'Auto' Network when testing

[jftwp]

We have finished our initial config/setup of SRM (1.0 U1) and we're almost ready to run our first test. On page 61 of the SRM Admin Guide, it states:

"Select the network to use during recovery plan tests, or select Auto to create an isolated network from the Test Network drop‐down menu. If the recovery plan includes virtual machines that run Linux and are DHCP clients, be sure that the test network includes a DHCP server. If your VI cluster at the recovery site uses VMware HA, you must create and use use a test VLAN that spans ESX hosts. HA clusters do not work correctly on other test network configurations."

My recovery site resources are indeed using HA. Does this mean I CANNOT use the 'Auto' network to create/use an isolated network for running RP tests? Why? The guide doesn't really elaborate any further about this.

If I do need to create my own isolated network, then I take it I just need to create a new vswitch (with no physical NICs in it) across all ESX hosts at my recovery site, and use that network instead of the built-in SRM 'Auto' network? Thanks for any clarification/explanation regarding this.

[Michelle_Laveri]

Here's the why....

When you use "auto"... the system will create an INTERNAL vSwitch with a portgroup called "test-bubble" nettwork.

Internal vSwitches do NOT have physical nics attached to them.

So if VM-1 needs to speak VM-2. During the recovery plan test, the VM-1 might come up on ESX1, and VM-2 might come up on ESX2. As such they will not see each other and will not communicate. Other functionality such as VMotion, DRS, HA which relies on vSwitch which has vmnic mapped to it will not function.

Try to see this "test-bubble" as safety valve. If some runs a recovery plan (without engaging their brain) they will not accidentally create IP and NetBIOS name conflicts.

Ideally, you should use the inventory map - to map your portgroup settings on the protected site (vlan10) to a valid portgroup on the recovery site (vlan50) - as well as re-IP-ing the VM. This means normal functionality (VMotion, DRS, HA) will work in the recovery site as normal...

Make sense?

Regards

Mike Laverick

RTFM Education

http://www.rtfm-ed.co.uk

Author of the SRM Book:http://stores.lulu.com/rtfm

Free PDF or at-cost Hard Copy

[jftwp]

Yes, all makes sense and I understand all. I've done the inventory mappings and so forth, so that in a REAL failover, which is hopefully never, the VMs will come up on the expected production virtual switch resource/s common/consistent throughout the recovery site's cluster.

However, the way it's worded in the Admin guide suggests that you can't just set to 'Auto' for a test SRM when the recovery site uses HA. I really could care less whether HA 'works' when I'm just trying to test my recovery plans to get the VMs running within-yes, absolutely desired-an isolated test bubble network. When I'm testing, I really don't care if they talk to each other or not. For me, it's not necessary at all for testing these VMs---I just want to see them all 'turn on' as spec'd in the plan/s. I have great confidence that IF/when I run a 'real' plan to failover, the mapped inventory config will bring them up on the specified port group/s that service my production VMs in the recovery site today/now.

So then, can I just use 'Auto' after all? Perhaps I'm reading 'too much' into the way the Admin Guide words that particular sentence and its relation to HA being in the recovery site's cluster, etc. etc.

[admin]

I guess the important thing to remember is that if you use the Auto option and have multiple VMs come online at the recovery site on different ESX hosts they will not be able to communicate with one another. We have created a dedicated and isolated VLAN that all of our recovery site ESX hosts have access to and use this VLAN for our test scenarios. This way we can have multiple VMs come up on different hosts and they have the ability to communicate with each other.

I have not had a chance to play with SRM 4 on vSphere so I can't comment as to how a distributed vSwitch would play in here, but that should also be an option that may not require a dedicated/isolated VLAN.

-alex

[Michelle_Laveri]

Yup, i can see how the term "auto" can be un-nerving.

for me I want my recovery plans when "tested" to be as realistic as possible... and that not just confirming the VMs come up and wot not. But they communicate and behave with other as expected - and that for me includes things like if vms must start in the right order for muilt-tier application to work problem - from a service dependency perspective...

Regards

Mike Laverick

RTFM Education

http://www.rtfm-ed.co.uk

Author of the SRM Book:http://stores.lulu.com/rtfm

Free PDF or at-cost Hard Copy

[Michelle_Laveri · Accepted solution]

Another thought.

I guess this question - raises a whole another issue....

What constitute a real "test" of a recovery or DR plan. Click the test button... OR bi-annually, clicking the run button - and having the guys in over a weekend or longer - to actually do the failover/failback for real... Some of us look at the test button in SRM, as merely a "test" of the SRM software, than an realistic test of the DR plan.

It depends on how comprehensive you think you need to be....

Regards

Mike Laverick

RTFM Education

http://www.rtfm-ed.co.uk

Author of the SRM Book:http://stores.lulu.com/rtfm

Free PDF or at-cost Hard Copy

[jftwp]

Thanks all. I'll just be using the 'Auto' network for testing... that should suffice fine, regardless of HA being enabled on my recovery cluster resources.