Alerting: Add rule_matcher filter to Prometheus rules API (#115297)

**What is this feature?**

Add `rule_matcher` filter to the Prometheus-compatible list rules API: `/api/prometheus/grafana/api/v1/rules`. It allows to filter rules by static labels (not by alert instance labels).

**Special notes:**
  - Equality (`=`) and inequality (`!=`) matchers are pushed down to the database. Regex matchers (`=~`, `!~`) are applied in-memory at the API layer.
  - SQLite: Uses GLOB pattern matching
  - MySQL / PostgreSQL: Use JSON functions to compare label values


---------

Co-authored-by: Konrad Lalik <konradlalik@gmail.com>
This commit is contained in:
Alexander Akhmetov
2025-12-16 14:13:50 +01:00
committed by GitHub
co-authored by Konrad Lalik
parent c7c1dd4ead
commit c0295d06a3
18 changed files with 1137 additions and 71 deletions
@@ -33,6 +33,12 @@ import (
"go.opentelemetry.io/otel/trace"
)
const (
queryIncludeInternalLabels = "includeInternalLabels"
queryRuleMatcher = "rule_matcher"
queryInstanceMatcher = "matcher"
)
type RuleStoreReader interface {
GetUserVisibleNamespaces(context.Context, int64, identity.Requester) (map[string]*folder.Folder, error)
ListAlertRulesStoreV2
@@ -62,6 +68,20 @@ type PrometheusSrv struct {
// Package-level OpenTelemetry tracer per Grafana instrumentation conventions.
var tracer = otel.Tracer("github.com/grafana/grafana/pkg/services/ngalert/api/prometheus")
// badRequestError returns a Prometheus-compatible error response for bad request data.
func badRequestError(err error) apimodels.RuleResponse {
return apimodels.RuleResponse{
DiscoveryBase: apimodels.DiscoveryBase{
Status: "error",
Error: err.Error(),
ErrorType: apiv1.ErrBadData,
},
Data: apimodels.RuleDiscovery{
RuleGroups: []apimodels.RuleGroup{},
},
}
}
func NewPrometheusSrv(log log.Logger, manager state.AlertInstanceManager, status StatusReader, store RuleStoreReader, authz RuleGroupAccessControlService, provenanceStore ProvenanceStore) *PrometheusSrv {
return &PrometheusSrv{
log,
@@ -73,8 +93,6 @@ func NewPrometheusSrv(log log.Logger, manager state.AlertInstanceManager, status
}
}
const queryIncludeInternalLabels = "includeInternalLabels"
func getBoolWithDefault(vals url.Values, field string, d bool) bool {
f := vals.Get(field)
if f == "" {
@@ -188,15 +206,15 @@ func getPanelIDFromQuery(v url.Values) (int64, error) {
return 0, nil
}
func getMatchersFromQuery(v url.Values) (labels.Matchers, error) {
func getMatchersFromQuery(v url.Values, paramName string) (labels.Matchers, error) {
var matchers labels.Matchers
for _, s := range v["matcher"] {
for _, s := range v[paramName] {
var m labels.Matcher
if err := json.Unmarshal([]byte(s), &m); err != nil {
return nil, err
}
if len(m.Name) == 0 {
return nil, errors.New("bad matcher: the name cannot be blank")
return nil, fmt.Errorf("bad %s: the name cannot be blank", paramName)
}
matchers = append(matchers, &m)
}
@@ -454,6 +472,7 @@ type paginationContext struct {
stateFilterSet map[eval.State]struct{}
healthFilterSet map[string]struct{}
matchers labels.Matchers
ruleLabelMatchers labels.Matchers
labelOptions []ngmodels.LabelOption
limitAlertsPerRule int64
limitRulesPerGroup int64
@@ -476,6 +495,9 @@ func accumulateTotals(dest, source map[string]int64) {
// fetchAndFilterPage fetches one page from the store and applies filters
func (ctx *paginationContext) fetchAndFilterPage(log log.Logger, store ListAlertRulesStoreV2, span trace.Span, token string, remainingGroups, remainingRules int64) (pageResult, error) {
// Split matchers: only equality/inequality are supported by the store
storeMatchers := filterOutRegexMatchers(ctx.ruleLabelMatchers)
byGroupQuery := ngmodels.ListAlertRulesExtendedQuery{
ListAlertRulesQuery: ngmodels.ListAlertRulesQuery{
OrgID: ctx.opts.OrgID,
@@ -488,6 +510,7 @@ func (ctx *paginationContext) fetchAndFilterPage(log log.Logger, store ListAlert
DataSourceUIDs: ctx.dataSourceUIDs,
SearchTitle: ctx.title,
SearchRuleGroup: ctx.searchRuleGroup,
LabelMatchers: storeMatchers,
},
RuleType: ctx.ruleType,
Limit: remainingGroups,
@@ -534,6 +557,8 @@ func (ctx *paginationContext) fetchAndFilterPage(log log.Logger, store ListAlert
filterRulesByHealth(ruleGroup, ctx.healthFilterSet)
}
filterRulesByLabelMatchers(ruleGroup, ctx.ruleLabelMatchers)
if ctx.limitRulesPerGroup > -1 && int64(len(ruleGroup.Rules)) > ctx.limitRulesPerGroup {
ruleGroup.Rules = ruleGroup.Rules[0:ctx.limitRulesPerGroup]
}
@@ -546,6 +571,17 @@ func (ctx *paginationContext) fetchAndFilterPage(log log.Logger, store ListAlert
return result, nil
}
func filterOutRegexMatchers(matchers labels.Matchers) labels.Matchers {
var result labels.Matchers
for _, m := range matchers {
if m.Type == labels.MatchEqual || m.Type == labels.MatchNotEqual {
result = append(result, m)
}
}
return result
}
// paginateRuleGroups fetches pages until limits are satisfied applying filters at each step
func paginateRuleGroups(log log.Logger, store ListAlertRulesStoreV2, ctx *paginationContext, span trace.Span, maxGroups, maxRules int64, startToken string) ([]apimodels.RuleGroup, map[string]int64, string, error) {
allGroups := []apimodels.RuleGroup{}
@@ -644,21 +680,30 @@ func PrepareRuleGroupStatusesV2(log log.Logger, store ListAlertRulesStoreV2, opt
attribute.Int64("limit_rules", limitRulesPerGroup),
attribute.Int64("limit_alerts", limitAlertsPerRule),
)
matchers, err := getMatchersFromQuery(opts.Query)
matchers, err := getMatchersFromQuery(opts.Query, queryInstanceMatcher)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = err.Error()
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(err)
}
span.SetAttributes(attribute.Int("matcher_count", len(matchers)))
ruleLabelMatchers, err := getMatchersFromQuery(opts.Query, queryRuleMatcher)
if err != nil {
return badRequestError(err)
}
regexCount := 0
for _, m := range ruleLabelMatchers {
if m.Type == labels.MatchRegexp || m.Type == labels.MatchNotRegexp {
regexCount++
}
}
span.SetAttributes(
attribute.Int("rule_matcher_count", len(ruleLabelMatchers)),
attribute.Int("rule_matcher_regex_count", regexCount),
)
stateFilterSet, err := GetStatesFromQuery(opts.Query)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = err.Error()
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(err)
}
span.SetAttributes(
attribute.Int("state_filter_count", len(stateFilterSet)),
@@ -667,10 +712,7 @@ func PrepareRuleGroupStatusesV2(log log.Logger, store ListAlertRulesStoreV2, opt
healthFilterSet, err := GetHealthFromQuery(opts.Query)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = err.Error()
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(err)
}
span.SetAttributes(
attribute.Int("health_filter_count", len(healthFilterSet)),
@@ -808,6 +850,7 @@ func PrepareRuleGroupStatusesV2(log log.Logger, store ListAlertRulesStoreV2, opt
stateFilterSet: stateFilterSet,
healthFilterSet: healthFilterSet,
matchers: matchers,
ruleLabelMatchers: ruleLabelMatchers,
labelOptions: labelOptions,
limitAlertsPerRule: limitAlertsPerRule,
limitRulesPerGroup: limitRulesPerGroup,
@@ -833,6 +876,7 @@ func PrepareRuleGroupStatusesV2(log log.Logger, store ListAlertRulesStoreV2, opt
return ruleResponse
}
// nolint:gocyclo
func PrepareRuleGroupStatuses(log log.Logger, store ListAlertRulesStore, opts RuleGroupStatusesOptions, ruleStatusMutator RuleStatusMutator, alertStateMutator RuleAlertStateMutator, provenanceRecords map[string]ngmodels.Provenance) apimodels.RuleResponse {
ruleResponse := apimodels.RuleResponse{
DiscoveryBase: apimodels.DiscoveryBase{
@@ -846,41 +890,30 @@ func PrepareRuleGroupStatuses(log log.Logger, store ListAlertRulesStore, opts Ru
dashboardUID := opts.Query.Get("dashboard_uid")
panelID, err := getPanelIDFromQuery(opts.Query)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = fmt.Sprintf("invalid panel_id: %s", err.Error())
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(fmt.Errorf("invalid panel_id: %w", err))
}
if dashboardUID == "" && panelID != 0 {
ruleResponse.Status = "error"
ruleResponse.Error = "panel_id must be set with dashboard_uid"
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(errors.New("panel_id must be set with dashboard_uid"))
}
limitRulesPerGroup := getInt64WithDefault(opts.Query, "limit_rules", -1)
limitAlertsPerRule := getInt64WithDefault(opts.Query, "limit_alerts", -1)
matchers, err := getMatchersFromQuery(opts.Query)
matchers, err := getMatchersFromQuery(opts.Query, queryInstanceMatcher)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = err.Error()
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(err)
}
ruleLabelMatchers, err := getMatchersFromQuery(opts.Query, queryRuleMatcher)
if err != nil {
return badRequestError(err)
}
stateFilterSet, err := GetStatesFromQuery(opts.Query)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = err.Error()
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(err)
}
healthFilterSet, err := GetHealthFromQuery(opts.Query)
if err != nil {
ruleResponse.Status = "error"
ruleResponse.Error = err.Error()
ruleResponse.ErrorType = apiv1.ErrBadData
return ruleResponse
return badRequestError(err)
}
var labelOptions []ngmodels.LabelOption
@@ -913,6 +946,9 @@ func PrepareRuleGroupStatuses(log log.Logger, store ListAlertRulesStore, opts Ru
dataSourceUIDs := opts.Query["datasource_uid"]
searchRuleGroup := opts.Query.Get("search.rule_group")
// Split matchers: only equality/inequality are supported by the store
storeMatchers := filterOutRegexMatchers(ruleLabelMatchers)
alertRuleQuery := ngmodels.ListAlertRulesQuery{
OrgID: opts.OrgID,
NamespaceUIDs: namespaceUIDs,
@@ -924,6 +960,7 @@ func PrepareRuleGroupStatuses(log log.Logger, store ListAlertRulesStore, opts Ru
SearchTitle: title,
SearchRuleGroup: searchRuleGroup,
DataSourceUIDs: dataSourceUIDs,
LabelMatchers: storeMatchers,
}
ruleList, err := store.ListAlertRules(opts.Ctx, &alertRuleQuery)
if err != nil {
@@ -978,6 +1015,10 @@ func PrepareRuleGroupStatuses(log log.Logger, store ListAlertRulesStore, opts Ru
filterRulesByHealth(ruleGroup, healthFilterSet)
}
if len(ruleLabelMatchers) > 0 {
filterRulesByLabelMatchers(ruleGroup, ruleLabelMatchers)
}
if limitRulesPerGroup > -1 && int64(len(ruleGroup.Rules)) > limitRulesPerGroup {
ruleGroup.Rules = ruleGroup.Rules[0:limitRulesPerGroup]
}
@@ -1105,6 +1146,30 @@ func filterRulesByHealth(ruleGroup *apimodels.RuleGroup, withHealthFast map[stri
ruleGroup.Rules = filteredRules
}
func filterRulesByLabelMatchers(ruleGroup *apimodels.RuleGroup, matchers labels.Matchers) {
if len(matchers) == 0 {
return
}
filteredRules := make([]apimodels.AlertingRule, 0, len(ruleGroup.Rules))
for _, rule := range ruleGroup.Rules {
ruleLabels := rule.Labels.Map()
matches := true
for _, m := range matchers {
if !m.Matches(ruleLabels[m.Name]) {
matches = false
break
}
}
if matches {
filteredRules = append(filteredRules, rule)
}
}
ruleGroup.Rules = filteredRules
}
// This is the same as matchers.Matches but avoids the need to create a LabelSet
func matchersMatch(matchers []*labels.Matcher, labels map[string]string) bool {
for _, m := range matchers {