Marc, I ran into the sme thing after bringing our Solaris 8 system running 2.0.16 online. It cropped up after we starting seeing a lot of incoming email. We went through a rough week of watching & restarting master until Larry Greenfield came up with a work-around. We made a change to master/service.h which entailed changing MAX_USE from 100 to 1, and recompiling the whole lot. We then stopped sendmail & master, deleted all the files in /var/imap/deliverdb & deliverdb/db, and manually installed lmtpd in /usr/cyrus/bin. This set's lmtpd to exit after a single use, instead of hanging around & getting re-used (the basic thought was that memory was getting corrupted after multiple re-use, which in turn was corrupting the deliverdb files). The reason you want to only copy lmtpd is because service is used by all of the Cyrus daemons (like imapd & popd, etc). You will probably notice a minor hit in performance, but it may fix your problem. After doing this we have had no problems since then (for about 3 weeks now). -chris
On Friday, February 1, 2002, at 10:02 AM, Marc G. Fournier wrote: > > We're getting geared up to move from a *very* old Cyrus IMAP server > running on a sparc/solaris to a Dual x86 with FreeBSD ... currently, we > have 2.0.16 installed from ports, and about 9 users on the system, so > load > is minimal ... > > We do have our postmaster mailbox on that system, which gets the most > email for a campus of >4000 students, staff and faculty ... > > Lately, the server has been "locking up" more and more frequently ... > not > the machine/os, but, the lmtp stuff ... maillog shows: > > Jan 31 21:00:49 new-relay sendmail[81380]: g1110XM81307: SYSERR(root): > Could not connect to socket /var/spool/imap/socket/lmtp: Connection > refused by localhost > Jan 31 21:00:49 new-relay sendmail[81384]: g110TgM76189: SYSERR(root): > Could not connect to socket /var/spool/imap/socket/lmtp: Connection > refused by localhost > Jan 31 21:00:49 new-relay sendmail[81386]: g1110SM81279: SYSERR(root): > Could not connect to socket /var/spool/imap/socket/lmtp: Connection > refused by localhost > Jan 31 21:00:49 new-relay sendmail[81383]: g1110BM81243: SYSERR(root): > Could not connect to socket /var/spool/imap/socket/lmtp: Connection > refused by localhost > Jan 31 21:00:50 new-relay sendmail[81389]: g1110XM81305: SYSERR(root): > Could not connect to socket /var/spool/imap/socket/lmtp: Connection > refused by localhost > Jan 31 21:00:50 new-relay sendmail[81385]: g110iKM78675: SYSERR(root): > Could not connect to socket /var/spool/imap/socket/lmtp: Connection > refused by localhost > > to fix, I have to kill off the master process, and restart it, and then > she processes everything from the queue until the next time it > 'hangs' ... > > I've looked through the changes file for 2.1.1, and see one point about: > > " * fixed a problem with LMTP AUTH and unix domain sockets " > > But that appears to be about it ... > > Its as if the master process can't start up a new lmtp process after a > period of time? > > Thoughts? > > Thanks ... > > > Marc G. Fournier > [EMAIL PROTECTED] > Senior Systems Administrator Acadia > University > > "These are my opinions, which are not necessarily shared by my > employer" >
