We're running perfsonar on a group of our servers, and while doing some maintenance work found that we
- get giant logfiles under
- table bloat on the database of the perfsonar scheduler
I assume the toolkit is using table locking when it does ANYTHING on the database, and always cancels the autovacuum run:
2019-04-07 06:33:43 NZST [471-1] ERROR: canceling autovacuum task 2019-04-07 06:33:43 NZST [471-2] CONTEXT: automatic vacuum of table "pscheduler.public.run" 2019-04-07 06:34:33 NZST [981-1] ERROR: canceling autovacuum task 2019-04-07 06:34:33 NZST [981-2] CONTEXT: automatic vacuum of table "pscheduler.public.run"
A manual run of
vacuum analyze roughly halved the use of disk space under
I've found several web pages talking about autovacuum tuning, but I'm wondering if it wouldn't make sense to disable autovacuum and run a
vacuum analyze from a cron job.
Is there a best practice/recommendation for this kind of situation?
And maybe ask the perfsonar project to clean up their code in the long run ;}