On Tue, Jan 18, 2022 at 4:32 PM Andres Freund <andres@anarazel.de> wrote:
>
> I wonder if a very different approach could make sense here. Presumably this
> wouldn't need to be queried at a very high frequency, right? If so, what about
> storing the latest commit LSN for each backend in PGPROC? That could be
> maintained without a lock/atomics, and should be just about free.
> pg_last_committed_xact() then would have to iterate over all PGPROCs to
> complete the LSN, but that's not too bad for an operation like that. We'd also
> need to maintain a value for all disconnected backends, but that's also not a hot
> path.
Is something roughly like the attached what you'd envisioned? I
wouldn't expect the final implementation to be in commit_ts.c, but I
left it there for expediency's sake in demonstrating the idea since
pg_last_committed_xact() currently finds its home there.
I think we need a shared ProcArrayLock to read the array, correct? We
also need to do the global updating under lock, but given it's when a
proc is removed, that shouldn't be a performance issue if I'm
following what you are saying.
Thanks,
James Coleman