If you have not exported the mailboxes (to the tempfile you list, for example), is there any way to manually rebuild?
My issue is that the OS drive went down, took everything but /var with it. I've finally reinstalled Cyrus, but now need to get the mailboxes to work properly. Any ideas? Thanks, Matt Kenigson ----- Original Message ----- From: "Mika Iisakkila" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Tuesday, December 04, 2001 7:39 AM Subject: Re: Mailbox Backup > Tarjei Huse wrote: > > PS: Some people noted that in addition to backing up the /var/imap > > and /var/spool/imap folders, you need to export some of the .db files to > > txt. Does anyone want to fill me out on this?? > > To export mailboxes.db to a text file: > > ctl_mboxlist -d > tempfile > > To restore it later: > > ctl_mboxlist -u < tempfile > > (start I-might-be-wrong section:) > The point in this is that the fine option for rebuilding the master > mailbox database from scratch (reconstruct -m) DOES NOT WORK for > some reason in Cyrus 2.0.16, even if it is on the man page. > Once you screw up the mailboxes.db (or some of the huge files under > db/*, they are part of the same database), Cyrus has lost knowledge of > your mailboxes and their ACLs. That's when you'll need the tempfile you > created above to recreate the mailbox database first. > > The remaining mailbox data and their contents can be then salvaged from > the mail store with 'reconstruct -r user.*' after pulling them from > regular backups. Fixing quotas (quota -f) might also be necessary. > > Cyrus uses the DB3 libraries in a transactional manner, so a backup > taken from a running system will not be enough to restore it to > a consistent state. After a crash, you can use the DB utilities > to "roll back" the events to a previous safe point in time, which is > saved periodically (the checkpoint events in cyrus.conf you always > wondered about, but was afraid to ask...). If you want to know more, > read the gigabyte of fine documentation at www.sleepycat.com :-) > > I don't know what should be done with deliver.db and its contents > (might be safe to just delete before restart?) and whether > singleinstancestore on or off affects that. Would anyone > who actually knows this stuff care to comment? > > --mika >
