Change multipath policy to Round Robin on ESX 4.1 - any caveats?

[MichaelLeone]

I am in the process of upgrading my 6 host ESX 4.1 cluster to ESXi 5.0. I have upgraded one server so far, and - as part of the upgrade - I am changing my fiber channel datastore multipath policy to Round Robin from MRU (Most Recently Used), before exiting from maintenance mode. 

I should be able to do this on my ESX 4.1 hosts now, before upgrading them to 5.0. Are there any caveats to be aware of? I don't need to be in maintenance mode, right? And I shouldn't need to do this off-hourse. I *should* just be able to go to every datastore on my 4.1 hosts and change the multipath policy on the fly, without any negative impact to my environment.

Correct?

Just want to double check, before I go barging ahead ... 🙂

[MauroBonder]

Yes, you´re correct.

If you checked that your storage works best with RR you can change under Configuration> Storage> Properties of each LUN> Manage Path> Select RR> Change, without impact.

Can you check what is better policy for your storage here ->www.vmware.com/go/hcl

you need do this procedure for each lun .

[farkasharry]

Even if you can change the PSP for running hosts, If it is not a 30 host cluster in production, I would put the host first in maintenance mode just to be on the safe side.

[MichaelLeone]

You don't need to put hosts into maintenance mode, I did all my hosts (6) live, with no ill effects. And RR is available for all arrays, according to the VMware documentation. In my particular case, that's 2 HP EVA 8000s, and a EVA 8400. In fact, RR is the recommended multipath for EVAs.

[farkasharry]

I wrote: "Even if you can change the PSP for running hosts..."
You can change it, but depending on the applications running within the VM, it is for me always better - if possible with not too many effort - to do anything like this in maintenance mode. I might be over-careful but experiences tought me to be it!

[mcowger]

MichaelLeone wrote:

And RR is available for all arrays, according to the VMware documentation.

I'm pretty sure thats *not* the case.  Can you provide a link to the documentation stating that RR is acceptable for all arrays?

[MichaelLeone]

• KB Article: 1011340


• Updated: Jan 20, 2012

"Switching to Round Robin from MRU or Fixed is safe and supported for all arrays, Please check with your vendor for supported Multipathing policies for your storage array. Switching to a unsupported pathing policy can cause an outage."

<http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=101134...>

[mcowger]

Wow.

Thanks for pointing that out, I will address with VMware directly, because thats not even close to correct.

[Troy_Clavell]

I think the key statement in the KB's you linked is below.  Not ALL arrays support Round Robin.

Please check with your vendor for supported Multipathing policies for  your storage array. Switching to a unsupported pathing policy can cause  an outage

[mcowger]

Yeah, but right above that, they say: "Switching to Round Robin from MRU or Fixed is safe and supported for all arrays", which is sillyness.

[MichaelLeone]

In my particular case, my arrays are HP EVA 8000s, and an EVA 8400. And RR is supported, and is in fact the recommended multipath.

[mcowger]

Agreed - for your EVAs thats true.  But there are plenty of arrays for which its not.