Alerting: Prom writer to handle err-mimir-max-series-per-user as user error (#98334)

handle err-mimir-max-series-per-user as user error

pkg/services/ngalert/writer/prom.go

@@ -11,10 +11,11 @@ import (
 
 	"github.com/benbjohnson/clock"
 	"github.com/grafana/dataplane/sdata/numeric"
+	"github.com/m3db/prometheus_remote_client_golang/promremote"
+
 	"github.com/grafana/grafana/pkg/infra/log"
 	"github.com/grafana/grafana/pkg/services/ngalert/metrics"
 	"github.com/grafana/grafana/pkg/setting"
-	"github.com/m3db/prometheus_remote_client_golang/promremote"
 
 	"github.com/grafana/grafana-plugin-sdk-go/backend/httpclient"
 	"github.com/grafana/grafana-plugin-sdk-go/data"
@@ -26,6 +27,7 @@ const (
 	// Fixed error messages
 	MimirDuplicateTimestampError = "err-mimir-sample-duplicate-timestamp"
 	MimirInvalidLabelError       = "err-mimir-label-invalid"
+	MimirMaxSeriesPerUserError   = "err-mimir-max-series-per-user"
 
 	// Best effort error messages
 	PrometheusDuplicateTimestampError = "duplicate sample for timestamp"
@@ -268,6 +270,11 @@ func checkWriteError(writeErr promremote.WriteError) (err error, ignored bool) {
 			return errors.Join(ErrRejectedWrite, writeErr), false
 		}
 
+		// this can happen when user exceeded defined maximum of
+		if strings.Contains(msg, MimirMaxSeriesPerUserError) {
+			return errors.Join(ErrRejectedWrite, writeErr), false
+		}
+
 		// For now, all 400s that are not previously known are considered unexpected.
 		// TODO: Consider blanket-converting all 400s to be known errors. This should only be done once we are confident this is not a problem with this client.
 		return errors.Join(ErrUnexpectedWriteFailure, writeErr), false

pkg/services/ngalert/writer/prom_test.go

@@ -12,13 +12,14 @@ import (
 
 	"github.com/benbjohnson/clock"
 	"github.com/grafana/grafana-plugin-sdk-go/data"
-	"github.com/grafana/grafana/pkg/infra/log"
-	"github.com/grafana/grafana/pkg/setting"
 	"github.com/m3db/prometheus_remote_client_golang/promremote"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/prometheus/prompb"
 	"github.com/stretchr/testify/require"
 
+	"github.com/grafana/grafana/pkg/infra/log"
+	"github.com/grafana/grafana/pkg/setting"
+
 	"github.com/grafana/grafana/pkg/services/ngalert/metrics"
 	ngmodels "github.com/grafana/grafana/pkg/services/ngalert/models"
 )
@@ -221,6 +222,22 @@ func TestPrometheusWriter_Write(t *testing.T) {
 		require.Error(t, err)
 		require.ErrorIs(t, err, ErrRejectedWrite)
 	})
+
+	t.Run("max series limit fit under the client error category ", func(t *testing.T) {
+		msg := "send data to ingesters: failed pushing to ingester ingester-1: user=1: per-user series limit of 10 exceeded (err-mimir-max-series-per-user). To adjust the related per-tenant limit, configure -ingester.max-global-series-per-user, or contact your service administrator."
+		clientErr := testClientWriteError{
+			statusCode: http.StatusBadRequest,
+			msg:        &msg,
+		}
+		client.writeSeriesFunc = func(ctx context.Context, ts promremote.TSList, opts promremote.WriteOptions) (promremote.WriteResult, promremote.WriteError) {
+			return promremote.WriteResult{}, clientErr
+		}
+
+		err := writer.Write(ctx, "test", now, frames, 1, map[string]string{"extra": "label"})
+
+		require.Error(t, err)
+		require.ErrorIs(t, err, ErrRejectedWrite)
+	})
 }
 
 func extractValue(t *testing.T, frames data.Frames, labels map[string]string, frameType data.FrameType) float64 {