We found limiting the number of lmtp processes to 10 helped with locker problems. Let sendmail/postfix handle the queuing and lmtp the delivery.. When we tried to blast 10-20 messages/sec at our test server locking became a problem with the duplicate delivery db. Across 3 servers and 600 concurrent logins (lmtpd max=10 on each server) we never see any locking problems and usually have <1 second delivery times..
Rob Siemborski wrote: > > On Thu, 25 Sep 2003, Robert Covell wrote: > > > We are in the process of replacing our 1.5.24 Cyrus box with a new 2.1.15 > > box. After we got everything setup, tested and stabilized we rebooted the > > box and got: > > > > Sep 25 16:12:09 mail1b ctl_cyrusdb[152]: checkpointing cyrus databases > > Sep 25 16:12:09 mail1b lmtpd[155]: DBERROR db3: 4 lockers > > Sep 25 16:12:09 mail1b ctl_cyrusdb[152]: done checkpointing cyrus databases > > Sep 25 16:12:09 mail1b imapd[156]: DBERROR db3: 4 lockers > > Sep 25 16:12:11 mail1b imapd[172]: DBERROR db3: 4 lockers > > Note that in reality, esepcially for smaller sites, these messages on > their own aren't a severe concern. Only if the number is rapidly > increasing should they be worried about. > > > Is the recommendation of using skiplist with mailboxes.db still valid? Or > > do we need to shutdown the cyrus server before we shutdown? If shutting it > > down before a reboot what is the preferred method? We are still in a > > position to change the format if necessary before migrating 800+ users... > > Yes. You will need to recompile the cyrus server to do this. You will > also want to dump the mailbox list beforehand, and undump it afterwords > (as with any 1.x -> 2.x upgrade). > > -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- > Rob Siemborski * Andrew Systems Group * Cyert Hall 207 * 412-268-7456 > Research Systems Programmer * /usr/contributed Gatekeeper
