dovecot - Apr 2012 - old messages being delivered to blackberry

Several users are reporting old messages (sometime years old) are being 
delivered as new on their blackberry devices... I've confirmed that this 
happens to all blackberry users... I believe the problem is related to 
this error:

Maildir: Expunged message reappeared, giving a new UID

In fact, this is the only mail related error message on the server... 
the mail is stored on a glusterfs share, and I'm using director to 
prevent simultaneous access... although I wasn't able to get LMTP to 
work with Director...   to simplify things, I've temporarily only 
enabled 1 dovecot server anyway, to get everything else working first... 
I'll fiddle with lmtp later... so right now only 1 server is accessing 
the mailboxes.  Delivery is also done from that one server, from postfix.

Possibly caused by GlusterFS, the indexes were still getting corrupted, 
even with only 1 server... so I've moved the indexes to a local 
directory on each server.  I'm no longer getting index related errors... 
but the expunged error is still occuring... and the error seems to 
happen a few moments before the BB gets a bunch of messages... so it's 
very likely related.

I would like to continue using GlusterFS if possible; I haven't found 
any indication that it is outright unsupported, as long as I'm using 
Director...

I am using dovecot 2.0.16.

Any help would be appreciated!

Here is my configuration:

# 2.0.16: /etc/dovecot/dovecot.conf
# OS: Linux 3.1.9-1.4-desktop i686 openSUSE 12.1 (i586)
auth_mechanisms = plain login
disable_plaintext_auth = no
log_timestamp = "%Y-%m-%d %H:%M:%S "
mail_fsync = always
mail_location = maildir:~/Maildir:INDEX=/var/indexes/%u
mail_nfs_storage = yes
mail_privileged_group = vmail
managesieve_notify_capability = mailto
managesieve_sieve_capability = fileinto reject envelope 
encoded-character vacation subaddress comparator-i;ascii-numeric 
relational regex imap4flags copy include variables body enotify 
environment mailbox date ihave
passdb {
   args = /etc/dovecot/dovecot-sql.conf
   driver = sql
}
plugin {
   quota = dict:user::file:/var/vmail/%d/%n/.quotausage
   sieve = /var/vmail/%d/%n/.sieve
}
postmaster_address = postmaster at bigdata.nb.ca
service auth {
   unix_listener /var/spool/postfix/private/auth {
     group = postfix
     mode = 0660
     user = postfix
   }
   unix_listener auth-userdb {
     group = vmail
     mode = 0666
     user = vmail
   }
   user = root
}
service lmtp {
   inet_listener lmtp {
     port = 24
   }
   unix_listener /var/spool/postfix/private/dovecot-lmtp {
     mode = 0660
     user = postfix
   }
}
ssl_ca = </etc/postfix/smtpd.bundle
ssl_cert = </etc/postfix/smtpd.cert
ssl_key = </etc/postfix/smtpd.key
userdb {
   args = /etc/dovecot/dovecot-sql.conf
   driver = sql
}
protocol imap {
   imap_client_workarounds = delay-newmail
   mail_plugins = quota imap_quota
}
protocol pop3 {
   mail_plugins = quota
   pop3_uidl_format = %08Xu%08Xv
}
protocol lda {
   mail_plugins = sieve quota
}
protocol lmtp {
   mail_plugins = quota sieve
}

Luc

On 10.4.2012, at 16.45, Luc Belliveau wrote:
> Several users are reporting old messages (sometime years old) are being
delivered as new on their blackberry devices... I've confirmed that this
happens to all blackberry users... I believe the problem is related to this
error:
> 
> Maildir: Expunged message reappeared, giving a new UID
This means:

1. readdir() returns maildir file X
2. Later readdir() doesn't return file X, so Dovecot thinks the mail is
expunged and also tells the client the mail is expunged
3. Later readdir() once again returns file X, so Dovecot can't really do
anything except treat it as a new mail.
> Possibly caused by GlusterFS, the indexes were still getting corrupted,
even with only 1 server... so I've moved the indexes to a local directory on
each server.  I'm no longer getting index related errors... but the expunged
error is still occuring... and the error seems to happen a few moments before
the BB gets a bunch of messages... so it's very likely related.
> 
> I would like to continue using GlusterFS if possible; I haven't found
any indication that it is outright unsupported, as long as I'm using
Director...
If glusterfs can't do readdir() reliably, there's not a whole lot you
can do.. Except maybe switch to another mailbox format. But if you're
getting index corruption also with 1 Dovecot server, it doesn't look very
promising.