[pgpool-general: 3193] Re: zombie connections
Tatsuo Ishii
ishii at postgresql.org
Thu Sep 25 08:46:25 JST 2014
I think your issue is related to this commit and fixed:
http://git.postgresql.org/gitweb/?p=pgpool2.git;a=commit;h=cb735f22441001b6afdbb5bac72808db66094ca9
Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp
> Thanks!
>
> I will try. Can you just tell me if what I describe is similar to known
> issues with previous versions? I was not able to find posts on this.
>
> On Wed, Sep 24, 2014 at 8:00 AM, Tatsuo Ishii <ishii at postgresql.org> wrote:
>
>> Please try 3.3.4. If you still have a problem, please let us know.
>>
>> Best regards,
>> --
>> Tatsuo Ishii
>> SRA OSS, Inc. Japan
>> English: http://www.sraoss.co.jp/index_en.php
>> Japanese:http://www.sraoss.co.jp
>>
>> > Hi
>> >
>> > We are struggling with the bellow issues for a long time - any help, even
>> > partial, is highly appreciated!
>> >
>> > We use pgpool-II 3.3.3 with 4 Postrges 9.3 nodes. We are experiencing the
>> > following issues:
>> >
>> > 1. In the db servers I see many idle connections (processes) for a long
>> > time (even days). "show pool_pools" does not show those connections so
>> > pgpool is not aware of them anymore. All connections to DB come from
>> > pg-pool so possibly it does not close the connection before killing the
>> > child process
>> >
>> > 2. In "show pool_pools" I see many idle connections far behind their
>> > connection_life_time and even behind their pool child_life_time. The
>> system
>> > was inactive for a few hours so I know for sure these connections are not
>> > in use.
>> >
>> > 3. In "show pool_pools" I sometimes see multiple connections within the
>> > same pool with pool_connected=1. Also, I know for a fact that those
>> client
>> > processes were killed a long time ago.
>> >
>> >
>> > Here is our config:
>> >
>> > num_init_children=400
>> > max_pool=5
>> > child_life_time=300
>> > child_max_connections=0
>> > connection_life_time=60
>> > client_idle_limit=300
>> >
>> > (we have 4 backends)
>> >
>> >
>> > This is example of multiple connections in the same child with
>> > pool_connected=1 (how can it be?).
>> > Also those connections are already not active for many hours (inactive
>> > zombie connection also happens when pool_connected=0)
>> >
>> > pool_pid | start_time | pool_id | backend_id | database |
>> > username | create_time | majorversion | minorversion |
>> > pool_counter | pool_backendpid | pool_connected
>> > 4741 | 2014-09-23 06:44:46 | 0 | 1 | DB1 |
>> > user1 | 2014-09-23 07:15:58 | 3 | 0 | 7
>> > | 6911 | 1
>> > 4741 | 2014-09-23 06:44:46 | 0 | 2 | DB1 |
>> > user1 | 2014-09-23 07:15:58 | 3 | 0 | 7
>> > | 22316 | 1
>> > 4741 | 2014-09-23 06:44:46 | 0 | 3 | DB1 |
>> > user1 | 2014-09-23 07:15:58 | 3 | 0 | 7
>> > | 27441 | 1
>> >
>> >
>> > Any idea about what can cause this is highly valuable for us (even
>> partial
>> > idea).
>> >
>> > Thanks,
>> > Yaron
>>
>
>
>
> --
>
> I'm on Twitter (@YaronNaveh <http://twitter.com/#!/YaronNaveh>)
> Thanks!
>
> I will try. Can you just tell me if what I describe is similar to known
> issues with previous versions? I was not able to find posts on this.
>
> On Wed, Sep 24, 2014 at 8:00 AM, Tatsuo Ishii <ishii at postgresql.org> wrote:
>
>> Please try 3.3.4. If you still have a problem, please let us know.
>>
>> Best regards,
>> --
>> Tatsuo Ishii
>> SRA OSS, Inc. Japan
>> English: http://www.sraoss.co.jp/index_en.php
>> Japanese:http://www.sraoss.co.jp
>>
>> > Hi
>> >
>> > We are struggling with the bellow issues for a long time - any help, even
>> > partial, is highly appreciated!
>> >
>> > We use pgpool-II 3.3.3 with 4 Postrges 9.3 nodes. We are experiencing the
>> > following issues:
>> >
>> > 1. In the db servers I see many idle connections (processes) for a long
>> > time (even days). "show pool_pools" does not show those connections so
>> > pgpool is not aware of them anymore. All connections to DB come from
>> > pg-pool so possibly it does not close the connection before killing the
>> > child process
>> >
>> > 2. In "show pool_pools" I see many idle connections far behind their
>> > connection_life_time and even behind their pool child_life_time. The
>> system
>> > was inactive for a few hours so I know for sure these connections are not
>> > in use.
>> >
>> > 3. In "show pool_pools" I sometimes see multiple connections within the
>> > same pool with pool_connected=1. Also, I know for a fact that those
>> client
>> > processes were killed a long time ago.
>> >
>> >
>> > Here is our config:
>> >
>> > num_init_children=400
>> > max_pool=5
>> > child_life_time=300
>> > child_max_connections=0
>> > connection_life_time=60
>> > client_idle_limit=300
>> >
>> > (we have 4 backends)
>> >
>> >
>> > This is example of multiple connections in the same child with
>> > pool_connected=1 (how can it be?).
>> > Also those connections are already not active for many hours (inactive
>> > zombie connection also happens when pool_connected=0)
>> >
>> > pool_pid | start_time | pool_id | backend_id | database |
>> > username | create_time | majorversion | minorversion |
>> > pool_counter | pool_backendpid | pool_connected
>> > 4741 | 2014-09-23 06:44:46 | 0 | 1 | DB1 |
>> > user1 | 2014-09-23 07:15:58 | 3 | 0 | 7
>> > | 6911 | 1
>> > 4741 | 2014-09-23 06:44:46 | 0 | 2 | DB1 |
>> > user1 | 2014-09-23 07:15:58 | 3 | 0 | 7
>> > | 22316 | 1
>> > 4741 | 2014-09-23 06:44:46 | 0 | 3 | DB1 |
>> > user1 | 2014-09-23 07:15:58 | 3 | 0 | 7
>> > | 27441 | 1
>> >
>> >
>> > Any idea about what can cause this is highly valuable for us (even
>> partial
>> > idea).
>> >
>> > Thanks,
>> > Yaron
>>
>
>
>
> --
>
> I'm on Twitter (@YaronNaveh <http://twitter.com/#!/YaronNaveh>)
More information about the pgpool-general
mailing list