Zimbra Open Source Clustering - SAN

[maceee]

Hello to everyone,

I am intenting to go with Zimbra Open Source Edition.
I have 3 servers.

I want to install this on HA environment and have some questions.

I am between 2 solutions:

A) 2 Openfiler Servers San Storage (RAID 10 - ISCSI) with drbd and heartbeat, and 1 Server with Zimbra Open Source.

B) 2 Nodes (Local RAID1) with drbd and heartbeat and each one one instance of Zimbra Open Source.

Which you think would be the best setp?
I need mostly redudancy

What else you have to propose?

Thank you very much

[Krishopper]

Considering that clustering isn't supported in the OS edition, you are on your own for getting the "cluster" setup and working.

But between the two options, having two Zimbra servers with DRBD is probably your better bet, since you'll have another server ready to go if your mailbox server takes a hit.

[maceee]

Thank you for your reply.
Is anywhere any doc for clustering at open source?

thank you

[phoenix]

Is anywhere any doc for clustering at open source?

There isn't any Zimbra documentation because it's not supported.

[raj]

B) 2 Nodes (Local RAID1) with drbd and heartbeat and each one one instance of Zimbra Open Source

is a better option

Raj

[Klug]

Better than VMs hosted on SAN?

[raj]

Better than VMs hosted on SAN?

Yes..coz he wants to use iSCSI SAN..and the if you look into the Data Transfer on i/o for VM i/0 + DRBD i/0 he will double/triple up the iSCSI bandwidth and that itself is more flaky compared to 2 solid local datastore's and only 1 network usage of DRBD

He asked for a choice and given my experience in both i can tell you that option B is more stable, cheaper and less complex to setup

PS: not to mention that he should be worried more about the DRBD instead of anything else

Raj

[Klug]

In my own experience, iSCSI SAN (OpenFiler) + computing nodes (XenServer or ESXi) is easier to setup than DRBD 8)

But I agree DRBD would be faster (without bonding).

I also agree that DRBD is the thing he should worry...

[maceee]

thank you for all your replies.
I did talk with one vey experienced Zimbra administrator who well knows the system.

The first he did tell me is that topology B (Node A- Node B) each one one zimbra instance is not officially supported from Zimbra. although i have seen here a how to.

I think that in both topology a and topology b i need to run drdb or am i wrong?

first of all we talk about finally not open source but starter pack edition.

what suggested at me is one SAN box (openfiler) and 2 nodes running down of it with 2 instances of Zimbra. This way both instances will share same storage but how wou ld i implement all other clustering services zimbra needs as postgresql? Don't i need there also drdb and heartbeat between node a and node b???
I n this situation except my question if i need clustering suite betwwen node a and node b my concern is that if my SAN fails then all system will go down.
So second thought for this scenario is also SAN to be replicated too.
So to summarize with plan a:

A) node a - node b with 2 zimbra instances and drdb - heartbeat???
san a - san b with drdb and heartbeat so that storage is replicated.

My finally question here is if san a goes down and i bing san b up will zimbra keep working with storage in san b???
I am really curious about this one.
If this work i can lose one zimbra nad one san node without my system to goes down i think.


TOpology b is much easier i think.

B) node a - node b with 2 zimbra instances.
Here i will need for sure drdb and heartbeat between these 2 i suppose.Does it officially supported ot can work out of the box? I have seen sme how to here in forum.


if anyone have a better solution kindly let me know
Many thanks for your answers



* Also forgot to say.
Node a andnode b of zimbra except drdb will have local raid 1.

[maceee]

Anyone pls?