New MS install is slow to an extreme

[Peter Farrow]

DAve wrote:
> Peter Farrow wrote:
>> DAve wrote:
>>> Likely I am going to break threading, but I am reading via the 
>>> archive, I am far to slow to get the messages.
>>>
>>>> Check this option in mailscanner.conf
>>>> ClamAV Full Message Scan = no
>>>
>>> Set to no
>>>
>>>>
>>>> how do you do the bayes expire?
>>>
>>> MailScanner, same as all previous installs.
>>>
>>>>
>>>> if it's set to yes really slows down de process.
>>>>
>>>> Do you have /var/spool/MailScanner/incoming on tmpfs?
>>>
>>> Not at this time, I didn't previously.
>>>
>>>> Start by
>>>>     MailScanner --debug --debug-sa
>>>
>>> bash-2.05b# ./MailScanner --debug --debug-sa
>>> In Debugging mode, not forking...
>>> Trying to setlogsock(unix)
>>>
>>>
>>> *****
>>> If 'awk' (with support for the function strftime) was
>>> available on your $PATH then all the SpamAssassin debug
>>> output would have the current time added to the start of
>>> every line, making debugging far easier.
>>> *****
>>>
>>> SpamAssassin temp dir = 
>>> /var/spool/MailScanner/incoming/SpamAssassin-Temp
>>> [72851] dbg: logger: adding facilities: all
>>> [72851] dbg: logger: logging level is DBG
>>> [72851] dbg: generic: SpamAssassin version 3.2.4
>>> [72851] dbg: config: score set 0 chosen.
>>> [72851] dbg: dns: no ipv6
>>> [72851] dbg: dns: is Net::DNS::Resolver available? yes
>>> [72851] dbg: dns: Net::DNS version: 0.62
>>> Use of uninitialized value in concatenation (.) or string at 
>>> /usr/local/lib/perl5/site_perl/5.8.8/Mail/SpamAssassin.pm line 1088.
>>> Use of uninitialized value in concatenation (.) or string at 
>>> /usr/local/lib/perl5/site_perl/5.8.8/Mail/SpamAssassin.pm line 1090.
>>> [72851] dbg: config: read_scoreonly_config: cannot open "": No such 
>>> file or directory
>>> Building a message batch to scan...
>>> Have a batch of 30 messages.
>>> Stopping now as you are debugging me.
>>>
>>>> to see where the holdups are.
>>>> Check you haven't got a screwed SpamAssassin cache as well.
>>>
>>> I removed it an restarted MS, no change.
>>>
>>>> Which MTA are you using?
>>>
>>> Sendmail
>>>
>>> It is worth noting that spamassassin -D --lint runs fine. I checked 
>>> after I moved all our config over to the new version.
>>>
>>> Thanks everyone. I'm still looking at what might be wrong.
>>>
>>>
>>> DAve
>>>
>> I had this issue with the latest version on Centos 4.6.  As I was in 
>> a hurry at the time I rolled it back to 4.65.3-1 and the problem went 
>> away just to prove it wasn't any of my Milters or modules causing the 
>> problem.  I wasn't too bothered as I was upgrading from 4.5x so I 
>> still got a major upgrade..
>>
>> I did notice that it wasn't just slow as you describe, it actaully 
>> was just queing everything, I didn't leave it long enough to verify 
>> that though as I got 600 messages in the queue quite quickly.
>>
>> The problem in debug mode was exactly as you describe with it halting 
>> at "Trying to setlogsock(unix)"
>>
>> The error seems to be to do with the perl sys::syslog , the perl -> 
>> syslog interface routine.
>>
>> By rolling back to the slightly earlier MailScanner version the 
>> problem was instantly fixed...
>>
>> Pete
>>
>
> I upgeaded the sys::syslog from .18 to .24 with no change. But, during 
> debug I am getting this and it slows down consierably even with a 
> single message batch.
>
> rules: failed to run DNS_FROM_OPENWHOIS RBL test, skipping:
>         (Can't locate object method "check_rbl_envfrom" via package 
> "Mail::SpamAssassin::PerMsgStatus" at 
> /usr/local/lib/perl5/site_perl/5.8.8/Mail/SpamAssassin/Plugin/Check.pm 
> line 203.
> )
>
> This seems to be the killer. I have no rbl checks configured with SA, 
> so  I don't understand this.
>
> DAve
>
What version of Spamassassin are you running?

I followed a few related links and got to this entry in CPAN...

http://search.cpan.org/dist/Mail-SpamAssassin/lib/Mail/SpamAssassin/DnsResolver.pm

It mentions about a bug:

"This is a DNS resolution engine for SpamAssassin, implemented in order 
to reduce file descriptor usage by Net::DNS and avoid a response 
collision bug in that module."

This may or may not be significant...might be worth a look..

P.