Re: [PERFORM] Very slow planning performance on partition table

[Rural Hunter]

More information found. After the hang connection appears, I noticed 
there were several hundreds of connections of the same user. Since I use 
pgbouncer and I only set the pool size to 50 for each user, this is very 
strange.  I checked the pgbouncer side, 'show pools' showed the active 
server connection count is less than 50(only 35 actually). I also 
checked the client port which is shown in pg process list. It is not 
used at pgbouncer side when I did the check. So I stopped pgbouncer then 
the connection count from the user drops slowly. Finally all those 
connections disappeared. After that I restarted pgbouncer and it looks 
good again.
With this solution, I at least don't have to kill pg when the problem 
happens. But anyone has a clue why this happens? What I need to check 
for the root cause? One thing I forgot to check is the network status of 
those orphan connections at pg side. I will check it next time and see 
if they are in abnormal status.

--
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance