[Pgpool-general] NOTIFY/LISTEN forwarding?

[Tatsuo Ishii]

> I'd really like to get pgpool going, but I'm having another problem with it. 
> After connecting about 190 clients to it, clients start getting disconnected. 
> I've set the following parameters:
> 
> num_init_children = 32
> max_pool = 64
> child_life_time = 0
> connection_life_time = 0
> 
> Every few seconds, a few clients will get kicked off pgpool. Resources don't 
> seem to be cleaned up on the Postgres server when this happens, either, so it 
> looks like a "bad" event, not a "planned" event (like a timeout).

Could I take a look at debug print from pgpool? You need to add -d -n
option to pgpool to enable debugging output.
--
Tatsuo Ishii

> I'm not using replication or failover. I set the above parameters based on the 
> following assumptions, please let me know if these are wrong:
> 
> 1. Each child can have more than one client connected. I would have about 1200 
> clients per box, and I would expect no more than 20 of them to be talking at 
> once, on average. I figured 32 children, spreading the client load evenly 
> among them, would be ideal. Each server would ideally support 800-1200 clients 
> on average, up to around 1600, so this would leave some margin for growth 
> (2048 max clients).
> 
> 2. My clients are well-behaved about closing connections, so I don't need the 
> timeouts, and they're undesirable since my clients may have a great deal of 
> idle time in between queries.
> 
> What could I be doing wrong? Is there any other information I can provide? 
> pgpool 2.4 and Postgres 8.0.1. If it matters, I can successfully connect all 
> thousand clients directly to the Postgres backend, I'm just trying to reduce 
> resource usage there since 8000+ clients requires a ridiculous amount of RAM.
> 
> Regards,
> Chad
> _______________________________________________
> Pgpool-general mailing list
> Pgpool-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgpool-general
>