#10 Solaris, resolv-problem then using -A (autorestart) option

v0.2.6
closed
5
2005-03-17
2005-03-07
Fredrik Pettai
No

I had lots of errors from sid-filter not been able to
resolve many known domains on the internet. Big
companies, who certainly shoudn't have any problems
(atleast not at the same time) The logentries looked
like this...

Mar 7 13:59:01 aries sid-filter[536]: [ID 375559
mail.error] j27CvXX7006204 DNS timeout (16 volvo.com)
Mar 7 13:59:01 aries sid-filter[536]: [ID 313135
mail.error] j27CvXX7006204 sid_marid_check(): -1 ( <
marid volvo.com?)

If i just did a nslookup/dig on the nameserver itself
(same machine running sendmail/sid-filter. It would
give right information, without any delay. I also use
sid-filer with -T 60 to be certain that theres enought
of time to resolve even very-very slow/busy sites.

However, then i removed the -A option from sid-filter,
i noticed that all errors like this one above just
disappeared.

Any ideas why, and how we can confirm this error (and
hopefully fix it)?

Discussion

    • labels: --> Functionality
    • milestone: --> v0.2.6
    • assigned_to: nobody --> sm-msk
     
  • Logged In: YES
    user_id=1048957

    I'm running with -A on my home OpenBSD box and the only
    ones for which this error gets logged are domains for which
    the nameservers are not responding. I tested this with the
    "dig" command, outside of sid-filter.

    medusa# host -t txt central.ca.
    ;; connection timed out; no servers could be reached
    medusa# host -t txt imagineering-inc.com.
    ;; connection timed out; no servers could be reached
    medusa# host -t txt businessmicrovar.com.
    ;; connection timed out; no servers could be reached
    medusa# host -t txt pluto.atlasdev.com.
    ;; connection timed out; no servers could be reached
    medusa# host -t txt accelerated.com.
    ;; connection timed out; no servers could be reached

    I'll take "-A" off to test, but I don't think it'll make a
    difference.

     
  • Fredrik Pettai
    Fredrik Pettai
    2005-03-15

    Logged In: YES
    user_id=370342

    It's easy for me to reproduce the problem, but i don't know
    how that would help you in finding the problem?
    Unlike you, I get a lot of "timeouts" on domains that are
    legimate domains, which works fine with dig.
    (But as i wrote before, those "timeouts" only comes then
    running sid-filter with -A)

    Would a truss do any good?

     
  • Logged In: YES
    user_id=1048957

    Does this still happen with 0.2.7?

     
  • Fredrik Pettai
    Fredrik Pettai
    2005-03-17

    • status: open --> closed
     
  • Fredrik Pettai
    Fredrik Pettai
    2005-03-17

    Logged In: YES
    user_id=370342

    Yes, i still get some failures, but not near the quantities
    i got before. The queries who still insists in failing is
    because ar_waitreply() fails first.

    Mar 17 13:27:24 libra sid-filter[13985]: [ID 146386
    mail.error] j2HCQhF5017805 ar_waitreply() failed
    Mar 17 13:27:24 libra sid-filter[13985]: [ID 313135
    mail.error] j2HCQhF5017805 sid_marid_check(): -1 ( < marid
    mensa.se?)

    Hmmm.....

    I use sid-filter with the flag "-T 60" and the milter config
    in sendmail have a long time to finnish (5 minutes) before
    the milter will got to status "failed".

    Maybe the fix in libar did solve something then.
    The statistics tend to get better after a filter-restart.

    (But its strange thats only sid-filter has problem with
    libar timeouts.
    dk-filter (compiled with AR_LIB) doesnt report this
    ar_waitreply() problems, yet it uses the same "-T" and
    milter parameter)

    /P