Comments on: Configure VMware ESX(i) Round Robin on EMC Storage

By: Jared P

Jared P — Sun, 20 Nov 2011 17:53:03 +0000

Hi,

Great article. I just wanted to comment that my findings were similar to what Kevin K suggested (at least for ESX 4.1). Changing the SATP for each device with the esxcli command wasn’t necessary… the host reboot step took care of it.

By: jason

jason — Thu, 25 Aug 2011 18:38:09 +0000

Active/Passive is the recommended configuration for the NS-120. Although you have one server_3 data mover sitting idle, it’s needed for failover in the event the server_2 data mover fails. IIRC the Celerra NS-120 does not support failover or redundancy in an active/active data mover configuration. All of that said, configure your iSCSI target portal(s) on the Celerra. On the ESX/ESXi host side, there are numerous articles which discuss the configuration of multipathing with VMware swISCSI: Create a vSwitch, bind at least 2 vmnics to it, create 2 VMkernel portgroups (vmknics), configure each portgroup to use 1 active vmnic from the vSwitch, and set the other vmnic to not used (don’t configure for standby). Lastly you’d use esxcli in vSphere 4 to bind each portgroup to the swISCSI adapter. With vSphere 5 (released today), this can all be performed via a GUI method.

By: Ian

Ian — Thu, 25 Aug 2011 01:04:39 +0000

I have a Celerra NS120 and the SP’s are configured active passive. How do I configure iSCSI MPIO (from ESX 4.1) and what will happen in the case of a SP failover?
Second, I keep reading to change the failover mode to 4 to support ALUA. Is ALUA supported with iSCSI MPIO? How and where do I enable failover mode 4 via Unisphere for a Celerra? I’m running Flare 30.

By: Welcome to vSphere-land! » Storage Links

Welcome to vSphere-land! » Storage Links — Fri, 08 Jul 2011 15:09:55 +0000

[...] PSP and SATP in Plain English (Stephen Foskett) Configure VMware ESX(i) Round Robin on EMC Storage (boche.net) What’s that ALUA exactly? (Yellow Bricks) Pluggable Storage Architecture, [...]

By: vSphere Round Robin MultiPathing « Phil the Virtualizer

vSphere Round Robin MultiPathing « Phil the Virtualizer — Tue, 29 Mar 2011 19:24:51 +0000

[...] http://www.boche.net/blog/index.php/2010/02/04/configure-vmware-esxi-round-robin-on-emc-storage/ [...]

By: Oliver

Oliver — Tue, 01 Feb 2011 11:10:14 +0000

Thanks for your post, especially the part with navisphere not showing the correct failover mode. navcli shows failover mode=4 (which is correct), navisphere still shows 1.
Maybe you should point out, that useANO is only necessary when using single targets per initiator. And did you set IOPS=1 finally?

By: jason

jason — Thu, 18 Nov 2010 19:25:34 +0000

@Chris
I’m not sure how older versions of ESX would see it but I have the hardware to test that out in the lab. My thought is that it’s not going to be a supported configuration betweeen hypervisor and storage vendor. When I enabled ALUA on the NS-120, it was with the specific knowledge that my hypervisor was ALUA aware and it would be a supported configuration between the two partners. The NS-120 doesn’t ship with ALUA enabled out of the box. The failover mode = 4 configuration change was required by (then) Navisphere to turn on ALUA goodness. The question we are kicking around is how does a non-ALUA aware hypervisor see a non-optimal ALUA path: Is it active, or is it passive? From an array standpoint, it’s active, but non-optimal. My guess is the hypervisor will see 1 active optimal path but will not see the non-optimal path with useANO=0. If useANO=1, then my hunch is ESX will see both paths as active.

By: Chris

Chris — Thu, 18 Nov 2010 00:19:37 +0000

Jason. This is awesome. Thank you.

Question. Let’s say you have a mixture of ESX 3.5 U5 and 4.0 U2 hosts (I know its not the most optimal setup, but it’s not my env either), since the 3.5 hosts are not ALUA aware would they still see the array as Active/Passive with Failover mode 4 while the vSphere hosts would see it as Active/Active? Let me know what you think. Thanks!

By: How to enable ALUA and Round Robin Multipathing with ESX and Clariion CX4 « Marco Galli

Tue, 16 Nov 2010 13:28:00 +0000

[...] [...]

By: Jay Halek

Jay Halek — Tue, 20 Apr 2010 13:36:31 +0000

We struggled for days trying to get this to work on a Clariion. After vmware support and EMC support, we needed to go into Navisphere in Engineering mode (ctrl+shift+f12) and check all 4 host initiators as only one was being used. another hidden feature I guess.

By: Enabling Round Robin and MPIO on ESXi 4 | i n b a u d w e t r u s t

Enabling Round Robin and MPIO on ESXi 4 | i n b a u d w e t r u s t — Sat, 17 Apr 2010 20:32:04 +0000

[...] vStorage Multi Paths Options in vSphere Configure VMWare ESXi Round Robin on EMC Storage Best Practices for HP EVA, vSphere 4 and Round Robin [...]

By: jason

jason — Wed, 24 Feb 2010 01:21:19 +0000

@Rich I think that was the key difference. The LUNs had already been presented to the ESX hosts prior to configuring them for ALUA afterwards. In this scenario, the host required a reboot to recognize the change. I suppose removing the LUNs and readding them would have also worked, however that was not an option as they contained data and running VMs.

By: Techstarts

Techstarts — Tue, 23 Feb 2010 12:08:13 +0000

Thanks Rich, will look forward Active/Active vs RR. I might be able to produce them in few weeks. I’m reading mastering vSphere and that is where my basic understanding is coming from. I read today, if you change the path selection/Load balancing policy to RR(which is not detected automatically), you have to manually set it SATP. Hope it helps.

By: Rich

Rich — Sat, 20 Feb 2010 18:40:57 +0000

@Techstarts In my testing, I would agree with Kevin K. The SATP properly detected the CLARiiON in ALUA mode. But, it did not do so when I just changed the Failover Mode for the host. I had to unpresent and represent the LUN to have it detected. I’m guessing, as Kevin K. mentioned, that the evaluation of the rules only happens during the initial scan or at boot.

I’d also like to mention that in my testing, Round Robin provided a significant performance increase over MRU. This was on an HP EVA which is active/active and which ESX sets to MRU by default. I haven’t had an opportunity to compare MRU/RR/PP on the CLARiiON yet.

By: Techstarts

Techstarts — Thu, 18 Feb 2010 08:30:52 +0000

I’m kind of surprised to read that SATP is not detected automatically. In fact it is NMP’s function to detect array type and apply specific SATP. either i have not understood correctly or something is wrong in NMP. NMP must detect the array type. i.e what
esxcli nmp satp list –signifies.

Many thanks for this post. I believe Round Robin is not better without PowerPath VE, in fact Active/Active best and reliably if you have high consolidation ratio.