[check_postgres] pgbouncer pooler/backend

[Cédric Villemain]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Greg Sabino Mullane a écrit :
>> The idea is to keep the existent plugins and make them workable with pgbouncer
>> (if it is relevant).
>>
>> Many plugins are irrelevant for pgbouncer (last_analyze for example), and some
>> plugin lacks if we want to check_pgbouncer.
>>
>>
>> So I wonder if it is better to do a check_pgbouncer.pl or make check_postgres.pl
>> understood it can use a normal postgresql server or a modified one (then issue
>> relevant query for this kind of modified server) ?
> 
> I think pgbouncer is a fairly unique case: it should be 100% transparent to all
> current queries, right? We should make separate actions for pgbouncer
> information (or Slony/Bucardo/etc.) This should be a very small number of things
> to check for, right? Perhaps something like this
> 
> ./check_postgres_bouncer_clients --warning=20
> 
> We'd know to use the default pgbouncer port for these checks, of course. If I
> get time, I'll whip up a proof of concept, but anyone feel free to take a swipe
> at it, it's unlikely to happen before PostgreSQL Conference East, at any rate.
> 

Well I did some progress in that way, and I quickly find that some basic routine
have to be rewrite to pgbouncer (especialy, pgbouncer db can not be SELECTed,
only SHOW FOO).

It will perhaps be better to have a check_pgbouncer.pl and not trying to make
check_postgres an industry. (and sure check_pgbouncer can steal the most common
and usefull function from check_postgres.pl)


- --
Cédric Villemain
Administrateur de Base de Données
Cel: +33 (0)6 74 15 56 53
http://dalibo.com - http://dalibo.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAknM9rUACgkQo/dppWjpEvyIEwCg1hcnqFXE5iQUjYULHehRtoZ5
k/gAn0pkirWLLH0WAkmBGGPJWbQj6dMh
=RJG5
-----END PGP SIGNATURE-----