On Tue, 26 Apr 2005 14:17:35 +0300 Timo Sirainen <tss@iki.fi> wrote:
On Tue, 2005-04-26 at 12:01 +0100, Chris Wakelin wrote:
On Tue, 26 Apr 2005 12:57:26 +0300 Timo Sirainen <tss@iki.fi> wrote:
On Mon, 2005-04-25 at 15:51 +0100, Chris Wakelin wrote:
I've not seen the UIDVALIDITY issue in the last few days since using dovecot-stable-20050422 but I've noticed it takes much longer to open a message. Looking at protocol dumps, it's taking 3 seconds to do "UID FETCH xxx (UID RFC822.SIZE BODY[])" and another 3 seconds to do "UID STORE xxx +Flags (\Seen)" (using Thunderbird). Is this expected as a result of changing the sync code? Has anybody noticed this with 1.0-test67/68? (I might try and build test68 and see!)
I did some changes to it and updated the tarball, maybe it works better now?
It seems worse. It's getting into a sync_loop (logging "sync_loop XX" where XX increases) for 20 minutes with my 7191-message INBOX (I believe in giving it a thorough test!).
Oops, I forgot the debug message in there :) It's slow because master process starts slowing down the log writing and causing imap process to block on log output.
You could grep for i_info line in src/lib-storage/index/mbox/mbox-sync.c and remove it. I'll update the tarball also.
OK done, and it seems better. Mind you, I had mail delayed (and now delivered in random order) while it was syncing, which proves the locking works, I guess :)
It still seems to have the delay, though :(
Then when a new message arrived, it's doing it again :o
Do you have mbox_dirty_syncs = yes, or did you change mailbox before that? Within same SELECT, it should normally just read just the first and last two messages when new mail arrives..
I've left it at the default, so I guess that means mbox_dirty_syncs = yes.
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