cyrus imap hangs regularly

[Ilan Rabinovitch]

Hello,

I've been using Cyrus IMAP since the begining of this year and so far its
been working great.  I am currently running 2.1.11 on RedHat 8.0 (with all
the latest security updates).  

Lately I'm finding that the various cyrus processes go to "sleep" and never
wake up.  As a result sendmail is unable to deliver its mail to cyrus.  This
has happened 3 times in the last 48 hours.   When this happens users can connect to pop but not
IMAP, but again no new mail gets delivered from the MTA.

If I restart Cyrus it now takes over an hour to fully start.  It spends this
time: converting db files, ctl_cyrusdb -r, and ctl_deliver -r.  While these
are running users cannot connect to imap or pop.  The cyrus processes use
97% CPU on a pretty beefy machine for their entire duration.  When it completes they can
connect to both just fine and I can force sendmail to deliver all the mail
no problem.  However, this problem repeats just a few hours late.

I've checked and all my partitions have plenty of space....  So I'm not
really sure where to go with this.

The following log lines seemed interesting:

startup
======
Aug 12 20:38:39 igaming1 ctl_cyrusdb[7485]: recovering cyrus databases
Aug 12 21:04:46 igaming1 ctl_cyrusdb[7485]: DBERROR db4: Recovery function
for LSN 11 10312588 failed
Aug 12 21:04:47 igaming1 ctl_cyrusdb[7485]: DBERROR: dbenv->open
'/var/lib/imap/db' failed: No such process
Aug 12 21:04:47 igaming1 ctl_cyrusdb[7485]: DBERROR: init /var/lib/imap/db:
cyrusdb error
Aug 12 21:04:48 igaming1 ctl_cyrusdb[7485]: done recovering cyrus databases
Aug 12 21:04:48 igaming1 master[7483]: process 7485 exited, status 1 
Aug 12 21:04:48 igaming1 master[8076]: about to exec
/usr/libexec/cyrus/ctl_mboxlist
Aug 12 21:04:48 igaming1 ctl_mboxlist[8076]: running mboxlist recovery
Aug 12 21:04:48 igaming1 ctl_mboxlist[8076]: done running mboxlist recovery
Aug 12 21:04:48 igaming1 master[8077]: about to exec
/usr/libexec/cyrus/ctl_deliver
Aug 12 21:30:10 igaming1 ctl_deliver[8077]: DBERROR db4: Recovery function
for LSN 11 10312588 failed
Aug 12 21:30:10 igaming1 ctl_deliver[8077]: DBERROR: dbenv->open
'/var/lib/imap/db' failed: No such process
Aug 12 21:30:10 igaming1 ctl_deliver[8077]: DBERROR: init /var/lib/imap/db:
cyrusdb error
Aug 12 21:30:10 igaming1 master[7483]: process 8077 exited, status 1 
Aug 12 21:30:10 igaming1 master[7483]: ready for work
Aug 12 21:30:10 igaming1 master[8356]: about to exec
/usr/libexec/cyrus/ctl_mboxlist
Aug 12 21:30:10 igaming1 ctl_mboxlist[8356]: checkpointing mboxlist
Aug 12 21:30:10 igaming1 master[8357]: about to exec /usr/libexec/cyrus/imapd
Aug 12 21:30:10 igaming1 imap[8357]: executed
Aug 12 21:30:10 igaming1 master[8358]: about to exec /usr/libexec/cyrus/imapd
Aug 12 21:30:10 igaming1 imaps[8358]: executed
Aug 12 21:30:10 igaming1 master[8359]: about to exec /usr/libexec/cyrus/pop3d
Aug 12 21:30:10 igaming1 imapd[8357]: skiplist: recovered
/var/lib/imap/mailboxes.db (143 records, 14636 bytes) in 0 seconds

After this it spawns imap / pop3 processes and starts showing connections
from users.

Are my cyrus db's corrupt? I've run reconstruct the mailboxes but this
doesnt seem to change anything.

I appreciate any suggestions, advice, etc.
Thanks,
Ilan Rabinovitch 
ilan at fonz.net E-MAIL

The Southern California Llnux Expo: "Bringing businesses, academic
institutions and the Linux community together in a way that no other
conference does!" -- http://www.socallinuxexpo.org