Hi,
On 2023-04-04 17:33:25 -0700, Jeff Davis wrote:
> On Tue, 2023-04-04 at 14:55 -0400, Robert Haas wrote:
> > That's presumably OK, in the
> > sense that they'll go back to sleep and eventually wake up again, but
> > it means they might end up chronically behind sending out WAL to
> > cascading standbys.
>
> Without 0004, cascading logical walsenders would have worse wakeup
> behavior than logical walsenders on the primary. Assuming the fix is
> small in scope and otherwise acceptable, I think it belongs as a part
> of this overall series.
FWIW, personally, I wouldn't feel ok with committing 0003 without 0004. And
IMO they ought to be committed the other way round. The stalls you *can* get,
depending on the speed of WAL apply and OS scheduling, can be long.
This is actually why a predecessor version of the feature had a bunch of
sleeps and retries in the tests, just to avoid those stalls. Obviously that's
not a good path...
Greetings,
Andres Freund