[Dovecot] Duplicate Mails downloding issue

Steffen Kaiser skdovecot at smail.inf.fh-brs.de
Fri Jul 26 16:54:35 EEST 2013

Hash: SHA1

On Fri, 26 Jul 2013, Kavish Karkera wrote:

> Jul 26 12:25:37 popserver dovecot: pop3(kavish.karkera at example.com): Error: Couldn't open INBOX: Timeout while waiting for lock
> Jul 26 12:42:12 popserver dovecot: imap(kavish.karkera at example.com): Error: Timeout (180s) while waiting for lock for transaction log file /indexes//mail/v3store/example.com/kavish.karkera at example.com/.INBOX/dovecot.index.log
> it wouls be helpfull if you eloborate thsi logs in more detail. so we coluld resolve this ongoing issue.
> We have 3 POP/IMAP Servers, running as load balancers through DNS round robin.
> Dovecot version server 1 = 2.1.12
> Dovecot version server 2 = 2.1.15
> Dovecot version server 3 = 2.1.13
> Mails and Indexes are stored over NFS  and we do have these setting enabled
> mail_nfs_storage = yes
> # Mail index files also exist in NFS. Setting this to yes requires
> # mmap_disable=yes and fsync_disable=no.
> mail_nfs_index = yes

you have three servers, into which the same user logs in simultaneously, 
but they share the same storage area.

Say, kavish.karkera at example.com logs into server2 with IMAP and keeps the 
mailbox locked, because it IDLEs there. Now kavish.karkera at example.com 
logs into server1 and wants to access INBOX a second time in parallel to 
server2, which has the lock still. Because both servers do not 
communicating with each other, server2 do not give up the lock, in order 
to let server1 proceed.

If kavish.karkera at example.com would only log into server2 (or server1, it 
does not matter which server, as long as kavish.karkera at example.com do not 
log into two servers the same time), Dovecot would handle it. But because 
you use DNS round robin as load balancer, you cannot ensure this.

- -- 
Steffen Kaiser
Version: GnuPG v1.4.11 (GNU/Linux)


More information about the dovecot mailing list