[Dovecot] How safe is mbox_very_dirty_syncs?
I've tried turning on mbox_very_dirty_syncs for myself (and a few other brave souls running 1.0-alpha3 rather than 1.0-stable) again to see what perfomance gain it gives (see thread http://www.dovecot.org/list/dovecot/2005-July/007956.html).
Now we aren't running UW-IMAP at all, how safe is mbox_very_dirty_syncs assuming the only other process writing to mailboxes is our MTA (exim) which simply appends messages to the end? Is it safer still using Dovecot 1.0-alphas and the LDA? Does it matter if there are concurrent connections to the mailbox?
Best Wishes, Chris
-- --+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+- Christopher Wakelin, c.d.wakelin@reading.ac.uk IT Services Centre, The University of Reading, Tel: +44 (0)118 378 8439 Whiteknights, Reading, RG6 2AF, UK Fax: +44 (0)118 975 3094
Chris Wakelin wrote:
Now we aren't running UW-IMAP at all, how safe is mbox_very_dirty_syncs assuming the only other process writing to mailboxes is our MTA (exim) which simply appends messages to the end?
I have been using mbox_very_dirty_syncs from the time it was added. If I remember correctly there might have been some problems way back then but nothing in last 6 months or so and I have been hitting Dovecot with millions of messages during this time.
Is it safer still using Dovecot 1.0-alphas and the LDA?
I would always use Dovecot LDA. It does have some extra overhead due indexing, which is not so bad thing, but it also adds X-UID, Status, X-Keywords and Content-Lenght headers to mboxes which decreases need for writes and offset changes on mboxes while opening the mailbox.
Does it matter if there are concurrent connections to the mailbox?
No if your mbox_*_locks are set correctly.
-- Tomi Hakala
On Tue, 2005-09-27 at 10:59 +0100, Chris Wakelin wrote:
Now we aren't running UW-IMAP at all, how safe is mbox_very_dirty_syncs assuming the only other process writing to mailboxes is our MTA (exim) which simply appends messages to the end?
It should have been safe even with UW-IMAP and other software changing the mbox. Only problem with it is that it might not have catched external flag updates as early as possible.
mbox_very_dirty_syncs just makes Dovecot opportunistic about thinking the mails are where Dovecot last saw them, but if at any point Dovecot notices that they're not where they should be, it just re-reads the entire mbox.
Is it safer still using Dovecot 1.0-alphas and the LDA?
If you're using now 1.0-stable, I'd think 1.0.alpha3 is better/safer in general :)
Does it matter if there are concurrent connections to the mailbox?
Nope.
participants (3)
-
Chris Wakelin
-
Timo Sirainen
-
Tomi Hakala