Fixing inconsistent state on interrupted XFER

[J.McKeown at ru.ac.za]

We're trying to migrate without downtime from 2.2.13p1 to 2.4.10. We  
put a 2.4.10 front-end/mupdate server in place, turned our 2.2.13 imap  
server into a back end, built a new back end running 2.4.10 and then  
started XFERring mailboxes. (This didn't work particularly well as the  
front end issues LOCALAPPEND to the back end, which appears to be  
unsupported on 2.2.13, so we lost the ability to save a copy of sent  
mail).

Unfortunately at some point the old back end imapd was restarted,  
interrupting 4 simultaneous XFERs. I now have four mailboxes which  
exist on the old back end, with subfolders which have transferred to  
the new back end, and the mailbox dbs on the three servers are  
inconsistent.


I can't redo the XFER - the old back end reports that the mailboxes  
are already remote, on the new server. If I try and read the mailboxes  
on the new server, I get an error stating that it can't mmap new  
message file. I've tried removing the mailboxes from the mailbox db  
using cyr_dbtool on the new back end and on the front end, and running  
ctl_mboxlist on the old backend which claims to revert the mailboxes  
to being local to the old backend. It still fails during the XFER with  
a couple of different errors.

Does anyone have any tips for straightening out the various servers'  
ideas of what is where, and transferring the old inboxes from the  
2.2.13p1 back end to the 2.4.10 backend?

Thanks
jonathan