Adding env vars for the stuck app/driver cleanups + v0.6.3 (#253)

spotinst · Dec 13, 2024 · b08678d · b08678d
1 parent c90ad92
commit b08678d
Show file tree

Hide file tree

Showing 3 changed files with 26 additions and 3 deletions.
diff --git a/charts/bigdata-spark-watcher/Chart.yaml b/charts/bigdata-spark-watcher/Chart.yaml
@@ -2,8 +2,8 @@ apiVersion: v2
 name: bigdata-spark-watcher
 description: A Helm chart for the Spot Big Data Spark Watcher
 type: application
-version: 0.6.3
-appVersion: 0.6.2
+version: 0.6.4
+appVersion: 0.6.3
 home: https://github.com/spotinst/charts
 icon: https://docs.spot.io/_media/images/spot_mark.png
 sources:

diff --git a/charts/bigdata-spark-watcher/templates/deployment.yaml b/charts/bigdata-spark-watcher/templates/deployment.yaml
@@ -75,6 +75,20 @@ spec:
               value: {{ .Values.k8sEventLogCollector.bucketPrefix }}
             - name: KUBE_EVENT_LOG_COLLECTION_ENVIRONMENT
               value: {{ .Values.k8sEventLogCollector.environment }}
+            {{- if or .Values.stuckAppCleanupEnabled .Values.stuckDriverCleanupEnabled }}
+            - name: WATCHER_STUCK_SPARK_APP_CLEANUP_ENABLED
+              value: {{ .Values.stuckAppCleanupEnabled | quote }}
+            - name: WATCHER_STUCK_DRIVER_CLEANUP_ENABLED
+              value: {{ .Values.stuckDriverCleanupEnabled | quote }}
+            {{- with .Values.stuckCleanup }}
+            - name: WATCHER_STUCK_TTL
+              value: {{ .gracePeriod }}
+            - name: WATCHER_STUCK_TTL_LEEWAY
+              value: {{ .leeway }}
+            - name: WATCHER_STUCK_CHECK_FREQUENCY
+              value: {{ .period }}
+            {{- end }}
+            {{- end }}
             - name: AWS_CREDENTIALS_FILE
               value: /creds/aws
             - name: GCP_CREDENTIALS_FILE

diff --git a/charts/bigdata-spark-watcher/values.yaml b/charts/bigdata-spark-watcher/values.yaml
@@ -8,7 +8,7 @@ image:
   repository: 066597193667.dkr.ecr.us-east-1.amazonaws.com/private/bigdata-spark-watcher
   pullPolicy: IfNotPresent
   # Overrides the image tag whose default is the chart appVersion.
-  tag: 0.6.2-0fdc670d
+  tag: 0.6.3-79527191
 
 imagePullSecrets:
   - name: spot-bigdata-image-pull
@@ -42,6 +42,15 @@ k8sEventLogCollector:
   bucketPrefix: "spot-bigdata-logcollector"
   environment: ""
 
+# cleans up stuck failed apps without termination times
+stuckAppCleanupEnabled: true
+# cleans up orphaned drivers
+stuckDriverCleanupEnabled: true
+stuckCleanup:
+  gracePeriod: 10m
+  leeway: 2m
+  period: 2m
+
 serviceAccount:
   create: true
   annotations: {}