Postgresql low throughput, high server CPU after changes, CDN latency spike

Hey there!

First, let’s tackle your specific questions:

1. Rows Fetched vs Rows Read: Think of Rows Read as all the data your database looks at, while Rows Fetched is what it actually sends back to your app. It’s like scanning a whole book (Rows Read) but only copying a few pages (Rows Fetched).

2. Is a higher Rows Fetched to Rows Read ratio better? Generally, yes! It means your queries are more efficient, like finding exactly what you need in that book without flipping through every page.

3. Best practices for database fetching:

   • Use indexes wisely (like a book’s table of contents)
   • Only ask for the data you need (avoid SELECT *)
   • Use LIMIT to keep result sets manageable
   • Try pagination for large data sets
   • Keep optimizing those queries!

4. Is it the server or the database? Good question! To figure this out:

   • Keep an eye on both server and database metrics
   • Use query logging to spot slow queries
   • Check database wait events for bottlenecks
   • Monitor the network between your server and database

Now, looking at your metrics, that high cache hit rate and low load average suggest your database might not be the main culprit. The high server CPU usage points more towards an application issue.

Here’s what I’d suggest:

1. Take a close look at your recent changes, especially around database queries.
2. Hunt for any queries that might be running too often or inefficiently.
3. Check for new CPU-hungry tasks on your server.
4. Scan your app logs for anything unusual.
5. Give DigitalOcean’s built-in monitoring tools a shot, or try something simple like Prometheus with node_exporter:

https://docs.digitalocean.com/products/databases/postgresql/how-to/monitor-databases/

If you’re still stuck, don’t be afraid to roll back some changes temporarily. It might help you pinpoint what’s causing the slowdown.

- Bobby