Merge e32a353 into f896d90

charts/operator-wandb/Chart.lock

@@ -4,7 +4,7 @@ dependencies:
   version: 0.1.0
 - name: wandb-base
   repository: file://../wandb-base
-  version: 0.3.4
+  version: 0.4.0
 - name: console
   repository: file://charts/console
   version: 0.1.0
@@ -34,7 +34,7 @@ dependencies:
   version: 25.3.5
 - name: etcd
   repository: oci://registry-1.docker.io/bitnamicharts
-  version: 10.7.2
+  version: 10.7.3
 - name: bufstream
   repository: file://charts/bufstream
   version: 0.3.5
@@ -58,9 +58,9 @@ dependencies:
   version: 0.1.0
 - name: wandb-base
   repository: file://../wandb-base
-  version: 0.3.4
+  version: 0.4.0
 - name: wandb-base
   repository: file://../wandb-base
-  version: 0.3.4
-digest: sha256:98ce75756c091ccd2e7bfb092be88f84444f05e1493bc4adfc66208267a452dd
-generated: "2025-01-15T19:56:47.399339-06:00"
+  version: 0.4.0
+digest: sha256:d9e104084992aaca36330a56d4a5b825fd20f0bdf995070bf93f46a883375169
+generated: "2025-01-21T12:17:36.303852-06:00"

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: 0.23.9
+version: 0.24.0
 appVersion: 1.0.0
 icon: https://wandb.ai/logo.svg
 

charts/operator-wandb/templates/_kafka.tpl

@@ -89,3 +89,14 @@ Return the number of partitions for run-updates-shadow
 {{- define "wandb.kafka.runUpdatesShadowNumPartitions" -}}
 {{- print .Values.global.kafka.runUpdatesShadowNumPartitions -}}
 {{- end -}}
+
+{{- define "wandb.runUpdateShadowTopicProducer" -}}
+{{- if .Values.global.pubSub.enabled -}}
+pubsub:/{{ .Values.global.pubSub.project }}/{{ .Values.global.pubSub.runUpdateShadowTopic }}
+{{- else if .Values.global.beta.bufstream.enabled -}}
+kafka://$(KAFKA_BROKER_HOST):$(KAFKA_BROKER_PORT)/$(KAFKA_TOPIC_RUN_UPDATE_SHADOW_QUEUE)?producer_batch_bytes=1048576&num_partitions=$(KAFKA_RUN_UPDATE_SHADOW_QUEUE_NUM_PARTITIONS)&replication_factor=3
+{{- else -}}
+kafka://$(KAFKA_CLIENT_USER):$(KAFKA_CLIENT_PASSWORD)@$(KAFKA_BROKER_HOST):$(KAFKA_BROKER_PORT)/$(KAFKA_TOPIC_RUN_UPDATE_SHADOW_QUEUE)?producer_batch_bytes=1048576&num_partitions=$(KAFKA_RUN_UPDATE_SHADOW_QUEUE_NUM_PARTITIONS)&replication_factor=3
+{{- end -}}
+{{- end -}}
+

charts/operator-wandb/templates/_mysql.tpl

@@ -52,7 +52,7 @@ Return the db password
 Return the db connection string
 */}}
 {{- define "wandb.mysql" -}}
-mysql://$(MYSQL_USER):$(MYSQL_PASSWORD)@$(MYSQL_HOST):$(MYSQL_PORT)/$(MYSQL_DATABASE)
+mysql://$(MYSQL_USER):$(MYSQL_PASSWORD)@$(MYSQL_HOST):$(MYSQL_PORT)/$(MYSQL_DATABASE)?tls=preferred
 {{- end -}}
 
 

charts/operator-wandb/templates/gorilla.yaml

@@ -78,7 +78,7 @@ data:
   GORILLA_ARTIFACT_GC_ENABLED: "false"
   GORILLA_RUN_STORE_ONPREM_MIGRATE_SHADOW_RUN_UPDATES: "true"
   GORILLA_GLUE_TASK_PROVIDER: "memory://"
-  GORILLA_DEFAULT_REGION: "{{ default "minio-local" ((include "wandb.bucket" . | fromYaml).region) }}"
+  GORILLA_DEFAULT_REGION: "minio-local"
   {{- if .Values.global.executor.enabled }}
   GORILLA_TASK_QUEUE_WORKER_ENABLED: "false"
   {{- else }}

charts/operator-wandb/values.yaml

@@ -622,8 +622,14 @@ glue:
 
 api:
   # install: false
+  autoscaling:
+    horizontal:
+      enabled: true
+      maxReplicas: 3
+      minReplicas: 2
   service:
     enabled: true
+    loadBalancerHealthCheckPath: "/ready"
     type: ClusterIP
     ports:
       - port: 8081
@@ -673,6 +679,22 @@ api:
       value: '{{ (include "wandb.bucket" . | fromYaml).url }}'
     GORILLA_STORAGE_BUCKET:
       value: '{{ (include "wandb.bucket" . | fromYaml).url }}'
+    GORILLA_RUN_UPDATE_SHADOW_QUEUE:
+      value: >
+        {
+          "overflow-bucket": {
+            "store": {{ (include "wandb.bucket" . | fromYaml).url | quote}},
+            "name": "wandb",
+            "prefix": "wandb-overflow"
+          },
+          "addr": {{ include "wandb.runUpdateShadowTopicProducer" . | quote }}
+        }
+    KAFKA_CLIENT_PASSWORD:
+      valueFrom:
+        secretKeyRef:
+          name: '{{ include "wandb.kafka.passwordSecret" . }}'
+          key: '{{ include "wandb.kafka.passwordSecret.passwordKey" . }}'
+          optional: true
     GORILLA_PORT:
       value: "8081"
     AZURE_STORAGE_KEY:
@@ -687,6 +709,7 @@ api:
     "{{ .Release.Name }}-mysql-configmap": "configMapRef"
     "{{ .Release.Name }}-redis-secret": "secretRef"
     "{{ .Release.Name }}-redis-configmap": "configMapRef"
+    "{{ .Release.Name }}-kafka-configmap": "configMapRef"
     "{{ .Release.Name }}-global-secret": "secretRef"
     "{{ .Release.Name }}-gorilla-secret": "secretRef"
     "{{ .Release.Name }}-gorilla-configmap": "configMapRef"
@@ -747,10 +770,17 @@ api:
           path: /healthz
           port: 8081
         initialDelaySeconds: 30
-        periodSeconds: 1
+        periodSeconds: 5
         timeoutSeconds: 1
         successThreshold: 1
         failureThreshold: 3
+      readinessProbe:
+        httpGet:
+          path: /ready
+          port: api
+        periodSeconds: 5
+        successThreshold: 1
+        failureThreshold: 5
       resources:
         limits:
           cpu: "2"

charts/wandb-base/Chart.yaml

@@ -2,7 +2,7 @@ apiVersion: v2
 name: wandb-base
 description: A generic helm chart for deploying services to kubernetes
 type: application
-version: 0.3.4
+version: 0.4.0
 icon: https://wandb.ai/logo.svg
 
 maintainers:

charts/wandb-base/templates/service.yaml

@@ -1,14 +1,32 @@
 {{- if .Values.service.enabled }}
+{{- if .Values.global.createGCPLoadBalancerBackend }}
+{{- if eq .Values.global.cloudProvider "gcp" }}
+---
+apiVersion: cloud.google.com/v1
+kind: BackendConfig
+metadata:
+  name: {{ include "wandb-base.fullname" . }}-backend-config
+  labels:
+    {{- include "wandb-base.labels" . | nindent 4 }}
+spec:
+  timeoutSec: 120
+{{- end }}
+{{- end }}
+---
 apiVersion: v1
 kind: Service
 metadata:
   name: {{ include "wandb-base.fullname" . }}
   labels:
     {{- include "wandb-base.labels" . | nindent 4 }}
-  {{- if eq .Values.global.cloudProvider "aws" }}
   annotations:
-    alb.ingress.kubernetes.io/healthcheck-path: /ready
-  {{- end }}
+    {{- if and ( eq .Values.global.cloudProvider "aws" ) .Values.service.loadBalancerHealthCheckPath }}
+    alb.ingress.kubernetes.io/healthcheck-path: {{ .Values.service.loadBalancerHealthCheckPath }}
+    {{- end }}
+    {{- if and .Values.global.createGCPLoadBalancerBackend (eq .Values.global.cloudProvider "gcp") }}
+    cloud.google.com/neg: '{"ingress": true}'
+    cloud.google.com/backend-config: '{"default": "{{ include "wandb-base.fullname" . }}-backend-config"}'
+    {{- end }}
 spec:
   type: {{ .Values.service.type }}
   ports:
@@ -17,4 +35,4 @@ spec:
     {{- end }}
   selector:
     {{- include "wandb-base.selectorLabels" . | nindent 4 }}
-  {{ end }}
+{{- end }}

charts/wandb-base/values.yaml

@@ -164,6 +164,8 @@ cronJobs: {}
 # This is for setting up a service more information can be found here: https://kubernetes.io/docs/concepts/services-networking/service/
 service:
   enabled: true
+  # This path is currently only used for AWS ALB health checks
+  loadBalancerHealthCheckPath: ""
   # This sets the service type more information can be found here: https://kubernetes.io/docs/concepts/services-networking/service/#publishing-services-service-types
   type: ClusterIP
   ports: