We operate a four-processor Ultra Enterprise 4000 with 1 GB of
central memory and several thousand users (interactive login,
POP3 service, WWW proxy).
User applications and services (daemons) experience random
"lock ups" of 10 to 30 seconds apparently linked to resolver
routine calls. The computer does *NOT* rund named.
Is nscd to blaim (or to reconfigure) ?
nscd -g :
Yes cache is enabled
1874035 cache hits on positive entries
2780 cache hits on negative entries
215158 cache misses on positive entries
50551 cache misses on negative entries
87% cache hit rate
16850 queries deferred
57889 total entries
211 suggested size
nscd.conf man page suggests upping the hash bucket table size
when the number of cache entries exceeds the suggested size by
more than a factor of four or five.
[That's 'factor' and NOT 'order of magnitude', right ?]
Eric Luyten, VUB/ULB Computing Centre.