"idle in transaction" postgres connection
2011-08-18 09:34:25 GMT
I am using postgres-8.4.0 on a cluster setup with slony-2.0.4 being used for replication.
Recently , I saw a "idle in transaction" postgres process as below.
postgres 13052 14742 0 May13 ? 00:00:00 postgres: slon abc 172.16.1.1(49017) idle in transaction
I wonder what could have lead to that hung postgres process . I googled about it a lot and they say that it could be
because of abrupt netwotk issue between slony and postgres . But in my case slon was connected
to its local postgres database. So, network wont be an issue in this case . Restarting the slon/postgres resolves
the issue. But, there are other processes also in my application which connect to postgres using psqlODBC . Those
processes can also cause "idle in transaction" postgres connection . I can't afford restarting any of the processes
to resolve this issue. So, I was wondering if I can make use of tcp_keepalives_idle , tcp_keepalives_interval and
tcp_keepalives_count configurations in postgresql.conf to get rid of this "idle in transaction" processes
after a certain amount of time . Will this help or not ??