OPS-1842 · Incident Ops

Primary saturation on orders-db

SEV-2 · Waiting

Prometheus

orders_db_cpu above 92% for 14 minutes. Connections 480 of 500. Read replica is quiet.

Cause

Checkout retried after a payments timeout and pinned the primary. The write path is the primary. The replica is healthy.

Plan

Restart the primary in the allow-listed window. Do not fail over. Do not touch payments-db.

Draft

Restart orders-db primary

Allow-list · restart-statefulset · Jira OPS-1842

Footprint

Person gate

The write waits here.

Restart orders-db primary is on the allow-list (restart-statefulset). It does not run until you decide.