rpg.pbem.online

Image/Photo
If you are a #podmin running a #diaspora pod on PostgreSQL, you should look into upgrading to version 13. These are runserver response times for Geraspora, the yellow line represents the average response times, the green line represents the the 99.9 percentile. The blue marker is the point I upgraded from pg12 to pg13. There was no other change to the environment, and since we've had a weekend just now, the amount of requests per second and the server load even went up slightly.

As you'll notice, the peaks went down significantly, but even the average response time went down by a considerable amount. Parallel autovacuuming and the performance improvements around b-tree indexes and aggregate/grouping queries is clearly help stream queries a lot. Not enough to significantly change the responsiveness for users, but still noteworthy.

(Ignore that one peak in the night of 2020-12-11 to 2020-12-12, that was a unlucky coincidence with autovac running while the server was running its "aw crap an entirely new database"-backup, so that always causes some locking issues.)
On my own pod the stream query times dropped by about 50% (or about 1.5 seconds). PG13 is an incredible upgrade.

And while we're on the subject of things podmins should look into, consider using WAL-G for continuous, encrypted, point-in-time recovery (PITR) backups to S3-compatible storage. 😀 It's fast, secure, and reliable.
I will try it this evening or tomorrow morning.

Perhaps it can speed up my Matrix server as well