Troubleshooting clusters

Common cluster problems - stuck states, timeouts, degraded executors, and connection issues - with fixes.

Symptom

Likely cause

Fix

Cluster stuck in Creating

Insufficient capacity

Wait for resources; contact support if persistent

Cluster suspended unexpectedly

Auto-suspend idle timeout fired

Increase the idle timeout or disable auto-suspend for always-on clusters

Resume taking > 90 s

Large cluster (XL+) or cold nodes

Expected for large sizes; keep min executors > 0 for fast resume

Cluster in Failed state

Startup or configuration error

Check the error code in Activity History and follow Lifecycle states and error codes

Queries timing out

Query timeout too low or cluster under-provisioned

Raise the query timeout or scale up executors

Cluster degraded (some executors down)

Node issues or resource pressure

Queries still run, slower; the cluster may self-heal

Auto-resume not working

Auto-resume disabled, or the workspace is disabled

Verify auto-resume is enabled; check whether the workspace itself is disabled

Version update stuck

New version failing health checks

Check the deployment phase; e6data auto-rolls back if the new version is unhealthy - see Zero-downtime upgrades

Can't connect to a running cluster

Networking, not the cluster

Check network configuration and firewall rules - see Endpoints and cluster ingress

Use Activity History on the cluster as the first stop - it records each operation with timestamps and error detail. For the full error-code reference, see Lifecycle states and error codes.

See also