[pgpool-general: 1766] Re: Persistent "Idle in transaction" connections
ishii at postgresql.org
Wed May 22 19:40:00 JST 2013
No, connection_life_time is for between pgpool and PostgreSQL, not
client and pgpool. So client_idle_limit is the right choice. However
we recently found a problem which might be related to Juan's problem.
If there's an activity found in pg_stat_activity as descrived in the
commit message, you might want to try the patch.
SRA OSS, Inc. Japan
> Hi Juan,
> If I understand correctly I think you want pgpools "connection_life_time" parameter, and if you're still having problems with dropped connections look at your kernels tcp_keepalive parameters.
> ----- Original Message -----
>> From: Juan Jose Perez <jperez at oteara.com>
>> To: pgpool-general at pgpool.net
>> Sent: Wednesday, 22 May 2013, 10:42
>> Subject: [pgpool-general: 1764] Persistent "Idle in transaction" connections
>> Hi all,
>> Sorry if this question is answered somewhere but I couldn't find it.
>> We have a client application using pgpool 3.1.2 and postgres 9.1 in
>> master-slave with streaming replication.
>> This application is unreliable so it can be shutdown at any time.
>> When this happens, its connection with pgpool remains in "idle in
>> transaction" state forever. To avoid this I changed client_idle_limit
>> parameter to 600 (this value should be enough to free broken connections
>> before pgpool collapses), but it isn't closing those connections.
>> Is client_idle_limit the appropriate parameter to control
>> connections in this situation or there is another way to protect pgpool
>> from this client behavior?
>> Many thanks,
>> Juanjo Pérez
>> pgpool-general mailing list
>> pgpool-general at pgpool.net
> pgpool-general mailing list
> pgpool-general at pgpool.net
More information about the pgpool-general