Re: shared memory based stat collector (was: Sharing record typmods between backends)

On 2017-08-14 13:06:48 -0400, Robert Haas wrote:
> On Mon, Aug 14, 2017 at 1:02 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> > I previously thought that an option to occasionally WAL log the stats
> > file would be useful (e.g. just before a checkpoint). That'd make them
> > persistent, and available on the standby. But that'd still require
> > somehow dealing with stats being produced on the standby - I presume
> > we'd need multiple stats files and provide functions for merging them.
>
> Well, if you think the stats files might be too big to fit in memory,
> then you're probably imagining tens of gigabytes of data, and you're
> not going to want to write a WAL record that size, either.

Well, that's why I'm thinking of it having to be an option. You
presumably don't want it on relatively idle servers, but if you have big
databases where unnecessary vacuums are bad...

I think again, this all'd be better if we'd figure out a way to make
this use a proper table... We kind of have figured out how to store data
in a queryable manner, with buffering etc. Kind of. Wonder if there's
some chance of a hack where we have a shared table with a key like
(dboid, reloid, inreplication). If we were to create those (once for
replication, once for not) at table creation, we could
heap_inplace_update them even on a standby...

- Andres