Add topologySpreadConstraints (kubeflow#2091)

* Update README and documentation (kubeflow#2047) * Update docs Signed-off-by: Yi Chen <[email protected]> * Remove docs and update README Signed-off-by: Yi Chen <[email protected]> * Add link to monthly community meeting Signed-off-by: Yi Chen <[email protected]> --------- Signed-off-by: Yi Chen <[email protected]> Signed-off-by: jbhalodia-slack <[email protected]> * Add PodDisruptionBudget to chart (kubeflow#2078) * Add PodDisruptionBudget to chart Signed-off-by: Carlos Sánchez Páez <[email protected]> Signed-off-by: Carlos Sánchez Páez <[email protected]> Signed-off-by: Carlos Sánchez Páez <[email protected]> * PR comments Signed-off-by: Carlos Sánchez Páez <[email protected]> --------- Signed-off-by: Carlos Sánchez Páez <[email protected]> Signed-off-by: Carlos Sánchez Páez <[email protected]> Signed-off-by: jbhalodia-slack <[email protected]> * Set topologySpreadConstraints Signed-off-by: jbhalodia-slack <[email protected]> * Update README and increase patch version Signed-off-by: jbhalodia-slack <[email protected]> * Revert replicaCount change Signed-off-by: jbhalodia-slack <[email protected]> * Update README after master merger Signed-off-by: jbhalodia-slack <[email protected]> * Update README Signed-off-by: jbhalodia-slack <[email protected]> --------- Signed-off-by: Yi Chen <[email protected]> Signed-off-by: jbhalodia-slack <[email protected]> Signed-off-by: Carlos Sánchez Páez <[email protected]> Signed-off-by: Carlos Sánchez Páez <[email protected]> Co-authored-by: Yi Chen <[email protected]> Co-authored-by: Carlos Sánchez Páez <[email protected]>
zendesk · Jul 26, 2024 · 4108f54 · 4108f54
1 parent 51e4886
commit 4108f54
Show file tree

Hide file tree

Showing 6 changed files with 80 additions and 5 deletions.
diff --git a/charts/spark-operator-chart/Chart.yaml b/charts/spark-operator-chart/Chart.yaml
@@ -1,7 +1,7 @@
 apiVersion: v2
 name: spark-operator
 description: A Helm chart for Spark on Kubernetes operator
-version: 1.4.5
+version: 1.4.6
 appVersion: v1beta2-1.6.2-3.5.0
 keywords:
   - spark

diff --git a/charts/spark-operator-chart/README.md b/charts/spark-operator-chart/README.md
@@ -1,6 +1,6 @@
 # spark-operator
 
-![Version: 1.4.5](https://img.shields.io/badge/Version-1.4.5-informational?style=flat-square) ![AppVersion: v1beta2-1.6.2-3.5.0](https://img.shields.io/badge/AppVersion-v1beta2--1.6.2--3.5.0-informational?style=flat-square)
+![Version: 1.4.6](https://img.shields.io/badge/Version-1.4.6-informational?style=flat-square) ![AppVersion: v1beta2-1.6.2-3.5.0](https://img.shields.io/badge/AppVersion-v1beta2--1.6.2--3.5.0-informational?style=flat-square)
 
 A Helm chart for Spark on Kubernetes operator
 
@@ -132,6 +132,7 @@ See [helm uninstall](https://helm.sh/docs/helm/helm_uninstall) for command docum
 | sidecars | list | `[]` | Sidecar containers |
 | sparkJobNamespaces | list | `[""]` | List of namespaces where to run spark jobs |
 | tolerations | list | `[]` | List of node taints to tolerate |
+| topologySpreadConstraints | list | `[]` | Topology spread constraints rely on node labels to identify the topology domain(s) that each Node is in. Ref: [Pod Topology Spread Constraints](https://kubernetes.io/docs/concepts/workloads/pods/pod-topology-spread-constraints/) Specify topologySpreadConstraints without the labelSelector field, the labelSelector field will be set to "spark-operator.selectorLabels" subtemplate in the deployment.yaml file. |
 | uiService.enable | bool | `true` | Enable UI service creation for Spark application |
 | volumeMounts | list | `[]` |  |
 | volumes | list | `[]` |  |

diff --git a/charts/spark-operator-chart/templates/deployment.yaml b/charts/spark-operator-chart/templates/deployment.yaml
@@ -138,3 +138,13 @@ spec:
       tolerations:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      {{- if and .Values.topologySpreadConstraints (gt (int .Values.replicaCount) 1) }}
+      {{- $selectorLabels := include "spark-operator.selectorLabels" . | fromYaml -}}
+      {{- $labelSelectorDict := dict "labelSelector" ( dict "matchLabels" $selectorLabels ) }}
+      topologySpreadConstraints:
+      {{- range .Values.topologySpreadConstraints }}
+      - {{ mergeOverwrite . $labelSelectorDict | toYaml | nindent 8 | trim }}
+      {{- end }}
+      {{ else if and .Values.topologySpreadConstraints (eq (int .Values.replicaCount) 1) }}
+      {{ fail "replicaCount must be greater than 1 to enable topologySpreadConstraints."}}
+      {{- end }}
diff --git a/charts/spark-operator-chart/tests/deployment_test.yaml b/charts/spark-operator-chart/tests/deployment_test.yaml
@@ -299,3 +299,54 @@ tests:
             - key: key2
               operator: Exists
               effect: NoSchedule
+
+  - it: Should not contain topologySpreadConstraints if topologySpreadConstraints is not set
+    set:
+      topologySpreadConstraints: []
+    asserts:
+      - notExists:
+          path: spec.template.spec.topologySpreadConstraints
+
+  - it: Should add topologySpreadConstraints if topologySpreadConstraints is set and replicaCount is greater than 1
+    set:
+      replicaCount: 2
+      topologySpreadConstraints:
+      - maxSkew: 1
+        topologyKey: topology.kubernetes.io/zone
+        whenUnsatisfiable: ScheduleAnyway
+      - maxSkew: 1
+        topologyKey: kubernetes.io/hostname
+        whenUnsatisfiable: DoNotSchedule
+    asserts:
+      - equal:
+          path: spec.template.spec.topologySpreadConstraints
+          value:
+          - labelSelector:
+              matchLabels:
+                app.kubernetes.io/instance: spark-operator
+                app.kubernetes.io/name: spark-operator
+            maxSkew: 1
+            topologyKey: topology.kubernetes.io/zone
+            whenUnsatisfiable: ScheduleAnyway
+          - labelSelector:
+              matchLabels:
+                app.kubernetes.io/instance: spark-operator
+                app.kubernetes.io/name: spark-operator
+            maxSkew: 1
+            topologyKey: kubernetes.io/hostname
+            whenUnsatisfiable: DoNotSchedule
+
+  - it: Should fail if topologySpreadConstraints is set and replicaCount is not greater than 1
+    set:
+      replicaCount: 1
+      topologySpreadConstraints:
+      - maxSkew: 1
+        topologyKey: topology.kubernetes.io/zone
+        whenUnsatisfiable: ScheduleAnyway
+      - maxSkew: 1
+        topologyKey: kubernetes.io/hostname
+        whenUnsatisfiable: DoNotSchedule
+    asserts:
+      - failedTemplate:
+          errorMessage: "replicaCount must be greater than 1 to enable topologySpreadConstraints."
+
diff --git a/charts/spark-operator-chart/tests/poddisruptionbudget_test.yaml b/charts/spark-operator-chart/tests/poddisruptionbudget_test.yaml
@@ -17,21 +17,22 @@ tests:
 
   - it: Should render spark operator podDisruptionBudget if podDisruptionBudget.enable is true
     set:
+      replicaCount: 2
       podDisruptionBudget:
         enable: true
-    documentIndex: 0
     asserts:
       - containsDocument:
           apiVersion: policy/v1
           kind: PodDisruptionBudget
-          name: spark-operator-podDisruptionBudget
+          name: spark-operator-pdb
 
   - it: Should set minAvailable from values
     set:
+      replicaCount: 2
       podDisruptionBudget:
         enable: true
         minAvailable: 3
     asserts:
       - equal:
-          path: spec.template.minAvailable
+          path: spec.minAvailable
           value: 3
diff --git a/charts/spark-operator-chart/values.yaml b/charts/spark-operator-chart/values.yaml
@@ -143,6 +143,18 @@ podDisruptionBudget:
   # Require `replicaCount` to be greater than 1
   minAvailable: 1
 
+# -- Topology spread constraints rely on node labels to identify the topology domain(s) that each Node is in.
+# Ref: [Pod Topology Spread Constraints](https://kubernetes.io/docs/concepts/workloads/pods/pod-topology-spread-constraints/)
+# Specify topologySpreadConstraints without the labelSelector field, the labelSelector field will be set
+# to "spark-operator.selectorLabels" subtemplate in the deployment.yaml file.
+topologySpreadConstraints: []
+# - maxSkew: 1
+#   topologyKey: topology.kubernetes.io/zone
+#   whenUnsatisfiable: ScheduleAnyway
+# - maxSkew: 1
+#   topologyKey: kubernetes.io/hostname
+#   whenUnsatisfiable: DoNotSchedule
+
 # nodeSelector -- Node labels for pod assignment
 nodeSelector: {}