Lun Replication manual failover steps ?

[galibai]

Hi, I have gone through various sites about DR - Lun Replication. I have kind of idea about this. But I need to be sure that the steps that I am taking is right.

We have 2 datacenter - One is LA and other is NY. We are going to use SRM and also manual failover (because the LUNs are getting replicated). We dont want to put too much of money on SRM. So we thought we can use the Lun Replication and register the VMs at the other site.

Steps that needs to be taken, if they are going to be bi-directional site

1) Let us say if LA Datacenter has a disaster and we need to get up and running at NY what needs to be done ?

- Break the replication

- Not sure if this step is right, please tell me if this is wrong - On each ESX host in the Cluster where these VMs are going to poweron I will first have to EnablResignaure set to 1. (Do I need to this on all the nodes in the Cluster OR just one node will do the trick?

After the resignature, I will scan for the new Luns and then Once I get that I will register them to my VC server and power-on the VMs at NY.

Is this sufficient OR there is more to this? Also, how can I get the VMs back to LA?

Thanks

[weinstein5]

To be honest I would trust SRM - it is designed to do this for you and you have the ability to test you DR plan without affecting production while all these manual steps will impact your environment - What are the concerns around SRM?

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[galibai]

SRM is going to cost us money. - Licenses. In this bad economy every dollar is very imp.

Thanks

[kjb007]

Step 2 (resignature) will rename your datastore. If your vm's were previously registered, they will have to be removed, and re-registered using the new datstore name. I, instead, set the 'DisallowSnapshotLun' to 0, effectively disabling the disallow. This will allow you to attach the replicated LUNs to ESX, and rescan, and your datastore will be visible.

SRM is, of course, the better option, but I understand your point on costs.

-KjB

VMware vExpert

Don't forget to leave points for helpful/correct posts.

[AndreTheGiant]

See also:

Andre

[galibai]

Automation is good. But first let me sure about the manual steps. I am still not clear on this topic.

Thanks

[galibai]

The more, I read the more I am getting lost. I would sincerely appreciate some help here.

As per this link - http://www.yellow-bricks.com/2008/12/11/enableresignature-andor-disallowsnapshotlun/

I should be using option 3. - Enable Resignature = 0 and disallow snapshot = 0. But what will happen when I need to put the Luns from NY back to LA?

Any ideas? How are people managing such complex setup ?

Thanks

[kjb007]

The disallowSnapshotLun option is the one I mentioned above. It will allow you to add/remove your LUNs, and allow your ESX host to see those added LUNs without additional work required to make the datastores visible.

When you need to add the original LUNs back to the servers, you stop the vm's, remove the DR LUNs, and add back the original LUNs. You'll have to make sure you allow replication data to sync up again, if you are writing new data on the DR LUNs.

To put it short, you set DisallowSnapshotLun to 0, and add / remove LUNs as required.

-KjB

VMware vExpert

[galibai]

So here is what you are suggesting. I will write all of them in steps so that it can be helpful to my and rest of the people who are planning this.

Scenario - Luns are getting replicated both ways from NY to LA and from LA to NY.

There are multiple Clusters both in LA and in NY and I have tried that both of them are matching at least for the number of hosts.

Let say there is a disaster in LA and now my manager wants to move all the VMs from LA to NY -

Here are the steps that I need to take for DR

1) Ask the Storage team to stop the replication ( break the mirror)

2) Once the mirroring is broken I go to my one of the host in NY Cluster and select the DisallowSnapshot to 0.

3) With the above option set, I rescan the Luns and then I should be seeing all those Luns which were presented to LA DataCenter, right ?

4) I register all the VMs

5) Power-on all the VMs

Lets say, now I want to move all the VMs back to LA, steps that I will take

1) Re-establish the mirror

2) Power-down the VM's from the LUN which were originally hosted in LA.

3) Once the VMs are powered-off remove them from inventory

4) From the host in the LA Cluster - Rescan the Luns ( Not sure if this is right step)

5) Register the VMs back in VC

6) Power-on the VMs.

Are these steps right ?

Thanks all for your help

[kjb007]

For failback, when you are talking about moving back to LA, you are using a LUN which I believe was part of the replication before. Before you can re-establish the mirror, make sure the direction of the mirror is correct, since you were running in NY, the original LA vm's were writing and updating data. Do you need those updates? If so, you'll need to sync the data first, before shutting down, and then re-establishing the mirror.

Rescan is good, since you broke/re-established the mirror.

Since the vm's were already registered in the LA cluster, you will not need to re-register the vm's. They should already be there.

-KjB

VMware vExpert