- Readonly users — cannot set
max_execution_timedue to ClickHouse permissions. Use smaller batch sizes and shorter lookback periods, or create a non-readonly user with SELECT-only grants - Load balancer IPs — if ClickHouse sits behind a load balancer, client IPs show the LB address instead of real client IPs. Enable PROXY protocol to fix this. See ClickHouse Real Client IP
- No browser UI — interactive HTML report is generated as static files, not a live dashboard
- Single-cluster — Kubernetes resolution works against one cluster at a time
- Namespace creation —
clickspectre deploycreates the target namespace if it doesn't exist. This is the only write operation on Kubernetes
- Check DSN format:
clickhouse://user:password@host:port/database - Verify network connectivity
- Check ClickHouse is running:
clickhouse-client --query "SELECT 1"
Use smaller dataset (--lookback 7d --batch-size 1000 --max-rows 50000) or create a non-readonly user with GRANT SELECT ON *.* TO clickspectre.
- Check kubeconfig path
- Verify cluster connectivity:
kubectl cluster-info - Use
--resolve-k8s=falseto disable K8s resolution
- Increase timeout:
--query-timeout 10m - Reduce batch size:
--batch-size 50000 - Reduce max rows:
--max-rows 500000