[Dovecot] Ring SYNC appears to have got lost, resending after upgrade

[/dev/rob0]

This issue still occurs. It varies which of the four director 
instances gets it, but it seems that once one of them does, the only 
fix is to restart all four.

On Mon, Sep 09, 2013 at 07:41:10AM -0500, /dev/rob0 wrote:
> On Mon, Sep 09, 2013 at 11:13:36AM +0200, Patrick Westenberg wrote:
> > on Saturday I upgraded two dovecot servers from squeeze to wheezy 
> > and dovecot from 2.1.x to 2.2.5 (compiled from sources). After the 
> > upgrade everything worked fine at first.
> > 
> > On Sunday Morning I recognized these errors (they occured after a
> > reload for logging purpose on midnight) on one server:
> > 
> > director: Error: Ring SYNC appears to have got lost, resending
> > 
> > After reloading/restarting both dovecot services the error occured 
> > on both servers. After some research I deleted some "zlib"-File 
> > which isn't needed anymore in dovecot 2.2.x and reinstalled 
> > dovecot. The error message disappeared.
> > 
> > Today the error occured again (after the reload on midnight) and 
> > again on one node only until reloading/restarting the other node 
> > too. However, there is an additional error message:
> > 
> > Sep 09 10:27:07 director: Error: Ring SYNC appears to have got 
> > lost, resending
> > Sep 09 10:27:10 director: Panic: file login-connection.c: line 88
> > (login_host_callback): assertion failed: (strncmp(request->line,
> > "OK\t", 3) == 0)
> 
> I had the same issue (CentOS 6.4 upgraded with third-party RPMs) on 
> Thu/Fri, and I asked Timo about it in IRC. Apparently a 2.2.6 release 
> is due soon. He gave me two hg links claimed to fix it:
> 
> http://hg.dovecot.org/dovecot-2.2/rev/f7a37b169f4a 
> http://hg.dovecot.org/dovecot-2.2/rev/9531ec8afe8b
> 
> However I did have the lost ring SYNC error recur after the cluster 
> was upgraded to the RPM packages currently in Dovecot's EE repo 
> (non-free, pay for access) which does include these fixes.
> 
> Restart of all director instances worked for me. Actually I stopped 
> all, then started all.
> 
> So far so good. We're going to go live with this cluster soon, I 
> hope.

-- 
  http://rob0.nodns4.us/ -- system administration and consulting
  Offlist GMX mail is seen only if "/dev/rob0" is in the Subject: