Exercise: The Postgres Primary at Its Limit

1. The Page Fires at 02:14

You are on call. The primary Postgres behind your main application is struggling at peak, and three people have already proposed a fix in the incident channel.

Here is everything the dashboard tells you.

• Machine: 16 vCPU, 64 GB RAM, one primary, no replicas.
• CPU: 95 percent, sustained, for the roughly ninety minutes a day that peak lasts.
• Latency: p99 on reads is 180 ms, against a normal 55 ms. Writes look unchanged.
• Database size: 340 GB, growing about 8 GB a month.