seamusob
Contributor
Contributor

San Disaster Recovery

I have a client who want to protect themselves against san failure within the same datacenter they do not want to replicate to a dr site. To that end they have bought two HP EVA's with san replication between the two. Both these EVA's will be presented to the same ESX cluster all hosts will be running ESX 4

Two vmfs luns will be used to host the vmdk files and these will be replicated to the second HP EVA via block level san replication.

I think I know the answer but I just wanted to throw it out for verification before we test the solution the process as I see it is below

1. One san blows up all production luns dissapear.

2. Replication is halted on the san and the copied data is marked active

3. A rescan is done on the ESX hosts so that they can pick up the replicated luns

4. Give new signature and mount lun

5. Import guests using vcenter.

Any advice on this approach would be appreciated

0 Kudos
8 Replies
idle-jam
Immortal
Immortal

looks good, but i would advise of going to vcenter site recovery manager (SRM) which would automate the whole tasks. and it is not really expensive as it goes by the number of VM required for protection.

seamusob
Contributor
Contributor

I agree but this will be in the same datacenter and they do not have the money for srm this might come later. They had a massive san failure last year which is the reason for now they are going down this route

0 Kudos
idle-jam
Immortal
Immortal

all is ok except for 5. Import guests using vcenter. you are suppose to go to each of the folder and register the .vmx as a virtual machine.

seamusob
Contributor
Contributor

yes you are right though I am going to try and do it with powercli and script it as much as I can I do understand though that I have to browse the datastores and import the vmx file that is what I meant

0 Kudos
squebel
Contributor
Contributor

Just my 2 cents here, but if they are forking out that kind of investment for two arrays with synchronous replication, why not just spend a little more and do the SRM solution so that way everything is automated (to an extent) and can be managed easier than worrying about scripts and whatnot. Plus, documentation that any baffoon can follow could be written to hit the "Go" button not to mention including support from VMware. Like I said, just my 2 cents. Might be worth another mention to your customer.

0 Kudos
seamusob
Contributor
Contributor

I am going to speak to them tomorrow about it but you are definitely right. I will let you know how I get onSmiley Happy

0 Kudos
JohnADCO
Expert
Expert

Are you sure this is the way to set this up?

I only ask because our new san arrays handle this as part of their HA.   The hosts only see some of the paths drop off.   Two completely separate independent san arrays though they do have a direct connect 10G connection between them.

0 Kudos
bulletprooffool
Champion
Champion

Seems a reasonable plan - and something that I have implemented before.

The thing to do is script the failover - possibly run a nightly report to capture resource Pools etc as a quick import will not carry this information.

I scripted for this scenrio before and had full cluster failovers running in minutes - but the script was one of my first PowerCli scripts and was VERY badly written. When I get round to re-writing it, I will post it.

In the mean time though, have a look at Gabe's Virtual world http://www.gabesvirtualworld.com/cheap-disaster-recovery/  - he has a small collection of PowerCli scripts that should manage this for you.

One day I will virtualise myself . . .
0 Kudos