Zimbra offers Open Source email server software and shared calendar for Linux and the Mac
Go Back   Zimbra :: Forums > Zimbra Collaboration Suite > Administrators

Welcome to the Zimbra :: Forums!
Welcome, if you would like to post a comment please register. We also encourage you to explore all things Zimbra with our team and members of the community.

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
  #1 (permalink)  
Old 07-23-2009, 08:38 AM
Starter Member
 
Posts: 1
Default Mail Stores via CIFS issue & advice?

Hi,

I have been testing ZCS OS and we are looking at rolling this out as a solution in the future.

I am running ZCS OS 5.0.18 on Ubuntu Server 8.04, running as a vm on ESX 3.5 (soon esx4).

I have been looking at ways to host the mail stores that will provide a flexibility for growth, backup and administration.

The 1st option I have looked at is hosting this on our NetApp San via CIFS. I have mounted a share and was testing. I found this performed fine while using the zimbra web interface, but as soon as I moved to external IMAP clients (outlook 2003, Thunderbird etc), I started to get issues.

When moving messages into other folders I'd get;

the mail server responded: UID COPY Failed - in the client.

and

imap - UID COPY failed
com.zimbra.common.service.ServiceException: system failure: Caught IOException executing com.zimbra.cs.imap.TcpImapHandler@1522e40
ExceptionId:ImapSSLServer-4:1248347888210:529fe58d136a4262

in the zimbra mailbox.log.


As soon as I moved the Zimbra mail stores back to the local disk (/opt/zimbra/stores) all imap clients behaved normally.

Doing some digging this looks like it's related to zimbra's use of hardlinks and the fact that cifs (and nfs?) don't support them. Can someone confirm this and offer any advice on using CIFS or NFS to host the zimbra mail stores?


Thanks.
Reply With Quote
  #2 (permalink)  
Old 07-23-2009, 04:00 PM
Active Member
 
Posts: 46
Default

I run my Zimbra mail store off an ext3 partition on a LVM Volume on a iSCSI Lun. If you already have the NetApp SAN backend, run it over the FC instead of the iSCSI.
If you have the snapdrive from NetApp, you can expend the filesystem with just one command. You do have to shutdown the service when you expend the lun but it will take less than 10 minutes. Even without the NetApp software, you can do this with just several steps but all can be done with about 10 minutes downtime. I use CentOS/Redhat 5 and the included system-config-lvm can resize filesystem faster than commandline resize2fs. I think it is somehow able to bypass the filesystem check requirement of the resize2fs. This can be done safely and remotely all without rebooting the server.
Reply With Quote
  #3 (permalink)  
Old 07-23-2009, 11:32 PM
Advanced Member
 
Posts: 186
Lightbulb

Hi all,
Instead of mount the NFS or CIFS file system to /opt/zimbra/store, why don't you guys try to mount it on /opt. So that the configuration data will also go to the SAN. For backup&recovery strategies this will help I guess.
Regards
Adeel
Reply With Quote
  #4 (permalink)  
Old 07-24-2009, 09:53 AM
Active Member
 
Posts: 46
Default

Yes, the lun hosts my entire /opt in my case as well.
By the way since you are running on a vm, there is another way of doing this, you would create a sperated virtual disk for your /opt partition and make sure it is on top of an LVM volume. When you need to expand /opt, you can either add another virtual disk or replace it with a bigger virtual disk. As a feature of LVM, there is no downtime require to increase or migrate physical volumes. The downtime windows is require only on the last step (i.e. resize2fs). You would choose this option over iSCSI or FC if the FC adapter is for your host system and your vm boot off the virtual disk image that live on the NetApp anyway.
Reply With Quote
  #5 (permalink)  
Old 07-24-2009, 01:20 PM
nrc nrc is offline
Special Member
 
Posts: 156
Default

Zimbra recommends against putting your mailstore on NFS. I imagine that they would never think that someone would use CIFS. Permissions, ACLs, and links don't map seamlessly from the Linux world to the CIFS world. Running your mailstore in that environment would be a bad idea.

See Zimbra's recommendations on disk performance here:

http://wiki.zimbra.com/index.php?tit...ge_Deployments
Reply With Quote
  #6 (permalink)  
Old 07-25-2009, 02:32 AM
Advanced Member
 
Posts: 186
Post

One having FC shouldn't be having only mail store on an FC share.
In all cases, /opt should be mounted rather than /opt/zimbra/store.
Consider if /opt/zimbra/store have some link with some other storage (very reliable ), and we loose /opt/zimbra due to some storage corruption or hardware failure, we still have store with us i.e. mailboxes. But the recovery is not something piece of cake as we lake the mysql data. Please tell me if some one knows about taking backup and restoring the mysql of zimbra. This will really make the recovery easy in cases in which we have only /opt/zimbra/store in some other storage.
FYI http://www.zimbra.com/forums/adminis...es-zimbra.html

Regards
Adeel
Reply With Quote
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes


Similar Threads

Why Join?

Registering let's you ask questions, makes it easier to search, displays any files attached to posts, and notifies you about replies.

blog.zimbra.com




 

SEO by vBSEO ©2011, Crawlability, Inc.