Remote mailbox creation error / unified murder
Janne Peltonen
janne.peltonen at helsinki.fi
Sat Jun 30 16:47:59 EDT 2007
Hi!
Using mutt, connecting via another backend, trying to create a folder
(ie. save a message to a new folder), result:
imap_copy_messages [a0009 NO [TRYCREATE] Mailbox does not exist]?
Connecting directly to the correct backend, result:
Append messages to imaps://jmmpelto@i08.mappi.helsinki.fi/INBOX.tataeiole? ([yes]/no):
So apparently what happens is, the mailbox gets created on the correct
backend, but the information isn't immediately reflected on the
mailboxes list of the proxying 'backend'. Actually, the change won't be
propagated to any of the mupdate slaves (that /are/ listening).
The same goes with deletion: the mailboxes list of the proxying backend
(and, therefore, my client) sees doesn't get updated. What am I missing
this time... My test setup, with the 2.3.8-4 invoca rpm based cyrus
doesn't express this behaviour (on the other hand, it a has much smaller
mailboxes list and much less members. But the mailboxes don't seem to
(dis)appear on the proxying members at all.. and turning off most
members of the to-be-production setup didn't help..)
(The setup to be in production in 2.3.8-7, with the patches to avoid
deletion of local mailboxes - looking at the code, the shouldn't cause
this kind of behaviour. And anyway, the test setup also has those
patches. I couldn't find any significant differences between the config
files, either. And I tried turning off the firewalls on both the mupdate
master and the nodes concerned, in case there was something in the way.)
Restarting the mupdate master server causes all the slaves to
re-synchronize their mailboxes lists. (And this, in turn, generates
a substantial amount of write load.) And this seems to be the only way
to propagate the changes.
Could anybody point to any help? I'm confused. Simon, if you're
listening: could any of the patches added between releases 2.3.8-4 and
-7 cause this?
--Janne
--
Janne Peltonen <janne.peltonen at helsinki.fi>
More information about the Info-cyrus
mailing list