[HDX] Bump clickhouse-operator to v0.0.6 + fix ClickHouse OOMKill

.changeset/bump-clickhouse-operator-v0.0.6.md

@@ -0,0 +1,5 @@
+---
+"helm-charts": patch
+---
+
+chore(deps): bump clickhouse-operator-helm to v0.0.6

.changeset/clickhouse-explicit-resources.md

@@ -0,0 +1,21 @@
+---
+"helm-charts": patch
+---
+
+fix(clickhouse): harden ClickHouse defaults for clickhouse-operator v0.0.6
+
+Two operator-default changes in v0.0.6 broke the single-replica ClickHouse
+deployment; the chart now overrides both:
+
+- Explicit `containerTemplate.resources` (2Gi memory, 500m CPU request). The
+  operator otherwise applies a 512Mi default (request == limit as of v0.0.6),
+  which is too low for the full ClickStack schema and OOMKills the server
+  (exit 137) under ingestion plus background merges.
+
+- `settings.enableDatabaseSync: false`. The operator now defaults this to true,
+  which creates the `default` database with the Replicated (DatabaseReplicated)
+  engine so table metadata lives in Keeper. That feature targets multi-replica
+  clusters; in a single-replica deployment a transient Keeper hiccup during
+  startup desyncs the Replicated database and silently drops all seeded tables,
+  which never come back. Keeping `default` Atomic stores tables on the
+  persistent data volume so they survive restarts.

charts/clickstack-operators/Chart.lock

@@ -4,6 +4,6 @@ dependencies:
   version: 1.7.0
 - name: clickhouse-operator-helm
   repository: oci://ghcr.io/clickhouse
-  version: 0.0.2
-digest: sha256:1daf572004da83b1836c8867f11198530652fee6905d4786a2d5eef87bc611cd
-generated: "2026-03-04T16:52:51.068188-06:00"
+  version: 0.0.6
+digest: sha256:5afcb0d78e0ceecf1a18f3f7dfb52ee2627b7acea2621ffa411ee7bfb530adf7
+generated: "2026-06-19T17:41:30.214406+02:00"

charts/clickstack-operators/Chart.yaml

@@ -13,6 +13,6 @@ dependencies:
     repository: https://mongodb.github.io/helm-charts
     alias: mongodb-operator
   - name: clickhouse-operator-helm
-    version: "~0.0.2"
+    version: "~0.0.6"
     repository: oci://ghcr.io/clickhouse
     alias: clickhouse-operator

charts/clickstack-operators/values.yaml

@@ -9,8 +9,8 @@ mongodb-operator:
 # See https://clickhouse.com/docs/clickhouse-operator/overview for all options
 clickhouse-operator:
   webhook:
-    enable: false
+    enabled: false
   certManager:
-    enable: false
+    enabled: false
   crd:
-    enable: true
+    enabled: true

charts/clickstack/tests/clickhouse-deployment_test.yaml

@@ -68,6 +68,20 @@ tests:
           path: spec.dataVolumeClaimSpec.resources.requests.storage
           value: 10Gi
 
+  - it: should set explicit container resources so the operator default does not OOMKill ClickHouse
+    templates:
+      - clickhouse/cluster.yaml
+    asserts:
+      - equal:
+          path: spec.containerTemplate.resources.requests.memory
+          value: 2Gi
+      - equal:
+          path: spec.containerTemplate.resources.limits.memory
+          value: 2Gi
+      - equal:
+          path: spec.containerTemplate.resources.requests.cpu
+          value: 500m
+
   - it: should resolve keeperClusterRef template expression
     templates:
       - clickhouse/cluster.yaml
@@ -152,3 +166,11 @@ tests:
           path: spec.settings.extraUsersConfig.users.app
       - isNotNull:
           path: spec.settings.extraUsersConfig.users.otelcollector
+
+  - it: should disable operator databaseSync so the default DB stays Atomic
+    templates:
+      - clickhouse/cluster.yaml
+    asserts:
+      - equal:
+          path: spec.settings.enableDatabaseSync
+          value: false

charts/clickstack/values.yaml

@@ -328,6 +328,12 @@ clickhouse:
         image:
           repository: clickhouse/clickhouse-server
           tag: "25.7-alpine"
+        resources:
+          requests:
+            cpu: 500m
+            memory: 2Gi
+          limits:
+            memory: 2Gi
       replicas: 1
       shards: 1
       keeperClusterRef:
@@ -339,6 +345,9 @@ clickhouse:
           requests:
             storage: 10Gi
       settings:
+        # Keep the `default` database Atomic; the Replicated engine the operator
+        # selects when this is true drops seeded tables on Keeper desync.
+        enableDatabaseSync: false
         extraUsersConfig:
           users:
             app:

integration-tests/full-stack/assert.sh

@@ -12,7 +12,7 @@ echo "Waiting for services to initialize..."
 sleep 30
 
 echo "Waiting for all pods to be ready..."
-kubectl wait --for=condition=Ready pods --all --timeout=600s || true
+kubectl wait --for=condition=Ready pods --all --field-selector=status.phase!=Succeeded --timeout=600s || true
 
 echo "Pod status:"
 kubectl get pods -o wide
@@ -24,7 +24,7 @@ echo "Checking ClickHouseCluster CR..."
 kubectl get clickhousecluster -o wide || true
 
 echo "Waiting for all pods to be ready (final check)..."
-kubectl wait --for=condition=Ready pods --all --timeout=600s
+kubectl wait --for=condition=Ready pods --all --field-selector=status.phase!=Succeeded --timeout=600s
 
 echo "Final pod status:"
 kubectl get pods -o wide