On-Call Agent: Postgres Saturation Webhook to Query Culprit Mitigation Plan

Handles database saturation incidents specifically. When connections or load spike, the agent looks directly at what the database is doing — long-running and blocking queries, lock waits, connection counts — and proposes targeted mitigations rather than generic restarts.

Use it when your most painful incidents are database pressure events and you need to know *which* query or client is the problem before deciding to kill a session, add an index, or throttle a caller. Read-only inspection plus human-approved action.

1. 1A monitoring system posts a webhook when Postgres connection or load saturation crosses threshold.
2. 2The agent runs read-only inspection queries against Postgres for active statements, slow queries, lock contention, and connection counts.
3. 3Logic ranks the culprits by impact and matches each to a candidate mitigation — terminate a runaway session, add a missing index, or rate-limit a noisy client.
4. 4It assembles a mitigation plan with the supporting evidence for each recommendation.
5. 5The agent posts the plan to Slack with each action gated behind approval; nothing is terminated or changed until an operator signs off.