| From: | Nathan Bossart <nathandbossart(at)gmail(dot)com> |
|---|---|
| To: | Thomas Munro <thomas(dot)munro(at)gmail(dot)com> |
| Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>, alvherre(at)alvh(dot)no-ip(dot)org, bharath(dot)rupireddyforpostgres(at)gmail(dot)com, pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Suppressing useless wakeups in walreceiver |
| Date: | 2022-11-14 23:14:39 |
| Message-ID: | 20221114231439.GA1470047@nathanxps13 |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Tue, Nov 15, 2022 at 09:42:26AM +1300, Thomas Munro wrote:
> That works for 020_pg_receivewal. I wonder if there are also tests
> that stream a bit of WAL first and then do wait_for_catchup that were
> previously benefiting from the 100ms-after-startup message by
> scheduling luck (as in, that was usually enough for replay)? I might
> go and teach Cluster.pm to log how much time is wasted in
> wait_for_catchup to get some observability, and then try to figure out
> how to optimise it properly. We could perhaps put the 100ms duct tape
> back temporarily though, if necessary.
Oh, I see. Since we don't check the apply position when determining
whether to send a reply, tests may need to wait a full
wal_receiver_status_interval. FWIW with my patch, the runtime of the
src/test/recovery tests seems to be back to what it was on my machine, but
I certainly wouldn't rule out scheduling luck.
--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Justin Pryzby | 2022-11-14 23:16:46 | Re: [RFC] building postgres with meson - v12 |
| Previous Message | Peter Geoghegan | 2022-11-14 23:07:05 | Re: HOT chain validation in verify_heapam() |