf("Expected RollingEvaluation, got %s", rule.Evaluation.Kind)
|
||||
}
|
||||
|
||||
// Verify evaluation window matches rule settings
|
||||
if window, ok := rule.Evaluation.Spec.(RollingWindow); ok {
|
||||
if !window.EvalWindow.Equal(rule.EvalWindow) {
|
||||
t.Errorf("Expected Evaluation EvalWindow %v, got %v", rule.EvalWindow, window.EvalWindow)
|
||||
}
|
||||
if !window.Frequency.Equal(rule.Frequency) {
|
||||
t.Errorf("Expected Evaluation Frequency %v, got %v", rule.Frequency, window.Frequency)
|
||||
}
|
||||
} else {
|
||||
t.Errorf("Expected RollingWindow spec, got %T", rule.Evaluation.Spec)
|
||||
}
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "schema v1 - uses critical threshold when no labels",
|
||||
initRule: PostableRule{},
|
||||
content: []byte(`{
|
||||
"alert": "NoLabelsTest",
|
||||
"schemaVersion": "v1",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "memory_usage", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 90.0,
|
||||
"matchType": "1",
|
||||
"op": "1"
|
||||
}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
validate: func(t *testing.T, rule *PostableRule) {
|
||||
if rule.RuleCondition.Thresholds == nil {
|
||||
t.Fatal("Expected Thresholds to be populated")
|
||||
}
|
||||
|
||||
specs, ok := rule.RuleCondition.Thresholds.Spec.(BasicRuleThresholds)
|
||||
if !ok {
|
||||
t.Fatalf("Expected BasicRuleThresholds, got %T", rule.RuleCondition.Thresholds.Spec)
|
||||
}
|
||||
spec := specs[0]
|
||||
// Should default to CriticalThresholdName when no severity label
|
||||
if spec.Name != CriticalThresholdName {
|
||||
t.Errorf("Expected threshold name '%s', got '%s'", CriticalThresholdName, spec.Name)
|
||||
}
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "schema v1 - overwrites existing thresholds and evaluation",
|
||||
initRule: PostableRule{},
|
||||
content: []byte(`{
|
||||
"alert": "OverwriteTest",
|
||||
"schemaVersion": "v1",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "cpu_usage", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}],
|
||||
"unit": "percent"
|
||||
},
|
||||
"target": 80.0,
|
||||
"targetUnit": "%",
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"thresholds": {
|
||||
"kind": "basic",
|
||||
"spec": [{
|
||||
"name": "existing_threshold",
|
||||
"target": 50.0,
|
||||
"targetUnit": "MB",
|
||||
"ruleUnit": "bytes",
|
||||
"matchType": "1",
|
||||
"op": "1"
|
||||
}]
|
||||
}
|
||||
},
|
||||
"evaluation": {
|
||||
"kind": "rolling",
|
||||
"spec": {
|
||||
"evalWindow": "10m",
|
||||
"frequency": "2m"
|
||||
}
|
||||
},
|
||||
"frequency":"7m",
|
||||
"evalWindow":"11m",
|
||||
"labels": {
|
||||
"severity": "critical"
|
||||
}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
validate: func(t *testing.T, rule *PostableRule) {
|
||||
if rule.RuleCondition.Thresholds == nil {
|
||||
t.Fatal("Expected Thresholds to be populated")
|
||||
}
|
||||
|
||||
specs, ok := rule.RuleCondition.Thresholds.Spec.(BasicRuleThresholds)
|
||||
if !ok {
|
||||
t.Fatalf("Expected BasicRuleThresholds, got %T", rule.RuleCondition.Thresholds.Spec)
|
||||
}
|
||||
|
||||
if len(specs) != 1 {
|
||||
t.Fatalf("Expected 1 threshold spec, got %d", len(specs))
|
||||
}
|
||||
|
||||
spec := specs[0]
|
||||
if spec.Name != "critical" {
|
||||
t.Errorf("Expected threshold name 'critical' (overwritten), got '%s'", spec.Name)
|
||||
}
|
||||
|
||||
if *spec.TargetValue != 80.0 {
|
||||
t.Errorf("Expected TargetValue 80.0 (overwritten), got %v", *spec.TargetValue)
|
||||
}
|
||||
if spec.TargetUnit != "%" {
|
||||
t.Errorf("Expected TargetUnit '%%' (overwritten), got '%s'", spec.TargetUnit)
|
||||
}
|
||||
|
||||
if rule.Evaluation == nil {
|
||||
t.Fatal("Expected Evaluation to be populated")
|
||||
}
|
||||
if window, ok := rule.Evaluation.Spec.(RollingWindow); ok {
|
||||
if !window.EvalWindow.Equal(rule.EvalWindow) {
|
||||
t.Errorf("Expected Evaluation EvalWindow to be overwritten to %v, got %v", rule.EvalWindow, window.EvalWindow)
|
||||
}
|
||||
if !window.Frequency.Equal(rule.Frequency) {
|
||||
t.Errorf("Expected Evaluation Frequency to be overwritten to %v, got %v", rule.Frequency, window.Frequency)
|
||||
}
|
||||
} else {
|
||||
t.Errorf("Expected RollingWindow spec, got %T", rule.Evaluation.Spec)
|
||||
}
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "schema v2alpha1 - uses explicit thresholds and evaluation",
|
||||
initRule: PostableRule{},
|
||||
@@ -645,19 +374,19 @@ func TestParseIntoRuleSchemaVersioning(t *testing.T) {
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 100.0,
|
||||
"matchType": "1",
|
||||
"op": "1"
|
||||
}
|
||||
"thresholds": {"kind": "basic", "spec": [{"name": "critical", "target": 100.0, "matchType": "1", "op": "1"}]}
|
||||
},
|
||||
"evaluation": {"kind": "rolling", "spec": {"evalWindow": "5m", "frequency": "1m"}},
|
||||
"notificationSettings": {"renotify": {"enabled": true, "interval": "4h", "alertStates": ["firing"]}}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectValidateError: true,
|
||||
},
|
||||
{
|
||||
name: "default schema version - defaults to v1 behavior",
|
||||
name: "missing schema version - rejected on write",
|
||||
initRule: PostableRule{},
|
||||
content: []byte(`{
|
||||
"alert": "DefaultSchemaTest",
|
||||
"alert": "MissingSchemaTest",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
@@ -672,24 +401,13 @@ func TestParseIntoRuleSchemaVersioning(t *testing.T) {
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 75.0,
|
||||
"matchType": "1",
|
||||
"op": "1"
|
||||
}
|
||||
"thresholds": {"kind": "basic", "spec": [{"name": "critical", "target": 75.0, "matchType": "1", "op": "1"}]}
|
||||
},
|
||||
"evaluation": {"kind": "rolling", "spec": {"evalWindow": "5m", "frequency": "1m"}},
|
||||
"notificationSettings": {"renotify": {"enabled": true, "interval": "4h", "alertStates": ["firing"]}}
|
||||
}`),
|
||||
kind: RuleDataKindJson,
|
||||
expectError: false,
|
||||
validate: func(t *testing.T, rule *PostableRule) {
|
||||
if rule.SchemaVersion != DefaultSchemaVersion {
|
||||
t.Errorf("Expected default schemaVersion '%s', got '%s'", DefaultSchemaVersion, rule.SchemaVersion)
|
||||
}
|
||||
if rule.RuleCondition.Thresholds == nil {
|
||||
t.Error("Expected Thresholds to be populated for default schema version")
|
||||
}
|
||||
if rule.Evaluation == nil {
|
||||
t.Error("Expected Evaluation to be populated for default schema version")
|
||||
}
|
||||
},
|
||||
kind: RuleDataKindJson,
|
||||
expectValidateError: true,
|
||||
},
|
||||
}
|
||||
|
||||
@@ -737,11 +455,7 @@ func TestParseIntoRuleThresholdGeneration(t *testing.T) {
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 100.0,
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"selectedQuery": "A",
|
||||
"targetUnit": "ms",
|
||||
"selectedQueryName": "A",
|
||||
"thresholds": {
|
||||
"kind": "basic",
|
||||
"spec": [
|
||||
@@ -907,428 +621,6 @@ func TestParseIntoRuleMultipleThresholds(t *testing.T) {
|
||||
assert.Equal(t, 1, len(vector))
|
||||
}
|
||||
|
||||
func TestAnomalyNegationEval(t *testing.T) {
|
||||
tests := []struct {
|
||||
name string
|
||||
ruleJSON []byte
|
||||
series *qbtypes.TimeSeries
|
||||
shouldAlert bool
|
||||
expectedValue float64
|
||||
}{
|
||||
{
|
||||
name: "anomaly rule with ValueIsBelow - should alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyBelowTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 2.0,
|
||||
"matchType": "1",
|
||||
"op": "2",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: -2.1}, // below & at least once, should alert
|
||||
{Timestamp: 2000, Value: -2.3},
|
||||
},
|
||||
},
|
||||
shouldAlert: true,
|
||||
expectedValue: -2.1,
|
||||
},
|
||||
{
|
||||
name: "anomaly rule with ValueIsBelow; should not alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyBelowTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 2.0,
|
||||
"matchType": "1",
|
||||
"op": "2",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`), // below & at least once, no value below -2.0
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: -1.9},
|
||||
{Timestamp: 2000, Value: -1.8},
|
||||
},
|
||||
},
|
||||
shouldAlert: false,
|
||||
},
|
||||
{
|
||||
name: "anomaly rule with ValueIsAbove; should alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyAboveTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 2.0,
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`), // above & at least once, should alert
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: 2.1}, // above 2.0, should alert
|
||||
{Timestamp: 2000, Value: 2.2},
|
||||
},
|
||||
},
|
||||
shouldAlert: true,
|
||||
expectedValue: 2.1,
|
||||
},
|
||||
{
|
||||
name: "anomaly rule with ValueIsAbove; should not alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyAboveTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 2.0,
|
||||
"matchType": "1",
|
||||
"op": "1",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: 1.1},
|
||||
{Timestamp: 2000, Value: 1.2},
|
||||
},
|
||||
},
|
||||
shouldAlert: false,
|
||||
},
|
||||
{
|
||||
name: "anomaly rule with ValueIsBelow and AllTheTimes; should alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyBelowAllTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 2.0,
|
||||
"matchType": "2",
|
||||
"op": "2",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`), // below and all the times
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: -2.1}, // all below -2
|
||||
{Timestamp: 2000, Value: -2.2},
|
||||
{Timestamp: 3000, Value: -2.5},
|
||||
},
|
||||
},
|
||||
shouldAlert: true,
|
||||
expectedValue: -2.1, // max value when all are below threshold
|
||||
},
|
||||
{
|
||||
name: "anomaly rule with ValueIsBelow and AllTheTimes; should not alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyBelowAllTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 2.0,
|
||||
"matchType": "2",
|
||||
"op": "2",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: -3.0},
|
||||
{Timestamp: 2000, Value: -1.0}, // above -2, breaks condition
|
||||
{Timestamp: 3000, Value: -2.5},
|
||||
},
|
||||
},
|
||||
shouldAlert: false,
|
||||
},
|
||||
{
|
||||
name: "anomaly rule with ValueOutsideBounds; should alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "AnomalyOutOfBoundsTest",
|
||||
"ruleType": "anomaly_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 7.0,
|
||||
"matchType": "1",
|
||||
"op": "7",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: -8.0}, // abs(−8) >= 7, alert
|
||||
{Timestamp: 2000, Value: 5.0},
|
||||
},
|
||||
},
|
||||
shouldAlert: true,
|
||||
expectedValue: -8.0,
|
||||
},
|
||||
{
|
||||
name: "non-anomaly threshold rule with ValueIsBelow; should alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "ThresholdTest",
|
||||
"ruleType": "threshold_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 90.0,
|
||||
"matchType": "1",
|
||||
"op": "2",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: 80.0}, // below 90, should alert
|
||||
{Timestamp: 2000, Value: 85.0},
|
||||
},
|
||||
},
|
||||
shouldAlert: true,
|
||||
expectedValue: 80.0,
|
||||
},
|
||||
{
|
||||
name: "non-anomaly rule with ValueIsBelow - should not alert",
|
||||
ruleJSON: []byte(`{
|
||||
"alert": "ThresholdTest",
|
||||
"ruleType": "threshold_rule",
|
||||
"version": "v5",
|
||||
"condition": {
|
||||
"compositeQuery": {
|
||||
"queryType": "builder",
|
||||
"queries": [{
|
||||
"type": "builder_query",
|
||||
"spec": {
|
||||
"name": "A",
|
||||
"signal": "metrics",
|
||||
"aggregations": [{"metricName": "test", "spaceAggregation": "p50"}],
|
||||
"stepInterval": "5m"
|
||||
}
|
||||
}]
|
||||
},
|
||||
"target": 50.0,
|
||||
"matchType": "1",
|
||||
"op": "2",
|
||||
"selectedQuery": "A"
|
||||
}
|
||||
}`),
|
||||
series: &qbtypes.TimeSeries{
|
||||
Labels: []*qbtypes.Label{
|
||||
{
|
||||
Key: telemetrytypes.TelemetryFieldKey{
|
||||
Name: "host",
|
||||
},
|
||||
Value: "server1",
|
||||
},
|
||||
},
|
||||
Values: []*qbtypes.TimeSeriesValue{
|
||||
{Timestamp: 1000, Value: 60.0}, // below, should alert
|
||||
{Timestamp: 2000, Value: 90.0},
|
||||
},
|
||||
},
|
||||
shouldAlert: false,
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
t.Run(tt.name, func(t *testing.T) {
|
||||
rule := PostableRule{}
|
||||
err := json.Unmarshal(tt.ruleJSON, &rule)
|
||||
if err != nil {
|
||||
t.Fatalf("Failed to unmarshal rule: %v", err)
|
||||
}
|
||||
|
||||
ruleThreshold, err := rule.RuleCondition.Thresholds.GetRuleThreshold()
|
||||
if err != nil {
|
||||
t.Fatalf("unexpected error from GetRuleThreshold: %v", err)
|
||||
}
|
||||
|
||||
resultVector, err := ruleThreshold.Eval(tt.series, "", EvalData{})
|
||||
if err != nil {
|
||||
t.Fatalf("unexpected error from Eval: %v", err)
|
||||
}
|
||||
|
||||
shouldAlert := len(resultVector) > 0
|
||||
|
||||
if shouldAlert != tt.shouldAlert {
|
||||
t.Errorf("Expected shouldAlert=%v, got %v. %s",
|
||||
tt.shouldAlert, shouldAlert, tt.name)
|
||||
}
|
||||
|
||||
if tt.shouldAlert && len(resultVector) > 0 {
|
||||
sample := resultVector[0]
|
||||
if sample.V != tt.expectedValue {
|
||||
t.Errorf("Expected alert value=%.2f, got %.2f. %s",
|
||||
tt.expectedValue, sample.V, tt.name)
|
||||
}
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestVersionDefaultsToV5(t *testing.T) {
|
||||
content := `{
|
||||
"alert": "cpu high",
|
||||
|
||||
@@ -147,7 +147,7 @@ func TestV2MinimalReadShape(t *testing.T) {
|
||||
|
||||
body := readBody(t, stored)
|
||||
|
||||
for _, field := range []string{"labels", "annotations", "description", "preferredChannels", "evalWindow", "frequency"} {
|
||||
for _, field := range []string{"labels", "annotations", "description"} {
|
||||
assert.NotContains(t, body, field)
|
||||
}
|
||||
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
19
tests/fixtures/alerts.py
vendored
19
tests/fixtures/alerts.py
vendored
@@ -32,12 +32,12 @@ def create_alert_rule(signoz: types.SigNoz, get_token: Callable[[str, str], str]
|
||||
|
||||
def _create_alert_rule(rule_data: dict) -> str:
|
||||
response = requests.post(
|
||||
signoz.self.host_configs["8080"].get("/api/v1/rules"),
|
||||
signoz.self.host_configs["8080"].get("/api/v2/rules"),
|
||||
json=rule_data,
|
||||
headers={"Authorization": f"Bearer {admin_token}"},
|
||||
timeout=5,
|
||||
)
|
||||
assert response.status_code == HTTPStatus.OK, f"Failed to create rule, api returned {response.status_code} with response: {response.text}"
|
||||
assert response.status_code == HTTPStatus.CREATED, f"Failed to create rule, api returned {response.status_code} with response: {response.text}"
|
||||
rule_id = response.json()["data"]["id"]
|
||||
rule_ids.append(rule_id)
|
||||
return rule_id
|
||||
@@ -45,11 +45,11 @@ def create_alert_rule(signoz: types.SigNoz, get_token: Callable[[str, str], str]
|
||||
def _delete_alert_rule(rule_id: str):
|
||||
logger.info("Deleting rule: %s", {"rule_id": rule_id})
|
||||
response = requests.delete(
|
||||
signoz.self.host_configs["8080"].get(f"/api/v1/rules/{rule_id}"),
|
||||
signoz.self.host_configs["8080"].get(f"/api/v2/rules/{rule_id}"),
|
||||
headers={"Authorization": f"Bearer {admin_token}"},
|
||||
timeout=5,
|
||||
)
|
||||
if response.status_code != HTTPStatus.OK:
|
||||
if response.status_code != HTTPStatus.NO_CONTENT:
|
||||
raise Exception( # pylint: disable=broad-exception-raised
|
||||
f"Failed to delete rule, api returned {response.status_code} with response: {response.text}"
|
||||
)
|
||||
@@ -89,19 +89,22 @@ def create_alert_rule_with_channel(
|
||||
|
||||
|
||||
def delete_all_rules(signoz: types.SigNoz, token: str) -> None:
|
||||
# the v3 list is paginated; 200 is its maximum page and more rules than
|
||||
# that would mean a seeding bug, not a second page
|
||||
response = requests.get(
|
||||
signoz.self.host_configs["8080"].get("/api/v2/rules"),
|
||||
signoz.self.host_configs["8080"].get("/api/v3/rules"),
|
||||
params={"limit": 200},
|
||||
headers={"Authorization": f"Bearer {token}"},
|
||||
timeout=5,
|
||||
)
|
||||
assert response.status_code == HTTPStatus.OK
|
||||
for rule in response.json()["data"]:
|
||||
for rule in response.json()["data"]["rules"]:
|
||||
delete_response = requests.delete(
|
||||
signoz.self.host_configs["8080"].get(f"/api/v1/rules/{rule['id']}"),
|
||||
signoz.self.host_configs["8080"].get(f"/api/v2/rules/{rule['id']}"),
|
||||
headers={"Authorization": f"Bearer {token}"},
|
||||
timeout=5,
|
||||
)
|
||||
assert delete_response.status_code == HTTPStatus.OK, f"failed to delete rule {rule['id']}: {delete_response.text}"
|
||||
assert delete_response.status_code == HTTPStatus.NO_CONTENT, f"failed to delete rule {rule['id']}: {delete_response.text}"
|
||||
|
||||
|
||||
@pytest.fixture(name="seed_alert_rules", scope="function")
|
||||
|
||||
@@ -573,24 +573,3 @@ def test_error_contract(
|
||||
error = response.json()["error"]
|
||||
assert error["code"] == expected_code, f"params {params!r}"
|
||||
assert expected_message_part in error["message"], f"params {params!r}: {error['message']}"
|
||||
|
||||
|
||||
def test_v2_list_still_serves_bare_array(
|
||||
signoz: SigNoz,
|
||||
create_user_admin: Operation, # pylint: disable=unused-argument
|
||||
get_token: Callable[[str, str], str],
|
||||
seed_alert_rules: Callable[[str, list[dict]], None],
|
||||
):
|
||||
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
|
||||
seed_alert_rules(SEED_CHANNEL_NAME, SEED_RULES)
|
||||
|
||||
response = requests.get(
|
||||
signoz.self.host_configs["8080"].get("/api/v2/rules"),
|
||||
headers={"Authorization": f"Bearer {token}"},
|
||||
timeout=5,
|
||||
)
|
||||
|
||||
assert response.status_code == HTTPStatus.OK
|
||||
data = response.json()["data"]
|
||||
assert isinstance(data, list), "deprecated v2 must keep returning a bare array"
|
||||
assert {rule["alert"] for rule in data} == {r["alert"] for r in SEED_RULES}
|
||||
|
||||
Reference in New Issue
Block a user