Horton

Re: Multiple hosts in connection string failed to failover in non-hot standby mode

Andreas Seltenreich <seltenreich@gmx.de>

View thread View on postgr.es
From: Andreas Seltenreich <seltenreich@gmx.de>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Hubert Zhang <zhubert@vmware.com>, tsunakawa.takay@fujitsu.com, pgsql-hackers@postgresql.org, Bruce Momjian <bruce@momjian.us>, Michael Meskes <michael.meskes@credativ.de>, Robert Treat <robert.treat@credativ.us>
Date: 2021-05-06T20:24:21Z
Lists: pgsql-hackers 
> On Thu, May 06, 2021 at 01:22:27PM -0400, Tom Lane wrote:
>> I'm curious though why it took this long for anyone to complain.
>> I'd supposed that people were running sqlsmith against HEAD on
>> a pretty regular basis.

Last time I ran it was November 27.  I'm neglecting it on my spare time
and there is hardly any opportunity to sneak it onto my agenda at work.
I'll do my best to try to get either of these fixed.

Justin Pryzby writes:
> I think it's also becase sqlsmith would need to run against the v14 *client*
> library.  I don't know about anyone else's workflow, but I tend not to "make
> install", but work with binaries out of ./tmp_install.

My playbooks don't grab the client libraries of the test target either.
I'll change them.

> There's a few changes needed on sqlsmith HEAD, but I guess nobody would have
> gotten that far if the connection was failing (or rather, detected as such).

Thanks for the patch.

regards,
andreas

Commits

  1. Clear conn->errorMessage at successful completion of PQconnectdb().

  2. Avoid ECPG test failures in some GSS-capable environments.

  3. Try next host after a "cannot connect now" failure.

  4. Uniformly identify the target host in libpq connection failure reports.

  5. Allow pg_regress.c wrappers to postprocess test result files.

  6. In libpq, always append new error messages to conn->errorMessage.