Mike Sanchez wrote: >>Old Solaris version on an old Sparc with not much RAM? > > > No, it's got 3G ram 4CPU, processes peaks of 100,000 msgs/hr. > Solaris is tuned, played with the clamd config options.
You wrote "Solaris 2.7" that's Solaris 7, I haven't seen it in a long time (I use 8 and 9, haven't tried 10), does it even support multiple CPUs? > When problem occurs under .80, clamd can't scan messages quick enough and > messages wait in queue. Killing and restarting clamd corrects the problem. > I suspect specific characteristics of a message cause it. The 20%-30% is > not total system load, it's what clamd uses. Normally it's nice and clamd > stays below 5%. > > Under .84 when problem occurs, clamd takes up 50%-60% of cpu. 60% of 4 CPUs (top reports 1 CPU loads, so the total could go up to 400% in your case and 60/400 is only 15% really, prstat doesn't have this quirck IIRC)? How many threads is clamd using? Even under high load if one thread takes too long the others should keep processing the mail and you shouldn't see the mail queue growing. This definetly points out to some misconfiguration, have you tried increasing MaxThreads? Or, most likely, does your script handle a single queue, no multiple threads and it doesn't make a difference if clamd could do more threads? This means your script could be the bottleneck. > Thanks for your suggestions, appreciated, just wanted to see if anyone else > has seen this or had any ideas I haven't already tried. Your setup is just sendmail + your perl script + clamd? -- René Berber _______________________________________________ http://lurker.clamav.net/list/clamav-users.html
