| From: | Jeff Davis <pgsql(at)j-davis(dot)com> |
|---|---|
| To: | pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Replication protocol pipelining edge case |
| Date: | 2021-01-05 23:23:34 |
| Message-ID: | 2c686eeeff7049e7500303248502aaebd876c649.camel@j-davis.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Tue, 2021-01-05 at 09:17 -0800, Jeff Davis wrote:
> The reason I ask is because it makes clients that do pipelining more
> complex. If the server always enters copy mode, then the client can
> send a CopyDone and then pipeline more requests behind it. But if the
> server may or may not enter copy mode, then the client needs to wait
> until it sees a CopyBothResponse before any new requests can be
> pipelined.
In an offline conversation with Andres, he pointed out that we can
never be sure that we'll enter COPY mode, because the START_REPLICATION
command could throw an error. Therefore we need to always wait for the
CopyBothResponse, and can't pipeline any new commands until we get it.
Regards,
Jeff Davis
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Craig Ringer | 2021-01-05 23:30:37 | Re: [PATCH] Runtime control of CLOBBER_CACHE_ALWAYS |
| Previous Message | Peter Smith | 2021-01-05 23:02:11 | Re: Single transaction in the tablesync worker? |