Helm: Add support for dedicated ruler query path (grafana#7964)

* Add support for dedicated ruler query path * Add feature to changelog * Default to false * Add ci test for ruler dedicated query path * rename test file + build tests * update comments * Update operations/helm/charts/mimir-distributed/CHANGELOG.md Co-authored-by: Dimitar Dimitrov <[email protected]> --------- Co-authored-by: Dimitar Dimitrov <[email protected]>
narqo · Jun 6, 2024 · 58d9ca1 · 58d9ca1
1 parent 626989a
commit 58d9ca1
Show file tree

Hide file tree

Showing 90 changed files with 5,485 additions and 0 deletions.
diff --git a/operations/helm/charts/mimir-distributed/CHANGELOG.md b/operations/helm/charts/mimir-distributed/CHANGELOG.md
@@ -29,6 +29,7 @@ Entries should include a reference to the Pull Request that introduced the chang
 
 ## main / unreleased
 
+* [FEATURE] Add support for a dedicated query path for the ruler. This allows for the isolation of ruler and user query paths. Enable it via `ruler.remoteEvaluationDedicatedQueryPath: true`. #7964
 * [CHANGE] Fine-tuned `terminationGracePeriodSeconds` for the following components: #7361 #7364
   * Alertmanager: changed from `60` to `900`
   * Distributor: changed from `60` to `100`

diff --git a/...ions/helm/charts/mimir-distributed/ci/offline/test-ruler-dedicated-query-path-values.yaml b/...ions/helm/charts/mimir-distributed/ci/offline/test-ruler-dedicated-query-path-values.yaml
@@ -0,0 +1,5 @@
+# Pin kube version so results are the same for running in CI and locally where the installed kube version may be different.
+kubeVersionOverride: "1.20"
+
+ruler:
+  remoteEvaluationDedicatedQueryPath: true
diff --git a/operations/helm/charts/mimir-distributed/templates/_helpers.tpl b/operations/helm/charts/mimir-distributed/templates/_helpers.tpl
@@ -446,6 +446,9 @@ Examples:
   "query-scheduler" "query_scheduler"
   "results-cache" "results-cache"
   "ruler" "ruler"
+  "ruler-querier" "ruler_querier"
+  "ruler-query-frontend" "ruler_query_frontend"
+  "ruler-query-scheduler" "ruler_query_scheduler"
   "smoke-test" "smoke_test"
   "store-gateway" "store_gateway"
   "tokengen" "tokengenJob"

diff --git a/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-dep.yaml b/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-dep.yaml
@@ -0,0 +1,171 @@
+{{- if .Values.ruler.remoteEvaluationDedicatedQueryPath }}
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ include "mimir.resourceName" (dict "ctx" . "component" "ruler-querier") }}
+  labels:
+    {{- include "mimir.labels" (dict "ctx" . "component" "ruler-querier" "memberlist" true) | nindent 4 }}
+  annotations:
+    {{- toYaml .Values.ruler_querier.annotations | nindent 4 }}
+  namespace: {{ .Release.Namespace | quote }}
+spec:
+  {{- if or (not .Values.ruler_querier.kedaAutoscaling.enabled) (.Values.ruler_querier.kedaAutoscaling.preserveReplicas) }}
+  # If replicas is not number (when using values file it's float64, when using --set arg it's int64) and is false (i.e. null) don't set it
+  {{- if or (or (kindIs "int64" .Values.ruler_querier.replicas) (kindIs "float64" .Values.ruler_querier.replicas)) (.Values.ruler_querier.replicas) }}
+  replicas: {{ .Values.ruler_querier.replicas }}
+  {{- end }}
+  {{- end }}
+  selector:
+    matchLabels:
+      {{- include "mimir.selectorLabels" (dict "ctx" . "component" "ruler-querier" "memberlist" true) | nindent 6 }}
+  strategy:
+    {{- toYaml .Values.ruler_querier.strategy | nindent 4 }}
+  template:
+    metadata:
+      labels:
+        {{- include "mimir.podLabels" (dict "ctx" . "component" "ruler-querier" "memberlist" true) | nindent 8 }}
+      annotations:
+        {{- include "mimir.podAnnotations" (dict "ctx" . "component" "ruler-querier") | nindent 8 }}
+    spec:
+      serviceAccountName: {{ template "mimir.serviceAccountName" . }}
+      {{- if .Values.ruler_querier.priorityClassName }}
+      priorityClassName: {{ .Values.ruler_querier.priorityClassName }}
+      {{- end }}
+      securityContext:
+        {{- include "mimir.lib.podSecurityContext" (dict "ctx" . "component" "ruler-querier") | nindent 8 }}
+      {{- with .Values.ruler_querier.initContainers }}
+      initContainers:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- if .Values.image.pullSecrets }}
+      imagePullSecrets:
+      {{- range .Values.image.pullSecrets }}
+        - name: {{ . }}
+      {{- end}}
+      {{- end }}
+      containers:
+        - name: ruler-querier
+          image: "{{ include "mimir.imageReference" . }}"
+          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          args:
+            - "-target=querier"
+            - "-config.expand-env=true"
+            - "-config.file=/etc/mimir/mimir.yaml"
+            - "-querier.scheduler-address={{ template "mimir.fullname" . }}-ruler-query-scheduler-headless.{{ .Release.Namespace }}.svc:{{ include "mimir.serverGrpcListenPort" .  }}"
+          {{- if .Values.ingester.zoneAwareReplication.migration.enabled }}
+            {{- if not .Values.ingester.zoneAwareReplication.migration.readPath }}
+            - "-ingester.ring.zone-awareness-enabled=false"
+            {{- end }}
+          {{- end }}
+          {{- if .Values.store_gateway.zoneAwareReplication.migration.enabled }}
+            {{- if not .Values.store_gateway.zoneAwareReplication.migration.readPath }}
+            - "-store-gateway.sharding-ring.prefix=collectors/"
+            - "-store-gateway.sharding-ring.zone-awareness-enabled=false"
+            {{- end }}
+          {{- end }}
+          {{- range $key, $value := .Values.ruler_querier.extraArgs }}
+            - "-{{ $key }}={{ $value }}"
+          {{- end }}
+          volumeMounts:
+            {{- if .Values.ruler_querier.extraVolumeMounts }}
+              {{ toYaml .Values.ruler_querier.extraVolumeMounts | nindent 12}}
+            {{- end }}
+            {{- if .Values.global.extraVolumeMounts }}
+              {{ toYaml .Values.global.extraVolumeMounts | nindent 12}}
+            {{- end }}
+            - name: config
+              mountPath: /etc/mimir
+            {{- if .Values.enterprise.enabled }}
+            - name: license
+              mountPath: /license
+            {{- end }}
+            - name: runtime-config
+              mountPath: /var/{{ include "mimir.name" . }}
+            - name: storage
+              mountPath: "/data"
+              subPath: {{ .Values.ruler_querier.persistence.subPath }}
+            - name: active-queries
+              mountPath: /active-query-tracker
+          ports:
+            - name: http-metrics
+              containerPort: {{ include "mimir.serverHttpListenPort" . }}
+              protocol: TCP
+            - name: grpc
+              containerPort: {{ include "mimir.serverGrpcListenPort" . }}
+              protocol: TCP
+            - name: memberlist
+              containerPort: {{ include "mimir.memberlistBindPort" . }}
+              protocol: TCP
+          livenessProbe:
+            {{- toYaml .Values.ruler_querier.livenessProbe | nindent 12 }}
+          readinessProbe:
+            {{- toYaml .Values.ruler_querier.readinessProbe | nindent 12 }}
+          resources:
+            {{- toYaml .Values.ruler_querier.resources | nindent 12 }}
+          securityContext:
+            {{- toYaml .Values.ruler_querier.containerSecurityContext | nindent 12 }}
+          env:
+            {{- with .Values.global.extraEnv }}
+              {{- toYaml . | nindent 12 }}
+            {{- end }}
+            {{- with .Values.ruler_querier.env }}
+              {{- toYaml . | nindent 12 }}
+            {{- end }}
+            {{- $cpu_request := dig "requests" "cpu" nil .Values.ruler_querier.resources }}
+            {{- if $cpu_request }}
+              {{- $cpu_request_doubled := include "mimir.parseCPU" (dict "value" $cpu_request) | float64 | mulf 2 | ceil }}
+              {{- $cpu_request_plus_four := include "mimir.parseCPU" (dict "value" $cpu_request) | float64 | addf 4 | ceil }}
+            - name: "GOMAXPROCS"
+              value: {{ max $cpu_request_doubled $cpu_request_plus_four | toString | toYaml }}
+            {{- end }}
+            {{- $jaeger_queue_size := dig "jaegerReporterMaxQueueSize" nil .Values.querier }}
+            {{- if $jaeger_queue_size }}
+            - name: "JAEGER_REPORTER_MAX_QUEUE_SIZE"
+              value: {{$jaeger_queue_size | toString | toYaml }}
+            {{- end }}
+          envFrom:
+            {{- with .Values.global.extraEnvFrom }}
+              {{- toYaml . | nindent 12 }}
+            {{- end }}
+            {{- with .Values.ruler_querier.extraEnvFrom }}
+              {{- toYaml . | nindent 12 }}
+            {{- end }}
+{{- if .Values.ruler_querier.extraContainers }}
+{{ toYaml .Values.ruler_querier.extraContainers | indent 8}}
+{{- end }}
+      {{- with .Values.ruler_querier.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.ruler_querier.affinity}}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- include "mimir.lib.topologySpreadConstraints" (dict "ctx" . "component" "ruler-querier") | nindent 6 }}
+      {{- with .Values.ruler_querier.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      terminationGracePeriodSeconds: {{ .Values.ruler_querier.terminationGracePeriodSeconds }}
+      volumes:
+        - name: config
+          {{- include "mimir.configVolume" . | nindent 10 }}
+        {{- if .Values.enterprise.enabled }}
+        - name: license
+          secret:
+            secretName: {{ tpl .Values.license.secretName . }}
+        {{- end }}
+        - name: runtime-config
+          configMap:
+            name: {{ template "mimir.fullname" . }}-runtime
+        {{- if .Values.ruler_querier.extraVolumes }}
+        {{ toYaml .Values.ruler_querier.extraVolumes | nindent 8}}
+        {{- end }}
+        {{- if .Values.global.extraVolumes }}
+        {{ toYaml .Values.global.extraVolumes | nindent 8}}
+        {{- end }}
+        - name: storage
+          emptyDir: {}
+        - name: active-queries
+          emptyDir: {}
+{{- end }}
diff --git a/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-pdb.yaml b/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-pdb.yaml
@@ -0,0 +1,3 @@
+{{- if .Values.ruler.remoteEvaluationDedicatedQueryPath }}
+{{- include "mimir.lib.podDisruptionBudget" (dict "ctx" $ "component" "ruler-querier" "memberlist" true) }}
+{{- end }}
diff --git a/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-servmon.yaml b/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-servmon.yaml
@@ -0,0 +1,3 @@
+{{- if .Values.ruler.remoteEvaluationDedicatedQueryPath }}
+{{- include "mimir.lib.serviceMonitor" (dict "ctx" $ "component" "ruler-querier" "memberlist" true) }}
+{{- end }}
diff --git a/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-so.yaml b/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-so.yaml
@@ -0,0 +1,46 @@
+{{- if .Values.ruler.remoteEvaluationDedicatedQueryPath }}
+{{- if .Values.ruler_querier.kedaAutoscaling.enabled }}
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "mimir.resourceName" (dict "ctx" . "component" "ruler-querier") }}
+  labels:
+    {{- include "mimir.labels" (dict "ctx" . "component" "ruler-querier") | nindent 4 }}
+  annotations:
+    {{- toYaml .Values.ruler_querier.annotations | nindent 4 }}
+  namespace: {{ .Release.Namespace | quote }}
+spec:
+  advanced:
+    horizontalPodAutoscalerConfig:
+      {{- with .Values.ruler_querier.kedaAutoscaling.behavior }}
+      behavior:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+  maxReplicaCount: {{ .Values.ruler_querier.kedaAutoscaling.maxReplicaCount }}
+  minReplicaCount: {{ .Values.ruler_querier.kedaAutoscaling.minReplicaCount }}
+  pollingInterval: {{ .Values.kedaAutoscaling.pollingInterval }}
+  scaleTargetRef:
+    name: {{ include "mimir.resourceName" (dict "ctx" . "component" "ruler-querier") }}
+    apiVersion: apps/v1
+    kind: Deployment
+  triggers:
+  - metadata:
+      query: sum(max_over_time(cortex_query_scheduler_inflight_requests{container="ruler-query-scheduler",namespace="{{ .Release.Namespace }}",quantile="0.5"}[1m]))
+      serverAddress: {{ include "mimir.kedaPrometheusAddress" (dict "ctx" $) }} 
+      threshold: {{ .Values.ruler_querier.kedaAutoscaling.querySchedulerInflightRequestsThreshold | quote }}
+      {{- if .Values.kedaAutoscaling.customHeaders }}
+      customHeaders: {{ (include "mimir.lib.mapToCSVString" (dict "map" .Values.kedaAutoscaling.customHeaders)) | quote }}
+      {{- end }}
+    name: cortex_querier_hpa_default
+    type: prometheus
+  - metadata:
+      query: sum(rate(cortex_querier_request_duration_seconds_sum{container="ruler-querier",namespace="{{ .Release.Namespace }}"}[1m]))
+      serverAddress: {{ include "mimir.kedaPrometheusAddress" (dict "ctx" $) }} 
+      threshold: {{ .Values.ruler_querier.kedaAutoscaling.querySchedulerInflightRequestsThreshold | quote }}
+      {{- if .Values.kedaAutoscaling.customHeaders }}
+      customHeaders: {{ (include "mimir.lib.mapToCSVString" (dict "map" .Values.kedaAutoscaling.customHeaders)) | quote }}
+      {{- end }}
+    name: cortex_querier_hpa_default_requests_duration
+    type: prometheus
+{{- end }}
+{{- end }}
diff --git a/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-svc.yaml b/operations/helm/charts/mimir-distributed/templates/ruler-querier/ruler-querier-svc.yaml
@@ -0,0 +1,27 @@
+{{- if .Values.ruler.remoteEvaluationDedicatedQueryPath }}
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "mimir.resourceName" (dict "ctx" . "component" "ruler-querier") }}
+  labels:
+    {{- include "mimir.labels" (dict "ctx" . "component" "ruler-querier" "memberlist" true) | nindent 4 }}
+    {{- with .Values.ruler_querier.service.labels }}
+    {{- toYaml . | nindent 4 }}
+    {{- end }}
+  annotations:
+    {{- toYaml .Values.ruler_querier.service.annotations | nindent 4 }}
+  namespace: {{ .Release.Namespace | quote }}
+spec:
+  type: ClusterIP
+  ports:
+    - port: {{ include "mimir.serverHttpListenPort" .}}
+      protocol: TCP
+      name: http-metrics
+      targetPort: http-metrics
+    - port: {{ include "mimir.serverGrpcListenPort" . }}
+      protocol: TCP
+      name: grpc
+      targetPort: grpc
+  selector:
+    {{- include "mimir.selectorLabels" (dict "ctx" . "component" "ruler-querier" "memberlist" true) | nindent 4 }}
+{{- end }}