Bug#501945: nscd runs out of filehandles opening multiple instances of /var/run/nscd/socket and uses 100% CPU
retitle 501945 nscd runs out of filehandles opening multiple instances of /var/run/nscd/socket and uses 100% CPU
found 501945 2.3.6.ds1-13etch8
As a data point: the same nscd failure occurred an hour ago, while
backups were running, on (as far as I can tell so far) a single system.
This particular system was last rebooted two days ago, so long uptime is
not an issue. While this problem was occurring, the slapd process on the
primary LDAP server was quite busy, causing its load average to rise
high enough to also trigger alarms. Nagios noticed the load average rise
before it noticed problems related to the runaway nscd process, but I
don't have data to indicate which actually occurred first.
As backups were running at the time, it could have been triggered by
something AMANDA is doing; however, cfengine was also busy doing its
four-times-an-hour traverse over a largeish directory tree.
Interactive access via the shell continued to work, albeit slowly.
While the problem was occurring, Samba and NRPE did not respond within
ten seconds. After I restarted nscd, everything returned to normal.
Apologies for the serveral typoes in the original bug report --
evidently I didn't proofread it well enough to notice 'of of' or
William Aoki KD7YAF firstname.lastname@example.org 5-1924