| From: | Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> |
|---|---|
| To: | Peter Smith <smithpb2250(at)gmail(dot)com> |
| Cc: | Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>, Andres Freund <andres(at)anarazel(dot)de>, Abbas Butt <abbas(dot)butt(at)enterprisedb(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Zahid Iqbal <zahid(dot)iqbal(at)enterprisedb(dot)com> |
| Subject: | Re: Logical replication keepalive flood |
| Date: | 2021-09-17 05:05:27 |
| Message-ID: | CAA4eK1JcFVay-O3PZnBFJMomYozJSiq1vzzYZQtTDE91JTH92w@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Thu, Aug 12, 2021 at 8:03 AM Peter Smith <smithpb2250(at)gmail(dot)com> wrote:
>
> That base data is showing there are similar numbers of keepalives sent
> as there are calls made to WalSndWaitForWal. IIUC it means that mostly
> the loop is sending the special keepalives on the *first* iteration,
> but by the time of the *second* iteration the ProcessRepliesIfAny()
> will have some status already received, and so mostly sending another
> keepalive will be deemed unnecessary.
>
> Based on this, our idea was to simply skip sending the
> WalSndKeepalive(false) for the FIRST iteration of the loop only! PSA
> the patch 0002 which does this skip.
>
I think we should also keep in mind that there are cases where it
seems we are not able to send keep-alive at the appropriate frequency.
See the discussion [1]. This is to ensure that we shouldn't
unintentionally hamper some other workloads by fixing the workload
being discussed here.
[1] - https://www.postgresql.org/message-id/20210913.103107.813489310351696839.horikyota.ntt%40gmail.com
--
With Regards,
Amit Kapila.
| From | Date | Subject | |
|---|---|---|---|
| Next Message | houzj.fnst@fujitsu.com | 2021-09-17 05:08:36 | RE: Column Filtering in Logical Replication |
| Previous Message | Amit Kapila | 2021-09-17 04:48:11 | Re: walsender timeout on logical replication set |