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

On Tue, Dec 13, 2016 at 11:40:54AM -0500, Robert Haas wrote:
> Let's confine ourselves to fixing one problem at a time. I think we
> can get where we want to be in this case by adding one new column and
> some new rows to pg_stat_activity.

Agreed. Let's also remove the abuse of WAL senders with the query field
at the same time.

> Michael, is that something you're
> going to do? If not, one of my esteemed colleagues here at
> EnterpriseDB will have a try.

If I had received feedback on the other thread, I would have coded a
proposal of patch already. But as long as SCRAM is not done I will
restrain from taking an extra project. I am fine to do reviews as I already
looked at ways to solve the problem though. So if anybody has room to do
it please be my guest.

Regarding the way to solve things, I think that having in ProcGlobal an
array of PGPROC entries for each auxiliary process is the way to go, the
position of each entry in the array defining what the process type is.
That would waste some shared memory, but we are not talking about that
much here. That would as well remove the need of having checkpointerLatch,
walWriteLatch and the startup fields in ProcGlobal.
--
Michael