Re: PG12 autovac issues

Hi,

On 2020-03-17 17:18:57 -0500, Justin King wrote:
> As you can see in this table, there are only ~80K rows, but billions
> of updates. What we have observed is that the frozenxid reaches the
> 200M mark fairly quickly because of the amount of activity.

And each of those updates is in a separate transaction? Is that
required? I.e. any chance to perform multiple of those updates in one
transaction?

Have you considered just increasing the vacuum limit? It's mostly there
because it can increase space usage a bit, but given today's systems its
not a usually a problem unless you have hundreds of postgres instances
on a single system. It's pretty common to just increase that to 1.8
billion or so.

From a single stats snapshot we can't actually understand the actual xid
consumption - is it actually the xid usage that triggers the vacuums?

What makes you think it is a problem that you have all these vacuums? If
you actually update that much, and you have indexes, you're going want a
lot of vacuums?

> What is interesting is that this happens with the 'postgres' and
> 'template1' databases as well and there is absolutely no activity in
> those databases.

That's normal. They should be pretty darn quick in v12?

Greetings,

Andres Freund