Merge 7efa21d into 973d79d

charts/operator-wandb/Chart.lock

@@ -11,6 +11,9 @@ dependencies:
 - name: weave-trace
   repository: file://charts/weave-trace
   version: 0.1.0
+- name: executor
+  repository: file://charts/executor
+  version: 0.1.0
 - name: parquet
   repository: file://charts/parquet
   version: 0.1.0
@@ -41,5 +44,5 @@ dependencies:
 - name: yace
   repository: file://charts/yace
   version: 0.1.0
-digest: sha256:bca2b6781737da6806e4485605cf9ce87b1428944b14cb88f082024cc3500bbd
-generated: "2024-11-08T13:42:24.351184+05:30"
+digest: sha256:17ce8e027aefeca83fbae29fcb79da3dada09de63e9704b88c4144e975cbbce5
+generated: "2024-11-18T15:15:21.132048-08:00"

charts/operator-wandb/Chart.yaml

@@ -2,7 +2,7 @@ apiVersion: v2
 name: operator-wandb
 description: A Helm chart for deploying W&B to Kubernetes
 type: application
-version: 0.18.18
+version: 0.18.19
 appVersion: 1.0.0
 icon: https://wandb.ai/logo.svg
 
@@ -28,6 +28,10 @@ dependencies:
     version: "*.*.*"
     repository: file://charts/weave-trace
     condition: weave-trace.install
+  - name: executor
+    version: "*.*.*"
+    repository: file://charts/executor
+    condition: executor.install
   - name: parquet
     version: "*.*.*"
     repository: file://charts/parquet

charts/operator-wandb/charts/app/templates/_deployment.tpl

@@ -271,6 +271,10 @@ spec:
               value: {{ .Values.artifactsGc.NumWorkers | quote }}
             - name: GORILLA_ARTIFACTS_GC_DELETE_FILES_NUM_WORKERS
               value: {{ .Values.artifactsGc.DeleteFilesNumWorkers | quote }}
+            - name: GORILLA_TASK_QUEUE
+              value: "{{ include "app.redis" . | trim }}"
+            - name: GORILLA_TASK_QUEUE_MONITOR_PORT
+              value: "10000"
             {{- if index .Values.global "weave-trace" "enabled" }}
             - name: GORILLA_INTERNAL_JWT_SUBJECTS_TO_ISSUERS
               value: {{ tpl (include "app.internalJWTMap" .) . }}

charts/operator-wandb/charts/executor/.helmignore

@@ -0,0 +1,23 @@
+# Patterns to ignore when building packages.
+# This supports shell glob matching, relative path matching, and
+# negation (prefixed with !). Only one pattern per line.
+.DS_Store
+# Common VCS dirs
+.git/
+.gitignore
+.bzr/
+.bzrignore
+.hg/
+.hgignore
+.svn/
+# Common backup files
+*.swp
+*.bak
+*.tmp
+*.orig
+*~
+# Various IDEs
+.project
+.idea/
+*.tmproj
+.vscode/

charts/operator-wandb/charts/executor/Chart.yaml

@@ -0,0 +1,13 @@
+---
+apiVersion: v2
+name: executor
+description: Chart for deploying the executor
+type: application
+version: 0.1.0
+appVersion: "0.33.0"
+home: https://wandb.ai
+icon: https://wandb.ai/logo.svg
+maintainers:
+  - name: wandb
+    email: support@wandb.com
+    url: https://wandb.com

charts/operator-wandb/charts/executor/templates/_helpers.tpl

@@ -0,0 +1,129 @@
+{{/*
+Expand the name of the chart.
+*/}}
+{{- define "executor.name" -}}
+{{- default .Chart.Name .Values.nameOverride | trunc 63 | trimSuffix "-" }}
+{{- end }}
+
+{{/*
+Create a default fully qualified app name.
+We truncate at 63 chars because some Kubernetes name fields are limited to this (by the DNS naming spec).
+If release name contains chart name it will be used as a full name.
+*/}}
+{{- define "executor.fullname" -}}
+{{- if .Values.fullnameOverride }}
+{{- .Values.fullnameOverride | trunc 63 | trimSuffix "-" }}
+{{- else }}
+{{- $name := default .Chart.Name .Values.nameOverride }}
+{{- if contains $name .Release.Name }}
+{{- .Release.Name | trunc 63 | trimSuffix "-" }}
+{{- else }}
+{{- printf "%s-%s" .Release.Name $name | trunc 63 | trimSuffix "-" }}
+{{- end }}
+{{- end }}
+{{- end }}
+
+{{/*
+Create chart name and version as used by the chart label.
+*/}}
+{{- define "executor.chart" -}}
+{{- printf "%s-%s" .Chart.Name .Chart.Version | replace "+" "_" | trunc 63 | trimSuffix "-" }}
+{{- end }}
+
+{{/*
+Common labels
+*/}}
+{{- define "executor.labels" -}}
+helm.sh/chart: {{ include "executor.chart" . }}
+{{ include "executor.selectorLabels" . }}
+{{- if .Chart.AppVersion }}
+app.kubernetes.io/version: {{ .Chart.AppVersion | quote }}
+{{- end }}
+app.kubernetes.io/managed-by: {{ .Release.Service }}
+{{- end }}
+
+{{/*
+Returns the extraEnv keys and values to inject into containers.
+
+Global values will override any chart-specific values.
+*/}}
+{{- define "executor.extraEnv" -}}
+{{- $allExtraEnv := merge (default (dict) .local.extraEnv) .global.extraEnv -}}
+{{- range $key, $value := $allExtraEnv }}
+- name: {{ $key }}
+  value: {{ $value | quote }}
+{{- end -}}
+{{- end -}}
+
+{{/*
+Returns a list of _common_ labels to be shared across all
+executor deployments and other shared objects.
+*/}}
+{{- define "executor.commonLabels" -}}
+{{- $commonLabels := default (dict) .Values.common.labels -}}
+{{- if $commonLabels }}
+{{-   range $key, $value := $commonLabels }}
+{{ $key }}: {{ $value | quote }}
+{{-   end }}
+{{- end -}}
+{{- end -}}
+
+{{/*
+Returns a list of _pod_ labels to be shared across all
+executor deployments.
+*/}}
+{{- define "executor.podLabels" -}}
+{{- range $key, $value := .Values.pod.labels }}
+{{ $key }}: {{ $value | quote }}
+{{- end }}
+{{- end -}}
+{{/*
+Selector labels
+*/}}
+{{- define "executor.selectorLabels" -}}
+app.kubernetes.io/name: {{ include "executor.name" . }}
+app.kubernetes.io/instance: {{ .Release.Name }}
+{{- end }}
+
+{{/*
+Create the name of the service account to use
+*/}}
+{{- define "executor.serviceAccountName" -}}
+{{- if .Values.serviceAccount.create }}
+{{- default (include "executor.fullname" .) .Values.serviceAccount.name }}
+{{- else }}
+{{- default "default" .Values.serviceAccount.name }}
+{{- end }}
+{{- end }}
+
+{{- define "executor.redis" -}}
+{{- $cs := include "wandb.redis.connectionString" . }}
+{{- $ca := include "wandb.redis.caCert" . }}
+{{- if $ca }}
+{{- printf "%s?tls=true&caCertPath=/etc/ssl/certs/redis_ca.pem&ttlInSeconds=604800" $cs -}}
+{{- else }}
+{{- print $cs -}}
+{{- end }}
+{{- end }}
+
+{{- define "executor.bucket" -}}
+{{- $bucketValues := .Values.global.defaultBucket }}
+{{- if .Values.global.bucket.provider }}
+{{- $bucketValues = .Values.global.bucket }}
+{{- end }}
+{{- $bucket := "" -}}
+{{- if eq $bucketValues.provider "az" -}}
+{{- $bucket = printf "az://%s/%s" $bucketValues.name (default "" $bucketValues.path) -}}
+{{- end -}}
+{{- if eq $bucketValues.provider "gcs" -}}
+{{- $bucket = printf "gs://%s/%s" $bucketValues.name (default "" $bucketValues.path) -}}
+{{- end -}}
+{{- if eq $bucketValues.provider "s3" -}}
+{{- if and $bucketValues.accessKey $bucketValues.secretKey -}}
+{{- $bucket = printf "s3://%s:%s@%s/%s" $bucketValues.accessKey $bucketValues.secretKey $bucketValues.name (default "" $bucketValues.path) -}}
+{{- else -}}
+{{- $bucket = printf "s3://%s/%s" $bucketValues.name (default "" $bucketValues.path) -}}
+{{- end -}}
+{{- end -}}
+{{- trimSuffix "/" $bucket -}}
+{{- end -}}

charts/operator-wandb/charts/executor/templates/deployment.yaml

@@ -0,0 +1,172 @@
+{{- if .Values.install }}
+{{- $imageCfg := dict "global" $.Values.global.image "local" $.Values.image -}}
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ include "executor.fullname" . }}
+  labels:
+    {{- include "wandb.commonLabels" . | nindent 4 }}
+    {{- include "executor.commonLabels" . | nindent 4 }}
+    {{- include "executor.labels" . | nindent 4 }}
+    {{- if .Values.deployment.labels -}}
+    {{-   toYaml .Values.deployment.labels | nindent 4 }}
+    {{- end }}
+  annotations:
+    {{- include "wandb.deploymentAnnotations" . | nindent 4 }}
+    {{- if .Values.deployment.annotations -}}
+    {{-   toYaml .Values.deployment.annotations | nindent 4 }}
+    {{- end }}
+spec:
+  {{- if not .Values.autoscaling.enabled }}
+  replicas: {{ .Values.replicaCount }}
+  {{- end }}
+  selector:
+    matchLabels:
+      {{- include "wandb.selectorLabels" $ | nindent 6 }}
+      {{- include "executor.labels" . | nindent 6 }}
+  template:
+    metadata:
+      {{- with .Values.pod.annotations }}
+      annotations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      labels:
+        {{- include "wandb.podLabels" . | nindent 8 }}
+        {{- include "executor.labels" . | nindent 8 }}
+        {{- include "executor.commonLabels" . | nindent 8 }}
+        {{- include "executor.podLabels" . | nindent 8 }}
+    spec:
+      {{- with .Values.imagePullSecrets }}
+      imagePullSecrets:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      serviceAccountName: {{ include "executor.serviceAccountName" . }}
+      {{- include "wandb.nodeSelector" . | nindent 6 }}
+      {{- include "wandb.priorityClassName" . | nindent 6 }}
+      {{- include "wandb.podSecurityContext" .Values.pod.securityContext | nindent 6 }}
+      # needed to ensure ensure Kafka consumers handle pod termination gracefully and avoid data loss
+      terminationGracePeriodSeconds: 60
+      containers:
+        - name: {{ .Chart.Name }}
+          {{- include "wandb.containerSecurityContext" .Values.container.securityContext | nindent 10 }}
+          image: "{{ .Values.image.repository }}:{{ .Values.image.tag }}"
+          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          volumeMounts:
+            {{- if ne (include "wandb.redis.caCert" .) "" }}
+            - name: {{ include "executor.fullname" . }}-redis-ca
+              mountPath: /etc/ssl/certs/redis_ca.pem
+              subPath: redis_ca.pem
+            {{- end }}
+          env:
+            - name: GOMEMLIMIT
+              valueFrom:
+                resourceFieldRef:
+                  resource: limits.memory
+            {{- if ne .Values.traceRatio 0.0 }}
+            - name: GORILLA_TRACER
+              value: "otlp+grpc://{{ .Release.Name }}-otel-daemonset:4317?trace_ratio={{ .Values.traceRatio }}"
+            {{- end }}
+            - name: ONLY_SERVICE
+              value: gorilla-executor
+            - name: LOGGING_ENABLED
+              value: 'true'
+            - name: MYSQL_PORT
+              value: "{{ include "wandb.mysql.port" . }}"
+            - name: MYSQL_HOST
+              value: "{{ include "wandb.mysql.host" . }}"
+            - name: MYSQL_DATABASE
+              value: "{{ include "wandb.mysql.database" . }}"
+            - name: MYSQL_USER
+              value: "{{ include "wandb.mysql.user" . }}"
+            - name: MYSQL_PASSWORD
+              valueFrom:
+                secretKeyRef:
+                  name: {{ include "wandb.mysql.passwordSecret" . }}
+                  key: {{ .Values.global.mysql.passwordSecret.passwordKey }}
+            - name: MYSQL
+              value: "mysql://$(MYSQL_USER):$(MYSQL_PASSWORD)@$(MYSQL_HOST):$(MYSQL_PORT)/$(MYSQL_DATABASE)"
+            - name: REDIS_PASSWORD
+              valueFrom:
+                secretKeyRef:
+                  name: {{ include "wandb.redis.passwordSecret" . }}
+                  optional: true
+                  key: REDIS_PASSWORD
+            - name: REDIS_PORT
+              value: "{{ include "wandb.redis.port" . }}"
+            - name: REDIS_HOST
+              value: "{{ include "wandb.redis.host" . }}"
+            - name: REDIS
+              value: "{{ include "executor.redis" . | trim }}"
+
+            - name: GORILLA_SETTINGS_CACHE
+              value: "{{ include "executor.redis" . | trim }}"
+            - name: GORILLA_TASK_QUEUE
+              value: "{{ include "executor.redis" . | trim }}"
+
+            - name: BUCKET
+              value: "{{ include "executor.bucket" . }}"
+            - name: AWS_REGION
+              value: {{ .Values.global.bucket.region | default .Values.global.defaultBucket.region }}
+            - name: AWS_S3_KMS_ID
+              value: "{{ .Values.global.bucket.kmsKey | default .Values.global.defaultBucket.kmsKey }}"
+            - name: AZURE_STORAGE_KEY
+              valueFrom:
+                secretKeyRef:
+                  name: "{{ include "wandb.bucket.secret" . }}"
+                  key: {{ .Values.global.bucket.accessKeyName }}
+                  optional: true
+
+            - name: G_HOST_IP
+              valueFrom:
+                fieldRef:
+                  fieldPath: status.hostIP
+
+            {{- if ne .Values.global.email.smtp.host "" }}
+            - name: GORILLA_EMAIL_SINK
+              value: "smtp://{{ .Values.global.email.smtp.user }}:{{ .Values.global.email.smtp.password }}@{{ .Values.global.email.smtp.host }}:{{ .Values.global.email.smtp.port }}"
+            {{- end }}
+
+            {{- include "executor.extraEnv" (dict "global" .Values.global "local" .Values) | nindent 12 }}
+            {{- include "wandb.extraEnvFrom" (dict "root" $ "local" .) | nindent 12 }}
+
+          livenessProbe:
+            httpGet:
+              path: /ready
+              port: 8087
+            initialDelaySeconds: 30
+            periodSeconds: 10
+            timeoutSeconds: 5
+            failureThreshold: 3
+            successThreshold: 1
+
+          readinessProbe:
+            httpGet:
+              path: /ready
+              port: 8087
+            initialDelaySeconds: 30
+            periodSeconds: 10
+            timeoutSeconds: 5
+            failureThreshold: 3
+            successThreshold: 1
+
+          resources:
+            {{- toYaml .Values.resources | nindent 12 }}
+      {{- with .Values.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      volumes:
+        {{- if ne (include "wandb.redis.caCert" .) "" }}
+        - name: {{ include "executor.fullname" . }}-redis-ca
+          secret:
+            secretName: "{{ .Release.Name }}-redis"
+            items:
+              - key: REDIS_CA_CERT
+                path: redis_ca.pem
+        {{- end }}
+{{- end }}