cascading (?) errors

Dan White dwhite at olp.net
Mon Nov 7 17:08:57 EST 2011


On 07/11/11 15:57 -0500, brian wrote:
>Cyrus-IMAP 2.2.13-14ubuntu3.1
>Postfix 2.5.5-1.1
>Spamassassin 3.2.5-4ubuntu0.0.1
>
>I'm not certain this is a Cyrus issue, but Cyrus is the thing that i've
>changed recently.
>
>I've suddenly got a bunch of the following errors in my log:
>
>Nov  7 13:48:56 artemis postfix/pipe[31890]: 5A9A57A0EE6:
>to=<admin at REMOVED>, relay=procmail, delay=3.2, delays=3/0.01/0
>/0.22, dsn=4.3.0, status=deferred (temporary failure. Command output:
>procmail: Error while writing to "/var/log/procmai
>l.log" procmail: [31891] Mon Nov  7 13:48:55 2011 procmail: No match on
>"^X-Spam-Flag: YES" procmail: Executing "/usr/sb
>in/cyrdeliver,-a,admin at REMOVED,-m,user.admin at REMOVED" procmail: Error
>while writing to "/usr/sbin/cyrdeliver" procmail:
>Assigning "LASTFOLDER=/usr/sbin/cyrdeliver -a admin at REMOVED -m
>user.admin at REMOVED"  Subject: Re: REMOVED   Folder: **Requeued**
>8802057 )
>
>I searched online and found that the problem was (maybe) a permissions
>issue. Figuring that I'd temporarily resolve it until learning more, I
>tried:
>
>sudo chmod ugo+w /var/log/procmail.log
>
>... and re-started Postfix. Now i'm seeing:
>
>postfix/pipe[32052]: 8A0BF7A0EE9: to=<admin at REMOVED>, relay=procmail,
>delay=1, delays=0.99/0/0/0, dsn=2.0.0, status=sent (delivered via
>procmail service)
>postfix/qmgr[31998]: 8A0BF7A0EE9: removed
>cyrus/master[32055]: about to exec /usr/lib/cyrus/bin/lmtpd
>cyrus/lmtpunix[32055]: executed
>cyrus/lmtpunix[32055]: accepted connection
>cyrus/lmtpunix[32055]: lmtp connection preauth'd as postman
>cyrus/lmtpunix[32055]: verify_user(REMOVED!user^admin) failed: Mailbox
>does not exist

Is 'admin' a cyrus or domain administrator? If so, do you have the same
problem if it is not an administrator?

The domain.name!user^admin should be domain.name!user.admin I believe, so
you appear to have ended up with a top level mailbox named
'user^admin at domain.name' rather than a user mailbox for
'admin at domain.name'.

>and:
>
>postfix/pipe[32199]: B05177A1189: to=<admin at REMOVED>, relay=procmail,
>delay=1.1, delays=1/0/0/0.01, dsn=2.0.0, status=sent (delivered via
>procmail service)
>postfix/qmgr[31998]: B05177A1189: removed
>cyrus/lmtpunix[32192]: verify_user(REMOVED!user^admin) failed: Mailbox
>does not exist
>cyrus/lmtpunix[32202]: accepted connection
>cyrus/lmtpunix[32202]: lmtp connection preauth'd as postman
>cyrus/lmtpunix[32202]: verify_user(REMOVED!user^admin) failed: Mailbox
>does not exist
>postfix/pipe[32190]: 5A9A57A0EE6: to=<admin at REMOVED>, relay=procmail,
>delay=956, delays=950/0.5/0/5.3, dsn=4.3.0, status=deferred (temporary
>failure)
>
>
>The msg saying the mailbox doesn't exist is very worrying. As it
>happens, the procmail errors started showing up around the same time I
>created several new mailboxes. (For a different domain--the mailbox
>above has existed for years.) I logged in to cyradmin and listed the
>mailboxes. It seems fine:
>
>user/admin at REMOVED (\HasChildren)
>(shows also all the child fiolders)
>
>The one configuration setting that I changed at the time I created these
>new mailboxes was to change unixhierarchysep from "no" to "yes" as the
>new accounts have the dot.separated@ format. Could this be why the old
>mailboxes can't be found? Please don't tell me that I can use only one
>or the other format.

It's probably most significant when an admin connects, which is one
reason why it's not a good idea for admins to have their own mailboxes.

There's a warning about this at:

http://www.cyrusimap.org/docs/cyrus-imapd/2.4.12/altnamespace.php

>Next, i came across the following:
>
>spamc[32164]: connect to spamd on 127.0.0.1 failed, retrying (#1 of 3):
>Connection refused
>spamc[32164]: connect to spamd on 127.0.0.1 failed, retrying (#2 of 3):
>Connection refused
>spamc[32164]: connect to spamd on 127.0.0.1 failed, retrying (#3 of 3):
>Connection refused
>spamc[32164]: connection attempt to spamd aborted after 3 retries
>
>
>And then, while I was tailing mail.log, it appeared that one of the
>mailbox owners was trying to log in:
>
>cyrus/imaps[32403]: imaps TLS negotiation failed:
>modemcable228.38-48-24.mc.videotron.ca [24.48.38.228]
>cyrus/imaps[32403]: Fatal error: tls_start_servertls() failed
>cyrus/master[17882]: process 32403 exited, status 75
>cyrus/master[17882]: service imaps pid 32403 in BUSY state: terminated
>abnormally
>cyrus/master[17882]: process 32406 exited, status 75
>artemis cyrus/master[17882]: service imaps pid 32406 in BUSY state:
>terminated abnormally
>
>Yikes! So it seems that I may have several different problems on my
>hands. But does anyone recognise anything here that would cause all or
>any of this to occur suddenly?

A way to attack this piecemeal wise would be to create an email message and
feed it to cyrdeliver directly, to rule out any problems with procmail.

You can also dump your mailbox list with ctl_mboxlist, to see if you have
any corruption corruption or unexpected output.

-- 
Dan White


More information about the Info-cyrus mailing list