support-bundle Helm chart to collect diagnostic data from a ClickHouse Private cluster. The resulting archive can be shared with ClickHouse support to speed up incident resolution.
Prerequisites
The chart relies on Troubleshoot, a kubectl plugin for cluster diagnostics. Install it via Krew:Generating and Running the Bundle
The chart does not install any resources into the cluster. Instead, it renders aSupportBundle manifest that is consumed by the kubectl support-bundle plugin.
Given a ClickHouse cluster named default-xx-01, render the chart locally and run the bundle:
tar.gz archive to the current directory.
What is Collected
The bundle collects data from two sources: the ClickHouse Operator namespace and the cluster namespace (derived asns-<cluster-name>).
ClickHouse Operator (toggle with supportBundle.collectors.operator.enabled):
- Container logs from the operator deployment
- Kubernetes resources in the operator namespace (Pods, Services, ConfigMaps, Events, etc.)
- Helm release values for charts installed in the operator namespace
supportBundle.collectors.clickhouseCluster.enabled):
- ClickHouse Server container logs
- ClickHouse Keeper container logs
- Kubernetes resources in the cluster namespace
- Helm release values for charts installed in the cluster namespace
- Output of
SELECT version()from the first ClickHouse Server pod - Output of the Keeper
mntrcommand from the first Keeper pod
Customizing Collection
The defaults are tuned for most scenarios but can be adjusted. For example, to collect only the last 24 hours of logs and skip operator data:supportBundle.collectors.operator.namespace accordingly.