Postgres Autovacuum Progress Monitoring

Question

I have an autovacuum (to prevent wraparound) happening on one of my tables for almost 4 days. I'm wondering if there is any way for me to tell if it's not keeping up with the amount of updates/deletes that are happening.

For reference, I am using Postgres 9.6 on Google CloudSQL. They have a known issue where pg_stat_progress_vacuum returns null for everything.

I've also been checking pg_stat_all_tables and have noticed that n_dead_tup for my table is going up instead of down, but I'm not sure if this number would be aware of the autovacuuming that is going on.

Answer 1

Bad luck. If you have a hosted database, and the monitoring view is broken, there's nothing you can do.

To speed up autovacuum as much as possible, increase maintenance_work_mem and lower autovacuum_vacuum_cost_delay. But that won't have an effect on an already running autovacuum worker.