Bind 9.3.0 strange behavior with memory on EL3

[Michael Lang]

Hi all,

ive been reporting a problem previous, where Bind (9.3.0) goes on 99%
CPU (Intel) and doesnt answer any more queries. I found that the IOStat
was very high too, as the swapspace on IDE Discs wasnt performing good
enough. The Machine has 1 GB Ram without doing any unusual stuff
(default Services SSH, Local MTA, Syslog,...) except Bind. Ive
configured to views with CacheLimits on 256MB for one and 8Mb for the
other (Public is the big, the smaller is internal). Even these two
Limits and the RSS and Size of the named process arent filling up the
Memory, but the kernel starts to swap. Ive limited the Public View to
128MB for Cache, and the machine didnt start swaping. But today the
Process again started to use 99% CPU and didnt response to any more
queries. The Logfiles show up tons of Lines containing these records:

# grep 'still overmem' named.run
29-Oct-2004 07:05:41.648 cache cleaner: still overmem, reset and try
again
29-Oct-2004 07:06:08.001 cache cleaner: still overmem, reset and try
again
29-Oct-2004 07:06:34.244 cache cleaner: still overmem, reset and try
again
29-Oct-2004 07:06:56.500 cache cleaner: still overmem, reset and try
again
29-Oct-2004 07:07:16.570 cache cleaner: still overmem, reset and try
again

the didnt show up, earlyer (these log entrys have some time between, but
on the critical time there where 250 Entrys in one second)

I didnt find anything about such a problem on the Lists.
Any suggestions ?

Kind regards
Michael Lang

-- 
Michael Lang <michi+bind-users at relay3.jackal-net.at>