trino-admin-cluster-health

What problem does it solve?

It helps you quickly diagnose why a Trino cluster is slow, unhealthy, or failing by surfacing coordinator and worker status, query states, memory pressure, and operational signals from REST, JMX/metrics, logs, and the Web UI.

Core Features & Use Cases

• Cluster health checks via REST: Validate coordinator readiness, enumerate nodes, inspect cluster-wide running/queued/blocked query counts, and view or terminate problematic queries.
• Memory and stability visibility via JMX metrics: Export key runtime and OOM-related counters to Prometheus and use alerting to catch worker instability and resource pressure early.
• Operational diagnosis from query state and logs: Identify BLOCKED/QUEUED conditions, correlate them with OOM kills and worker loss patterns, and perform safe rolling shutdowns.

Use case: you observe unexpected queuing and intermittent failures during peak load, so you check cluster endpoints, confirm blocked/running/query queue depth, verify OOM kills or free-memory drops, then determine whether a graceful worker shutdown or configuration change (resource/memory limits, spill behavior) is needed.

Quick Start

Use the trino-admin-cluster-health skill to diagnose a slow Trino cluster by checking /v1/info, /v1/cluster, /v1/node, and inspecting blocked or long-running queries, metrics, and relevant server logs.