I have an environment using UW-Imap 2004g on multiple machines that
share common mailboxes via a Sun QFS Shared file system mounting from
SAN. These mbox formats work fine so far using .locking techniques. 

 

I'm looking at converting to MBX or Mix format (finally!) and am
doubtful that the flock implementation will work out of the box. In my
testing, there have been some random corruption issues with MBX style
boxes when I scatter usage among servers to the same mailbox. QFS
supports both flock and fcntl locking (I'm on Solaris 10, so I guess it'
s fcntl) but c-client is putting those inode exclusive lock files in
/tmp. QFS also controls file access at the FS level so that only one
client can write to a file at any given time unless I specifically
specify multi-writer ability at mount time, which I'm not doing. QFS is
a server-client model; a metadata server controls access and locking
while data access is direct. I'm wondering if the separate servers not
knowing about the /tmp.lockfiles is the reason for the corruption. It's
mainly only noticeable when larger attachments are being delivered
(snarfed).

 

Are there any modifications I can do to share the location of these tmp
lock files on one of my shared filesystems to ensure the flock'ing
success on MBX/Mix formats between multiple servers? Will that help or
hurt things? Am I just out of luck when it comes to exclusive locking
and multiple servers? The current environment provides good redundancy
and load balancing as it is now. I'd like to convert to newer mailbox
formats to get better performance for larger mailboxes.
_______________________________________________
Imap-uw mailing list
[email protected]
http://mailman2.u.washington.edu/mailman/listinfo/imap-uw

Reply via email to