robert delius royar wrote:
The earlier version did not require any thread-related parameters to
configure. Perhaps this problem is a thread-related issue. Note that
sendmail was compiled with threads enabled (nm shows calls to
pthread_* in libmilter.so)
[As an aside, when I did try 0.90 clamd, it quickly ran up to 80% of
my CPU time (on a dual CPU system). The only way to kill it was a
SIGKILL.]
This sounds like the threading problems mentioned previously with
FreeBSD and 0.9rc clamd. We solved it by forcing clamd to use libthr
rather than libpthread using /etc/libmap.conf, and the solution also
appeared to work for another admin a few days ago (see the "ScanStream:
accept timeout, unkillable" thread).
However, we use FreeBSD 6, not 4, so I don't know if there's an
equivalent or similar solution for you.
We haven't tried 0.90 itself as it's not yet in the ports tree, but from
the "load under 0.90" thread detailing issues Solaris users are having
with 0.90 and the similar or same symptoms, I'm thinking it might be
that Clam's threading code had some tweaking done and the problem's
there, rather than with the OSes themselves. However, I readily admit
that I am not an expert and that this speculation is likely worth
considerably less than 2 cents.... :-)
Craig.
------
_______________________________________________
Help us build a comprehensive ClamAV guide: visit http://wiki.clamav.net
http://lurker.clamav.net/list/clamav-users.html
