On Wed, 2007-09-05 at 15:24 +0200, Peter Eriksson wrote:

I have a user that gets client timeout errors when he tries to delete/clean up his massive "Spam" folder (many thousands of mail messages).

He uses Thunderbird and has it configured to use "Move to Trash" when he deletes messages.

I find the following in Dovecot's log file when it happens:

Aug 28 10:38:22 ifm.liu.se dovecot: [ID 107833 mail.error] IMAP(bpn): Timeout while waiting for lock for transaction log file /home/bpn/Maildir/.Trash/dovecot.index.log

This happens if mailbox synchronization has been running for 2 minutes by one process and another process is also trying to sync the mailbox. Synchronization is also done when committing copied/saved mails, like:

• lock log file
• rename() all new files to destination mailbox's new/ or cur/
• unlock

I guess the rename() stage takes a long time with that user. I hadn't really thought before that this could be a problem.

Any suggestions on what I can tuned/fix in Dovecot to solve this issue?

You could grow the hardcoded timeouts.

If you're using dotlocks, src/lib-index/mail-transaction-log.c:

#define LOG_DOTLOCK_TIMEOUT 60 #define LOG_DOTLOCK_STALE_TIMEOUT 60

Otherwise src/lib-index/mail-index-private.h:

#define MAIL_INDEX_LOCK_SECS 120

It's a bit stupid that these are using different defines with different values. I'll go fix that for v1.1.