Problem recover replica

Manel Gimeno Zaragozá magiza83 at hotmail.com
Fri Feb 3 03:51:17 EST 2012


Hello agin,



Is there any option to check the integrity of the replica information? I
 mean, some command or script to run and validate that Master and 
Replica have the same information.



Regards.

Manel Gimeno Zaragoza
magiza83 at hotmail.com


From: magiza83 at hotmail.com
To: brong at fastmail.fm
Subject: RE: Problem recover replica
Date: Thu, 2 Feb 2012 16:13:10 +0100
CC: info-cyrus at lists.andrew.cmu.edu







> On Thu, Feb 02, 2012 at 01:11:24PM +0100, Manel Gimeno Zaragozá wrote:
> 
> What version?



I'm running Cyrus 2.4.13.

> 
> > Everything is working fine between Master & 
Replica until Replica goes down (maintenance or whatever). Once is down,
 in master there are still connections and modification. If I bring up 
Replica the changes made during the "blackout" are no fully transmitted 
to it. Some of the are transmitted, but not all.
> > 
> > Which is the best way to resincronize the missing parts to Replica? 
> 
> sync_client -r -f $file for each file in the conf/sync/ directory.



I've been trying sync_client -r -f log-PID and it works. All the missing changes are applied 
without problem, I should do some script that controls this files and apply them.
anyway the log-PID is not deleted when its processed. I guess we should control the processed log-PIDs manually and move or delete them, am i right?


> 
> > I was wordering to rsync imap folders 
(/var/spool/imap & /var/lib/imap), and It works fine in my test 
environment, but in produccion it could be GB to sincronize and I think 
this is not optimum.
> 
> No, that's awful.  Don't do that.
> 
>
 > Other option was "sync_client -l -u <all_user>" but I'm 
guessing that in production I will have the same problem due to the 
amount of data to sincronize.
> 
> It's not too bad.  We run it occasionally if we suspect things have
> got confused for other reasons.

> 
> > On the other hand, I have also doubts about how 
"sync_client -r" and "sync_server" are initiated. In my test machines I 
should run manually them each time I start cyrus-imapd, I though that it
 could be automatic one I start the service cyrus-imapd, but it is not.
> 
> I have attached the script that we run.  It has lots of hooks into
> our own systems of course.
> 
> I would LIKE to make something more generic that's part of Cyrus
> itself, rather than having external tooling.  Of course, our external
> tooling is a bit special-cased as well.  But at least having a basic
> "keep trying to replicate all the records" would be good.
> 
> Bron.



Thanks for the response.



Manel.

Manel Gimeno Zaragoza
magiza83 at hotmail.com


 		 	   		  

----
Cyrus Home Page: http://www.cyrusimap.org/
List Archives/Info: http://lists.andrew.cmu.edu/pipermail/info-cyrus/ 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.andrew.cmu.edu/pipermail/info-cyrus/attachments/20120203/1ef77421/attachment.html 


More information about the Info-cyrus mailing list