Re: Very slow planning performance on partition table

This was no error in the log of pgbouncer, but there is a sudden drop of
request count when the problem happened:
2014-07-30 11:36:51.919 25369 LOG Stats: 2394 req/s, in 339478 b/s, out
1422425 b/s,query 3792 us
2014-07-30 11:37:51.919 25369 LOG Stats: 2207 req/s, in 314570 b/s, out
2291440 b/s,query 5344 us
2014-07-30 11:38:51.919 25369 LOG Stats: 2151 req/s, in 288565 b/s, out
1945795 b/s,query 10016 us
[=========problem happens=========]
2014-07-30 11:39:51.919 25369 LOG Stats: 1061 req/s, in 140077 b/s, out
2652730 b/s,query 515753 us
[=========pgbouncer restart=========]
2014-07-30 11:40:52.780 10640 LOG File descriptor limit: 65535
(H:65535), max_client_conn: 5500, max fds possible: 6560
2014-07-30 11:40:52.781 10640 LOG Stale pidfile, removing
2014-07-30 11:40:52.782 10642 LOG listening on 0.0.0.0:xxxx
2014-07-30 11:40:52.782 10642 WARNING Cannot listen on ::/xxxx: bind():
Address already in use
2014-07-30 11:40:52.782 10642 LOG listening on unix:/tmp/.s.PGSQL.xxxx
2014-07-30 11:40:52.782 10642 LOG process up: pgbouncer 1.5.4, libevent
1.4.13-stable (epoll), adns: libc-2.11
2014-07-30 11:41:52.781 10642 LOG Stats: 2309 req/s, in 331097 b/s, out
3806033 b/s,query 4671 us
2014-07-30 11:42:52.782 10642 LOG Stats: 2044 req/s, in 285153 b/s, out
2932543 b/s,query 4789 us
2014-07-30 11:43:52.782 10642 LOG Stats: 1969 req/s, in 282697 b/s, out
560439 b/s,query 4607 us
2014-07-30 11:44:52.782 10642 LOG Stats: 2551 req/s, in 351589 b/s, out
3223438 b/s,query 4364 us