Re: Libpq support to connect to standby server as priority

Tom Lane wrote:
> Laurenz Albe <laurenz(dot)albe(at)cybertec(dot)at> writes:
> > Haribabu Kommi wrote:
> > > Added comments along the lines that you mentioned. And also try
> > > to update some more comments.
> > Looks ok to me, I'll mark it as "ready for committer".
>
> I don't like this patch at all: the business with keeping two connections
> open seems impossibly fragile and full of race conditions. (For instance,
> by the time you return the read-only session to the application, it might
> not be read-only any more. I also wonder what inquiry functions like
> PQsocket ought to return while in this state.) I think the feature
> definition needs to be re-thought to make that unnecessary.

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.

If the read-only connection turns writable after it has been tested,
but before it is returned, that can hardly be avoided.
I don't think that's so bad - after all, you asked for a read-only
connection *if possible*.
If you demand that the server be not promoted until the connection has
been returned to the client, you'd somehow have to block the server
from being promoted, right?

> Also, we really need to consider the interaction between this and the
> feature(s) being discussed in the thread at
>
> https://www.postgresql.org/message-id/flat/1700970.cRWpxnom9y%40hammer.magicstack.net

That's a good point.

Yours,
Laurenz Albe