Re: Why is my table not autovacuuming?

With a scale factor of 0.2, autovacuum won't kick in until there are an
estimated 0.2 * 71M = 14.2M dead tuples. For such large tables, it may be
prudent to adjust the autovacuum parameters to have a lower scale factor.
You can also set the scale factor to zero, and then the threshold alone
determines when autovacuum runs: you'd set the threshold to some number
much larger than the default 50.

https://blog.2ndquadrant.com/autovacuum-tuning-basics/
https://www.postgresql.org/docs/9.6/static/runtime-config-autovacuum.html

On Mon, Aug 28, 2017 at 11:45 AM Don Seiler <don(at)seiler(dot)us> wrote:

> PostgreSQL 9.2.18. I have a table with 71M rows (via select count). In
> pg_stat_activity, n_live_tups is 170K, n_dead_tups is slightly more than
> that (203K). The autovacuum parameter is on, the scale factor is 0.2
> (default) and threshold is 50 (default), autovacuum_max_workers is 3. There
> are no custom autovacuum thresholds set for this table. The last_autovacuum
> field is null.
>
> Obviously I would expect autovacuum/autoanalyze to kick in for this table
> so the stats aren't getting so far out of whack from reality. When I query
> pg_stat_activity, I never see any autovacuum jobs running, so it isn't a
> matter of too much work from what I can tell.
>
> Previously I had tables not being autovacuumed because of index corruption
> (which would also cause manual vacuum to fail), but I was just able to
> successfully vacuum this table manually and now pg_stat_activity show the
> accurate info.
>
> Wondering why my autovacuum daemon appears to be acting lazily?
>
> Don.
>
> --
> Don Seiler
> www.seiler.us
>