intermittent mupdate failures.

[Darrell Fuhriman]

We are currently working on migrating our UW servers to Cyrus. We
currently have one each mupdate master, frontend, and backend.

We're migrating users using a slightly modified version of the
imapsync script.

What we're seeing is that sometimes, a user's mailbox is not
properly propagated to the front end.  The update comes from the
backend to the mupdate master, but never goes from the master to
the frontend.  This is confirmed with network snoops,
additionally the mailboxes.db on the master and the backend are
correct, while mailboxes.db on the frontend is missing the
entries.

It seems to happen about 1/3 of the time, but I don't have enough
evidence to really trust that number.

In cyrus.conf on the frontend we have:

mupdate       cmd="mupdate" listen=3905 prefork=3

and on the mupdate master, we have:

mupdate       cmd="mupdate -m" listen=3905 prefork=3


Now, the (possible) ~1/3 failure rate with 3 mupdate daemons
seems to be an "interesting" co-incidence.

If we restart mupdate on the master, then, as would be expected
the "missing" mailboxes correctly get sent to the frontend
machine when it reconnects.

Has anyone else seen this issue?  As you might imagine, it's kind
of holding up our migration.

Darrell
---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html