Skip to content

Commit 4b9b1b4

Browse files
fix(capi): add startupProbe to capi-controller-manager to survive cert provisioning delay
The capi-controller-manager liveness probe has no initialDelaySeconds (defaults to 0), giving the container only 30s (3×10s) to serve /healthz. In controller-runtime, the health endpoint starts after the webhook server, which needs its TLS cert from cert-manager. During a loaded install, cert-manager is busy issuing many certs simultaneously and the CAPI webhook cert can take >30s — causing a kill loop (6+ restarts observed in CI). A startupProbe with failureThreshold=30 gives up to 300s for the initial startup, after which the regular livenessProbe takes over for ongoing health monitoring. This is the standard pattern for slow-starting containers with external dependencies. Co-Authored-By: Claude Sonnet 4.6 <[email protected]> Signed-off-by: Myasnikov Daniil <[email protected]>
1 parent 8c411d8 commit 4b9b1b4

1 file changed

Lines changed: 6 additions & 0 deletions

File tree

‎packages/system/capi-providers-core/files/core-components.yaml‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -15144,6 +15144,12 @@ spec:
1514415144
fieldPath: metadata.uid
1514515145
image: registry.k8s.io/cluster-api/cluster-api-controller:v1.10.1
1514615146
imagePullPolicy: IfNotPresent
15147+
startupProbe:
15148+
httpGet:
15149+
path: /healthz
15150+
port: healthz
15151+
failureThreshold: 30
15152+
periodSeconds: 10
1514715153
livenessProbe:
1514815154
httpGet:
1514915155
path: /healthz

0 commit comments

Comments
 (0)