CP-34996: change deployment strategy for agent when persistent volume is enabled

helm/templates/agent-deploy.yaml

@@ -16,6 +16,10 @@ spec:
   {{- else }}
   replicas: 1
   {{- end }}
+  {{- if .Values.server.persistentVolume.enabled }}
+  strategy:
+    type: Recreate
+  {{- end }}
   template:
     metadata:
       {{- include "cloudzero-agent.generateAnnotations" .Values.server.podAnnotations | nindent 8 }}

helm/tests/persistent_volume_test.yaml

@@ -0,0 +1,196 @@
+suite: test persistent volume configuration
+templates:
+  - templates/agent-deploy.yaml
+  - templates/agent-pvc.yaml
+tests:
+  # Test persistent volume disabled (default behavior)
+  - it: should use emptyDir when persistentVolume is disabled
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      server.persistentVolume.enabled: false
+    asserts:
+      # Verify emptyDir is used for storage
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: cloudzero-agent-storage-volume
+            emptyDir:
+              sizeLimit: 8Gi
+      # Verify strategy is not set (should use default RollingUpdate)
+      - isNull:
+          path: spec.strategy
+
+  # Test persistent volume enabled - basic configuration
+  - it: should use PVC when persistentVolume is enabled
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      server.persistentVolume.enabled: true
+      server.persistentVolume.size: 10Gi
+    asserts:
+      # Verify PVC is used for storage
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: cloudzero-agent-storage-volume
+            persistentVolumeClaim:
+              claimName: RELEASE-NAME-cz-server
+      # Verify strategy is set to Recreate
+      - equal:
+          path: spec.strategy.type
+          value: Recreate
+      # Verify volume mount is configured
+      - contains:
+          path: spec.template.spec.containers[1].volumeMounts
+          content:
+            name: cloudzero-agent-storage-volume
+            mountPath: /data
+            subPath: ""
+
+  # Test PVC is created when enabled
+  - it: should create PVC when persistentVolume is enabled
+    template: templates/agent-pvc.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      server.persistentVolume.enabled: true
+      server.persistentVolume.size: 15Gi
+      server.persistentVolume.storageClass: "fast-ssd"
+    asserts:
+      - isKind:
+          of: PersistentVolumeClaim
+      - equal:
+          path: metadata.name
+          value: RELEASE-NAME-cz-server
+      - equal:
+          path: spec.resources.requests.storage
+          value: 15Gi
+      - equal:
+          path: spec.storageClassName
+          value: fast-ssd
+
+  # Test PVC is not created when disabled
+  - it: should not create PVC when persistentVolume is disabled
+    template: templates/agent-pvc.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      server.persistentVolume.enabled: false
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  # Test existing PVC claim name is used
+  - it: should use existingClaim when specified
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      server.persistentVolume.enabled: true
+      server.persistentVolume.existingClaim: "my-existing-pvc"
+    asserts:
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: cloudzero-agent-storage-volume
+            persistentVolumeClaim:
+              claimName: my-existing-pvc
+      # Should still use Recreate strategy with existing claim
+      - equal:
+          path: spec.strategy.type
+          value: Recreate
+
+  # Test custom mount path
+  - it: should use custom mountPath when specified
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      server.persistentVolume.enabled: true
+      server.persistentVolume.mountPath: /custom/data/path
+      server.persistentVolume.subPath: "prometheus-data"
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[1].volumeMounts
+          content:
+            name: cloudzero-agent-storage-volume
+            mountPath: /custom/data/path
+            subPath: "prometheus-data"
+
+  # Test persistent volume with agent mode (Prometheus)
+  - it: should work with agent mode (Prometheus with --enable-feature=agent)
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      components.agent.mode: agent
+      server.persistentVolume.enabled: true
+    asserts:
+      # Verify Prometheus container is used
+      - equal:
+          path: spec.template.spec.containers[1].name
+          value: cloudzero-agent-server
+      # Verify strategy is Recreate
+      - equal:
+          path: spec.strategy.type
+          value: Recreate
+      # Verify PVC is mounted
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: cloudzero-agent-storage-volume
+            persistentVolumeClaim:
+              claimName: RELEASE-NAME-cz-server
+
+  # Test persistent volume with server mode (standard Prometheus)
+  - it: should work with server mode (standard Prometheus)
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      components.agent.mode: server
+      server.persistentVolume.enabled: true
+    asserts:
+      # Verify Prometheus container is used
+      - equal:
+          path: spec.template.spec.containers[1].name
+          value: cloudzero-agent-server
+      # Verify strategy is Recreate
+      - equal:
+          path: spec.strategy.type
+          value: Recreate
+      # Verify PVC is mounted
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: cloudzero-agent-storage-volume
+            persistentVolumeClaim:
+              claimName: RELEASE-NAME-cz-server
+
+  # Test persistent volume with federated mode
+  - it: should work with federated mode
+    template: templates/agent-deploy.yaml
+    set:
+      apiKey: "test-key"
+      existingSecretName: null
+      components.agent.mode: federated
+      server.persistentVolume.enabled: true
+    asserts:
+      # Verify Prometheus container is used
+      - equal:
+          path: spec.template.spec.containers[1].name
+          value: cloudzero-agent-server
+      # Verify strategy is Recreate
+      - equal:
+          path: spec.strategy.type
+          value: Recreate
+      # Verify PVC is mounted
+      - contains:
+          path: spec.template.spec.volumes
+          content:
+            name: cloudzero-agent-storage-volume
+            persistentVolumeClaim:
+              claimName: RELEASE-NAME-cz-server

helm/tests/server_deployment_name_independence_test.yaml

@@ -99,7 +99,7 @@ tests:
     set:
       apiKey: "test-key"
       existingSecretName: null
-      components.agent.mode: clustered
+      components.agent.mode: agent
       server.persistentVolume.enabled: true
       server.persistentVolume.size: 10Gi
       server.persistentVolume.accessModes:
@@ -117,7 +117,7 @@ tests:
     set:
       apiKey: "test-key"
       existingSecretName: null
-      components.agent.mode: clustered
+      components.agent.mode: agent
       server.persistentVolume.enabled: true
       server.persistentVolume.size: 10Gi
       server.persistentVolume.accessModes:

helm/values.schema.json

@@ -5410,6 +5410,42 @@
           }
         }
       }
+    },
+    {
+      "if": {
+        "properties": {
+          "components": {
+            "properties": {
+              "agent": {
+                "properties": {
+                  "mode": {
+                    "const": "clustered"
+                  }
+                },
+                "type": "object"
+              }
+            },
+            "type": "object"
+          }
+        }
+      },
+      "then": {
+        "properties": {
+          "server": {
+            "properties": {
+              "persistentVolume": {
+                "properties": {
+                  "enabled": {
+                    "const": false
+                  }
+                },
+                "type": "object"
+              }
+            },
+            "type": "object"
+          }
+        }
+      }
     }
   ],
   "anyOf": [

helm/values.schema.yaml

@@ -2364,3 +2364,38 @@ allOf:
                   oneOf:
                     - type: "null"
                     - const: federated
+
+  # Conditional: Persistent volume constraint for clustered mode
+  #
+  # If components.agent.mode is "clustered" (Alloy mode), then
+  # server.persistentVolume.enabled must be false.
+  #
+  # This is necessary because:
+  # 1. Alloy doesn't use the persistent volume mount (it uses /tmp/alloy)
+  # 2. Alloy is designed for horizontal scaling with multiple replicas
+  # 3. Persistent volumes with ReadWriteOnce can't be shared across pods
+  # 4. The Recreate deployment strategy (required for PV) prevents proper
+  #    rolling updates needed for multi-replica deployments
+  #
+  # In clustered/Alloy mode, metrics data should be transient and stored
+  # in remote storage via remote_write configuration.
+  - if:
+      properties:
+        components:
+          type: object
+          properties:
+            agent:
+              type: object
+              properties:
+                mode:
+                  const: clustered
+    then:
+      properties:
+        server:
+          type: object
+          properties:
+            persistentVolume:
+              type: object
+              properties:
+                enabled:
+                  const: false

tests/helm/schema/server.persistentVolume.agent-mode-enabled.pass.yaml

@@ -0,0 +1,11 @@
+# Valid: Agent mode (Prometheus) with persistent volume enabled
+# Persistent volumes work fine with non-clustered modes (agent, server, federated)
+apiKey: "test-key"
+existingSecretName: null
+components:
+  agent:
+    mode: agent
+server:
+  persistentVolume:
+    enabled: true
+    size: 10Gi

tests/helm/schema/server.persistentVolume.clustered-mode-disabled.pass.yaml

@@ -0,0 +1,10 @@
+# Valid: Clustered mode (Alloy) with persistent volume explicitly disabled
+# Persistent volumes are not supported in clustered mode
+apiKey: "test-key"
+existingSecretName: null
+components:
+  agent:
+    mode: clustered
+server:
+  persistentVolume:
+    enabled: false

tests/helm/schema/server.persistentVolume.clustered-mode-invalid.fail.yaml

@@ -0,0 +1,13 @@
+# Invalid: Clustered mode (Alloy) with persistent volume enabled
+# This should fail because persistent volumes are incompatible with clustered mode:
+# - Alloy uses /tmp/alloy for ephemeral storage, not the PV mount
+# - Clustered mode supports multiple replicas, but PVs with ReadWriteOnce can't be shared
+# - The Recreate deployment strategy (required for PV) prevents rolling updates
+apiKey: "test-key"
+existingSecretName: null
+components:
+  agent:
+    mode: clustered
+server:
+  persistentVolume:
+    enabled: true