John Rudd wrote:
It might be good to try running some tests with -L and some tests
without -L, to see if the same test is fast without the network check.
That would help isolate the problem to "is SpamAssassin slow" or "is
SpamAssassin spending a lot of time waiting for DNS results". It
probably IS the latter, but it's always good to isolate symptoms, IMO.
Also, are you doing any other network stuff? razor, pyzor, etc.?
Hmm. I did the spamassassin test with -L and it flew through the test.
I have done some dig requests using the hosts in the debug file and they
are getting returned ok. Strange thing is that sometimes 18 or 38 pass,
sometimes 25 of 38 pass there seems to be no pattern to them failing.
There is a firewall running (nothing that would stop the DNS queries as
I disabled it for a test and that made no effect).
I cant see any reason why some DNS requests are failing. I'm looking at
running bind and using that for most of the rDNS queries to see if that
speeds stuff up.
As a work around I have set rbl_timeout to 5 to speed up the failures.
Anyone else got any ideas I could try.
Cheers
Richard
