dovecot 2.1.x and 2.2.x has "just worked" with a minimal configuration for just about anything I've ever thrown at it.

Now, with 2.3.0, there is one user (and only one user) that continually sees horrible delays, which is not surprising given the logs:

Error: imap(user)<70175><356C8oljxrCBUppP>: Timeout (180s) while waiting for lock for transaction log file /path/to/home/directory/.imap/dovecot.list.index.log (WRITE lock held by pid 69995)

where 69995 is the pid of the first dovecot process and 70175 is the pid of, e.g., the first LIST command.

What's unique about this user is that he's the only one using linux (thunderbird) and his home directory is NFS-mounted.

I understand that something is incorrect with the locking, but there are a couple of issues that I do not understand:

1. Is there a default that changed between 2.2 and 2.3 that would cause this? Despite potential issues associated with NFS, I didn't see the problem with 2.2. There is nothing in my conf file related to locking, so if there was a change, it must be a default for 2.3 (from doveconf -a):

dotlock_use_excl = yes lock_method = fcntl mbox_dotlock_change_timeout = 2 mins mbox_lock_timeout = 5 mins mbox_read_locks = fcntl mbox_write_locks = dotlock fcntl

2. If the problem is related to NFS locks, why would the problem persist when the user runs thunderbird from the server where his home directory *is not* NFS-mounted?

Thanks for any help.

Cheers, Zube