Resolver issue - drop in qps and memory leak

[Dennis Perisa]

Hi all,

We are running a couple of resolvers using BIND 9.6.2-p2 on FreeBSD 7.1 and
are peaking at 10,000 queries per second (qps).

Recently, the qps on one resolver dropped dramatically to 3000 during the
peak period.  The secondary picked up the additional queries and we had a
few complaints about response times, but not much more than that.
Restarting named seems to have resolved it and we are back at 10,000
qps during peak - for now.

We also observed the following during the problem period:
- memory usage increased steadily from 10% to 15% (typically it remains
steady at 10%)
- many events in the general log like this: "general: info: sockmgr
0x800d46000: maximum number of FD events (64) received"

Appreciate any help to determine the root cause.

Regards,
Dennis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/bind-users/attachments/20110408/50fd149e/attachment.html>