flink-statement-triage

name	flink-statement-triage
description	Troubleshoot Confluent Cloud Flink SQL statements using flink-triage CLI, metrics, and exceptions. Use when a statement is FAILED, PENDING, has backpressure, zero output, or high state size.

Flink statement triage

Scope

Confluent Cloud for Flink only. No local or K8s Flink runtime.

When to use

User reports CC Flink statement failure, stuck deployment, lag, or asks to analyze a named statement.

Required inputs

• statement_name (e.g. perf-dml-passthrough)
• compute_pool_id (e.g. lfcp-xxxxx)

Workflow

1. Step 1 — ./assets/scripts/step1_verify.sh (offline, no cluster)
2. Step 2 — ./assets/scripts/step2_deploy_cc.sh then live triage

See docs/DEPLOYMENT.md.

Triage commands

uv run flink-triage run --statement NAME --pool POOL
uv run flink-triage run -s perf-dml-passthrough -p lfcp-example --dry-run  # Step 1 only

Confluent MCP: detect-flink-statement-issues, get-flink-statement-profile.

References

• docs/DEPLOYMENT.md
• docs/RUNBOOK.md