Re: exposing wait events for non-backends (was: Tracking wait event for latches)

On Mon, Dec 12, 2016 at 9:19 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

> So, one of the problems in this patch as committed is that for any
> process that doesn't show up in pg_stat_activity, there's no way to
> see the wait event information. That sucks. I think there are
> basically two ways to fix this:
>
> 1. Show all processes that have a PGPROC in pg_stat_activity,
> including auxiliary processes and whatnot, and use some new field in
> pg_stat_activity to indicate the process type.
>
> 2. Add a second view, say pg_stat_system_activity, to show the
> processes that don't appear in pg_stat_activity. A bunch of columns
> could likely be omitted, but there would be some duplication, too.
>
> Preferences?
>
>
​I'm inclined toward option 2.

A view over both that involves just the shared columns would give you the
benefits from option 1.

Question: is there a parent-child relationship involved here? Given a
record in today's pg_stat_activity is it useful/possible to link in all of
the pg_stat_system_activity​ process that are working to fulfill the
client-initiated task?

Even with "system" we'd probably want to distinguish between background
workers and true system maintenance processes. Or am I mis-interpreting
the scope of this feature?

David J.