Horton

Re: Libpq support to connect to standby server as priority

Laurenz Albe <laurenz.albe@cybertec.at>

View thread View on postgr.es
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Haribabu Kommi <kommi.haribabu@gmail.com>, Robert Haas <robertmhaas@gmail.com>, Jing Wang <jingwangian@gmail.com>, Pg Hackers <pgsql-hackers@postgresql.org>, "Tsunakawa, Takayuki" <tsunakawa.takay@jp.fujitsu.com>
Date: 2018-11-16T17:56:11Z
Lists: pgsql-hackers 
Tom Lane wrote:
> > As it is now, the patch doesn't keep two connections open.  It remembers
> > the index of the host of the first successful writable connection, but
> > closes the connection, and opens another one to that host if no read-only
> > host can be found.
> 
> Oh!  The reason I assumed it wasn't doing that is that such a behavior
> seems completely insane.  If the point is to keep down the load on your
> master server, then connecting only to immediately disconnect is not
> a friendly way to do that --- even without counting the fact that you
> might later come back and connect again.

That's why I had argued initially to keep the session open, but you
seem to dislike that idea as well.

> If that's the best we can do, we should forget the whole feature and
> just recommend putting slave servers first in your hosts list when
> you want prefer-slave.

If you know which is which, certainly.
But in a setup with automated failover you cannot be certain which is which.
That's what the proposed feature targets.

Yours,
Laurenz Albe

Commits

  1. Make test_target_session_attrs more robust against connection failure.

  2. Extend the abilities of libpq's target_session_attrs parameter.

  3. Mark default_transaction_read_only as GUC_REPORT.

  4. Introduce a new GUC_REPORT setting "in_hot_standby".

  5. Avoid spamming the client with multiple ParameterStatus messages.

  6. Avoid harmless Valgrind no-buffer-pin errors.

  7. Restructure libpq code to remove some duplicity

  8. libpq: Add target_session_attrs parameter.