[pgpool-general: 3017] pgpool watchdog pipe open error

[Ryan DeShone]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello all,

I'm experiencing an issue with pgpool 3.3.3 running on CentOS 6. If I
enable the connection watchdog and configured the upstream trusted
servers, when pgpool is started it will function fine for a while.
After some time, however, it logs the following error repeatedly:

pgpool[11188]: exec_ping: pipe open error:Too many open files

Eventually, the wd_lifecheck will fail and watchdog will change status
to DOWN even though the server is not, in fact, down.

Jul  1 04:43:47 provdb-a pgpool[11188]: wd_lifecheck: failed to
connect to any trusted servers
Jul  1 04:43:47 provdb-a pgpool[11188]: gethostbyname() failed:
Unknown host host: hostname
Jul  1 04:43:47 provdb-a pgpool[11188]: send_packet_4_nodes: packet
for hostname:9000 is canceled
Jul  1 04:43:57 provdb-a pgpool[11188]: wd_lifecheck: watchdog status
is DOWN. You need to restart pgpool


While investigating this, I found that the number of open files for
pgpool steadily increases as time goes on. It appears as if pgpool
watchdog may not be closing its pipes properly. Has anyone else
experienced this issue? Does anyone have a workaround for it? I'd
really like to be able to turn the watchdog on to hopefully avoid an
(admittedly, extremely unlikely in our setup) edge case which could
result in split-brain, but it relies on having watchdog working with
trusted IPs.

- -- 
Ryan DeShone
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJTvZMcAAoJEJ3SEJwY0lkxTWQIALCZC+R4oyK439cWehojBFkB
cpjdGLfEKmgxpL6/92Uw6ZxLllu8ht9sdMzk5jsKM1klsfgCDsj82g6cxVAOFMbl
8YE2D7Go9ZwvQQk8ggwt9bFxssFtUZI+69JtBKoFr1Ji3EpOPTblE63oOAbnnvVK
Cr2DAKJXF7lV1Invu5g8fLQiJPOrF7OCEahSKEOB2g/lj8qTV9ETvRbr6JgexN3d
yz7leCm4RQHX3plKF3PtL2v0qH3ysv3S8e5dYfkdixev5G5FiJYf4To2Re3ueEws
DiQcOa91Ky09cvtRR5IpdKpUPRf/kR26eTxiK2LB/elKAGx21G3igIQG3S67zJw=
=x8NZ
-----END PGP SIGNATURE-----