Re: Intermittent buildfarm failures on wrasse

Hi,

On April 15, 2022 2:14:47 PM EDT, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>Andres Freund <andres(at)anarazel(dot)de> writes:
>> On 2022-04-15 12:36:52 -0400, Tom Lane wrote:
>>> Yeah, I was also thinking about a flag in PGPROC being a more reliable
>>> way to do this. Is there anything besides walsenders that should set
>>> that flag?
>
>> Not that I can think of. It's only because of hs_feedback that we need
>> to. I guess it's possible that somebody build some extension that needs
>> something similar, but then they'd need to set that flag...
>
>Here's a WIP patch for that. The only exciting thing in it is that
>because of some undocumented cowboy programming in walsender.c, the
> Assert((proc->statusFlags & (~PROC_COPYABLE_FLAGS)) == 0);
>in ProcArrayInstallRestoredXmin fires unless we skip that.
>
>I could use some help filling in the XXX comments, because it's far
>from clear to me *why* walsenders need this to happen.

I'm out for the rest of the day due to family events (visiting my girlfriend's parents till Wednesday), I can take a stab at formulating something after.

If you want to commit before: The reason is that walsenders use their xmin to represent the xmin of standbys when using hot_standby_feedback. Since we're only transmitting global horizons up from standbys, it has to influence globally (and it would be hard to represent per db horizons anyway).

Andres
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.