SpamAssassin timeout

[spiff06]

I've adjusted some of the MailScanner parameters to block GIF spam. Works ok (except for false positives on daily Dilberts), but a few appear to go through because SpamAssassin times out: SpamCheck: not spam, SpamAssassin (not cached, timed out).

I've read (here) that it may have to do with Bayes database corruption.

I would appreciate if someone would kindly point me in the right direction to solve this.

Some more info:

• I have ConfigServer's (excellent) MailScanner package.
• There's a bunch of bayes_toks expire files in the /var/spool/mqueue/.spamassassin/
  directory. I tried sa-learn --force-expire, didn't report any errors, but didn't get rid of the expire files.

 

[chirpy]

Best way to find out where it is timing out is to run a lint test:

spamassassin --lint -D

If you're using our front-end, then it's best done in MailWatch > Tools/Links > SpamAssassin Lint (Test)

This helpfully gives timings for each stage. The most common reason for timeouts is if you have slow DNS resolvers.

 

[spiff06]

Thanks, Jonathan, for your reply.

I'm listing here the top entries that came up from that run (all others are <0.1, most <0.01):

dbg: dcc: got response: X-DCC-CollegeOfNewCaledonia-Metrics: server.domain.com 1189; Body=41432 Fuz1=41432 Fuz2=10409429 2.14872
dbg: diag: perl platform: 5.008001 linux 1.9023
dbg: rules: ran body rule __NONEMPTY_BODY ======> got hit: "I" 1.23673
dbg: plugin: Mail::SpamAssassin::Plugin::ReplaceTags=HASH(0x925e7b8) implements 'finish_parsing_end' 1.19388
dbg: info: leaving helper-app run mode 1.17555
dbg: rules: ran header rule __SANE_MSGID ======> got hit: "<1156263588@lint_rules> 0.46158
dbg: config: adding redirector regex: /^http:\/\/chkpt\.zdnet\.com\/chkpt\/\w+\/(.*)$/i 0.28856
dbg: rules: running header regexp tests; score so far=5.076 0.15876
dbg: bayes: tie-ing to DB file R/O /var/spool/mqueue/.spamassassin/bayes_toks 0.1128​

I wonder about the DCC. Anything else worth noticing in here?

 

[chirpy]

Nothing looks particularly bad there. Do you use a local DNS resolver on the server or an external one? Often external resolvers can be slow and cause lookups to take some time to return and using the local DNS server in /etc/resolv.conf can fix those issues.

 

[spiff06]

I have a local IPs at the top of /etc/resolv.conf, complemented with ServerMatrix IPs underneath to prevent the occasional upcp lockout (discussed here). Would the extra IPs actually cause the timeouts, or should the timeout setting be changed in SA config?

 

[chirpy]

The addition nameservers wouldn't cause any delay. The SA timeout is actually controlled in MailScanner.conf with:

SpamAssassin Timeout = 75

 

[spiff06]

Mmh, this appears to be a bad idea, on my server at least. When I add the above line to the bottom of MailScanner.conf (there was no timeout setting at all in the config file, by the way), and restart MailScanner, mail stops being delivered. The queue resumes instantly once I comment out the timeout line and restart MailScanner again. Did it a couple of times, the failure is consistent, and caused by the addition of this setting.

Does any other setting need to be changed (in exim.conf, for example), in order for this to work?