Clustering Zimbra

[Dionisas]

Hello, all,

I am new here, but got some questions. So maybe you can help me in finding out answers.

1. We are planing to use Zimbra for 5K users, 15TB (10Gb per user) space, with all Zimbra futures, mobile, outlook connectors, and etc. What type of architecture you could recommend to use. Will it be enough for good performance, use 1x1 clustering soliution, on HP DL385 G2 2,2Ghz, 4Gb RAM and HP MSA StorageWorks with RAID 1+0? If not why? What will you suggest? Service expandation?

2. Is it possible for clustering soliution to use not SAN storages? If no? Why? Is it possible some workaround?

3. If we will make some soliution, and we deside to use not 15TB, but 25Tb, will it be possible to add more?

4. Clustering soliution as we so is only active - passive. Is it possible to make active - active? Because in some of docs i've found this statement:

Future: Native Replication. Zimbra clustering can be characterized as “single copy” failover in the sense that, from perspective of ZCS, there is only a single disk image to worry about (RAID mirroring is hidden to Zimbra). The Zimbra Collaboration Suite will in the future also provides “dual copy” failover that does not rely on underlying shared disk. ZCS native replication is likely the better fit for failover across a WAN, such as for multi-site disaster recovery wherein shared network attached storage may be infeasible. Zimbra replication is also well suited to commodity server and storage hardware.

5. How about domain level qouta restriction? Is it possible? If no, when do you think to implement that?

6. Is it possible to use different COS in Domain level? For example a domain admin can create mailboxes, by differend COS.

7. Is it possible to use another Clustering method not RHEL, and still get support for Zimbra app?

Thanks,
For help.

[Klug]

1. You should check with sales and/or a local Zimbra VAR.

2. The problem with such a big setup will be disks IO (disks IO is _the_ bottleneck in Zimbra and such apps), I'm not sure you can achieve it with DAS. So SAN might be the only solution available. Anyway, see point 1.

3. Yes, you can add storage volumes on the fly.

4. Not yet and no ETA.

5. Not yet : Bug 6380 - Domain-wide Quota

6. I'm not sure I understand the question. You can have several COS for a single domain. And/or you can have one different COS per domain.

7. You can setup your Zimbra on a virtualisation infrastructure and use the virtualisation failover systems (vmWare and its VMWare HA).

[Dionisas]

1. Does Zimbra have any standart API's for managementing users/groups? Now web admin console. But some web comandline or smth else. For we have Service Desk, and some simple job as resizing user mailbox quotas, adding/deleting users and etc. we would like our SD to do. But we are wondering is it possible to use any standart Zimbra API? Because admin console for SD wont be available.

2. We desided to make 1x1 clustering soliuton on RHEL 4, as you suggested. But later if our server load will be big, we are thinking of services expandation. So if we expand for example LDAP service, can we do it on openSuse OS? Clustering soliution still be working i supose. Or not?

3. Maybe you know avarage of active server downtimes, while upgradint Zimbra? Is it possible upgrade/update to be done in 5 min for example? because we must ensure HA for about 99,5. Because as i understood from forums and other links. All bux fixes, and patches are making as system upgrade to new version. It is not possible just to run any simple script to fix bug and finish. Am I right?

4. If we desided to use 1x1 Clustering soliution with SAN storage and HSM. Can we use MSA for HSM? Or MSA is not possible at all because it is not shared storage?

[Klug]

1. Everything (user creation, etc) is available in SOAP, you should not have any issue using them from your Service Desk.

2. Actually I don't know how you can expand a clustered Zimbra : I don't think Zimbra is "different" (about expansion) once clustered but I'm not sure. You can also cluster each part of Zimbra separetely.

3. Upgrade downtime depends on the server power and "size of upgrade" : some upgrades are just binaries upgrades while other changes things in the database. I'm not sure you can achieve it in less than 5 minutes (especially with lots of users/data).

4. I'm using iSCSI storage for HSM (and backup) and it works very well. Which MSA enclosure are you thinking of using ? You can not use a single-attached enclosure in a clustered server, you need a double-attached one (or a FC or iSCSI SAN). However, you don't need GFS or any kind of shared filesystem because it's an active/passive cluster.

[Dionisas]

Thanks for answers.

1. Where could I find info's about Zimbras SOAP?

[Klug]

Some documentation is here : http://zimbra.svn.sourceforge.net/vi...soap-admin.txt

You can also search the forum and wiki (someone created PHP classes and someone else some perl code).

PHP SOAP vs Zimbra
Perl Module to simplify SOAP access to Zimbra