Re: [GENERAL] Strange performance degradation - Mailing list pgsql-performance

From Lorenzo Allegrucci
Subject Re: [GENERAL] Strange performance degradation
Date
Msg-id 4B0AF4B1.8040803@forinicom.it
Whole thread Raw
In response to Re: [GENERAL] Strange performance degradation  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: [GENERAL] Strange performance degradation
List pgsql-performance
Tom Lane wrote:
> Lorenzo Allegrucci <lorenzo.allegrucci@forinicom.it> writes:
>> So, my main question is.. how can just a plain simple restart of postgres
>> restore the original performance (3% cpu time)?
>
> Are you killing off any long-running transactions when you restart?

After three days of patient waiting it looks like the common
'<IDLE> in transaction' problem..

[sorry for >80 cols]

19329 ?        S     15:54 /usr/lib/postgresql/8.3/bin/postgres -D /var/lib/postgresql/8.3/main -c
config_file=/etc/postgresql/8.3/main/postgresql.conf
19331 ?        Ss     3:40  \_ postgres: writer process
19332 ?        Ss     0:42  \_ postgres: wal writer process
19333 ?        Ss    15:01  \_ postgres: stats collector process
19586 ?        Ss   114:00  \_ postgres: forinicom weadmin [local] idle
20058 ?        Ss     0:00  \_ postgres: forinicom weadmin [local] idle
13136 ?        Ss     0:00  \_ postgres: forinicom weadmin 192.168.4.253(43721) idle in transaction

My app is a Django webapp, maybe there's some bug in the Django+psycopg2 stack?

Anyway, how can I get rid those "idle in transaction" processes?
Can I just kill -15 them or is there a less drastic way to do it?

pgsql-performance by date:

Previous
From: Dave Youatt
Date:
Subject: Re: Performance degrade running on multicore computer
Next
From: Jason Dictos
Date:
Subject: Best possible way to insert and get returned ids