Alex,
> I have a server that's frequently pretty busy during the day, and just
> started to notice these messages periodically:
>
> Jun 8 13:35:39 mail01 amavis[28784]: (28784-272) SA TIMED OUT,
> backtrace: at /usr/lib/perl5/5.6.0/i386-linux/IO/Select.pm line
> 104\n\tIO::Select::can_read('IO::Select=ARRAY(0xe3ea068)', 10) called
> at /usr/lib/perl5/site_perl/5.6.0/i386-linux/Net/DNS/Resolver/Base.pm
> line
> 668\n\tNet::DNS::Resolver::Base::send_udp('Net::DNS::Resolver=HASH(0xe392b
> a4)', 'Net::DNS::Packet=HASH(0xe39f5d0)',
> '(m-v...@^@^...@^@^...@^@^...@^@^fpro164^itenaccave^cc...@^@^...@^a') [...]
>
> Is this a result of the server being too busy to process the request,
> or one of the services that it uses, such as a DNSBL, not responding
> in time? How can I troubleshoot this?
The reported stackdump just shows a routine that was executing at
the time when timer expired. It often shows the problem section
(like a database access), but may just show an innocent bystander
code which happened to be normally executing at the time.
If it happens often, check which code sections are reported
often. If it usually shows the same or closely related routine,
then a problem may indeed be there or close. If the reported
routines appear random, then your host is just generally overloaded.
Slow or dead DNSBL and URIBL servers/blacklists are normally not
a problem, as the time limiting on DNS replies works very well
in SA 3.3.* and 3.2.5. Regardless, a locally running caching DNS
server is almost a must with any substantial amount of mail traffic.
Mark
