Is this a valid SRM environment?

[mpovall]

I have two sites, lets call them SITE-A and SITE-B, and they have identical storage arrays. SITE-A and SITE-B both have vSphere environments, both have vCentre 4.1, and both have SRM 4.1. Both sites have production virtual machines.

Can I replicate volumes from storage in SITE-A to storage in SITE-B, replicate volumes from the storage in SITE-B to storage in SITE-A, and then use SITE-A as a recovery site for SITE-B and SITE-B as a recovery site for SITE-A, so that if SITE-A fails then SITE-B runs everything and vice versa.

Regards, Mike.

[Henriwithani · Accepted solution]

Yes you can do that. Just make sure that you have enough licenses.

Here is other discussion about it: http://communities.vmware.com/message/1204964

-Henri

[mal_michael]

Hi,

It is usually referenced as bi-directional protection and is valid and supported scenario:

http://www.vmware.com/products/site-recovery-manager/faqs.html

Additionally, make sure that at each site you have enough capacity to run "local" and recovered VMs.

Michael.

[mpovall]

Thankyou to Henri and Michael for confirming that I was not going insane.

The question was prompted by a problem that I have been having with SRM when using HP P4000 / LeftHand storage and v8.5 of the SRA.

Assume that I have installed vSphere hosts and P4000 storage in two sites with physical servers acting as vCentre servers in each site. In the CMC I create the relevant server entries for the Site A vSphere hosts, create the required volumes, present the volumes to the vSphere hosts and create the datastores. The Virtual Machines required for Site A are then created and become operational. Remote Copy snapshot schedules are configured that replicate the volumes associated with Site A VMs to the storage in Site B. SRM is now installed and configured, SRA is installed, Array Managers are configured and can see the replicated data stores, Inventory Mappings are configured, and Recovery Plans defined. When the test of a Recovery Plan is run it creates an entry in the CMC for the Site B vSphere host with a meaningless name of SRM_AG_x where x is a number. If a pre configured server entry exists in the CMC for the vSphere hosts in Site B then the test will fail at this stage.

We now have SRM implemented for Site A failover to Site B. I now create a volume in the storage at Site B, present it to the Site B vSphere hosts using the SRM_AG_x entries, create a datastore, and then create a new Virtual Machine in Site B. If I set up a Remote Copy Snapshot schedule to replicate this volume from Site B to Site A and then set up the Array Managers to recognise this replication, Configure Protection Group, Inventory Mappings and a Recovery Plan for this everything is fine. When I try to test this Recovery Plan to recover the Site B Virtual Machine on the Site A environment it fails because it can’t create the server entry in the CMC as the iSCSI initiator name already exists. I can’t get rid of the server entries as they are required for the Site A virtual machines.

HP support believes that this functionality where the SRA creates the server entry in the CMC is required to comply with a directive from VMware, though they are unable to show me any evidence of this, and that the bi-directional functionality that I expect is in their words 'challenging the design features of SRA'.

This change to the way that the SRA works appeared in v8.5 and is driving me crazy. If someone can tell me what I am doing wrong it would be very much appreciated.

Regards, Mike.

[mal_michael]

Hi,

Unfortunately I have no knowledge in LeftHand storage and cannot help you.

However, to increase the possibility of getting the answer to your  question, I would recommend you to open a new thread on the LeftHand  issue, because the current thread title is not related to the problem  and you have marked the thread as answered already.

Michael.

[TomHowarth]

Just a quick thought,  are you using a unique nameing conventions,  I seem to get the intimation thath you are not from your post.

[mpovall]

Hi Tom,

With regard to names I am assuming that you mean names of VMware hosts, vCentre servers, Storage modules, Management Groups, Clusters etc in which case I can say that all names are unique within a site and across the sites.

Cheers, Mike.

[TomHowarth]

No I was refering to the naming conventions at the array level.

[mpovall]

Hi Tom,

There are no duplications of names of anything in the environment. Volume names are related to the server that they are storing and the Remote volume name is prefixed with R- so the primary volume would be something like CRM4 and the remote volume would be R-CRM4.

Cheers, Mike.