[pgpool-general: 4563] Stuck DISCARD connexions

[Thomas SIMON]

Hi,

I recently upgrade pgpool from 3.3.4-1 to 3.4.3-1 , and upgraded in the 
same time postgres from 9.3 to 9.5.

Since this upgrade, I have connexions stuck, who are in "DISCARD" state 
on pgpool , and in "idle" state between postgres and pgpool.
Number of connexions grows slowly everyday. (total 435 on 1800 now)


It seems reset_query_list actions are done, but connexions never ends up.
I tried to change client_idle_limit and connection_life_time (0 
previsously) and reload pgpool, but no changes after more than one day.

pgpool parameters :
reset_query_list = 'ABORT; DISCARD ALL'
child_life_time = 300
child_max_connections = 20
connection_life_time = 3600
client_idle_limit = 86400



Below quick example :

ps -ef | grep pgpool
...
postgres 50593 54042  0 Mar10 ?        00:00:01 pgpool: talend 
talend_administrator talend1.xxx(36946) DISCARD
postgres 50713 54042  0 Mar10 ?        00:00:00 pgpool: talend 
talend_administrator talend1.xxx(39774) DISCARD
postgres 52132 54042  0 Mar16 ?        00:00:00 pgpool: talend 
talend_administrator talend1.xxx(47671) DISCARD
postgres 56773 54042  0 Mar16 ?        00:00:00 pgpool: talend 
talend_administrator talend1.xxx(48738) DISCARD
postgres 57400 54042  0 Mar10 ?        00:00:00 pgpool: talend 
talend_administrator talend1.xxx(39779) DISCARD
postgres 57695 54042  0 Mar10 ?        00:00:00 pgpool: talend 
talend_administrator talend1.xxx(39789) DISCARD
postgres 58165 54042  0 Mar16 ?        00:00:01 pgpool: talend 
talend_administrator talend1.xxx(60536) DISCARD
postgres 87664 54042  0 17:59 ?        00:00:00 pgpool: talend 
talend_administrator talend1.xxx(55556) idle
...

ps -ef | grep postgres
...
postgres 72306 47359  0 Mar16 ?        00:00:00 postgres: prod: talend 
talend_administrator localhost.localdomain(34000) idle
postgres 74487 47359  0 Mar10 ?        00:00:00 postgres: prod: talend 
talend_administrator localhost.localdomain(42283) idle
postgres 74490 47359  0 Mar10 ?        00:00:00 postgres: prod: talend 
talend_administrator localhost.localdomain(42285) idle
postgres 74524 47359  0 Mar10 ?        00:00:00 postgres: prod: talend 
talend_administrator localhost.localdomain(42301) idle
postgres 74620 47359  0 Mar10 ?        00:00:00 postgres: prod: talend 
talend_administrator localhost.localdomain(42367) idle
postgres 74627 47359  0 Mar10 ?        00:00:01 postgres: prod: talend 
talend_administrator localhost.localdomain(42380) idle
...



pcp_proc_info 20 localhost 9898 pcpuser `cat /root/.pcp_passwd` 0
...
database talend 2016-03-11 11:02:17 2016-03-11 11:06:20 3 0 1 37685 1
database talend 2016-03-11 11:02:17 2016-03-11 11:06:20 3 0 1 2488 1
database talend 2016-03-15 10:18:18 2016-03-15 10:18:29 3 0 1 14820 1
database talend 2016-03-15 10:18:18 2016-03-15 10:18:29 3 0 1 12508 1
database talend 2016-03-22 13:50:11 2016-03-22 14:31:50 3 0 1 49029 1
...


does anyone have an idea of why this happens ?
thanks,
thomas