Enable Round Robin connection between my VMware Platform and my HPE MSA 2050

[nrinaldi]

Hi, my name is Nicholas. I'm looking to connect my ESXi with my recently new HPE MSA 2050 Storage using a Round Robin policy.

Actually I'm using this storage connected to my ESXi (6.5 version) using one HBA (dual port) at 8gb/s connected directly to both of my storage controllers (A and B). Everything works fine, but when I want to change my multipathing policy from my vCenter (6.7 version) from "Recently Used" to "Round Robin" just one storage adaptor is used to transfer data (I/O tag is shown).

Am I missing something? HPE Support says that this feature should be enabled automatically, and that there is nothing to do from the Storage configurations.

Thanks!

[JWolberg82]

Hi Nicholas-

Go to the "Properties" tab under "Storage Devices" and scroll down to the very bottom to see the "Multipathing Policies" menus.  It's not obvious since it's below the visible area.

Jon

[nrinaldi]

Hi Jon. Thanks for your answer. What I meant to say is that even though I have enabled Round Robin, there is just one of the two HBA ports active.

[JWolberg82]

Hi Nicholas-

My apologies.  I just saw that. Have you tried this in the flash interface?  There are still a few things that do not translate properly to the HTML5 interface.

Jon

[nrinaldi]

I'm not sure what do you mean exactly with that. Here a screenshot of my Storage Devices from the vSphere Web Client:

I also attach a screenshot from the Storage Adapters tab:

Thanks!

[JWolberg82]

Can you try it using the "Flex" ( ala. Flash ) interface instead of HTML5?  When you browse to your vCenter it gives you the choice between the two. 

[nrinaldi]

I just edited the post above your reply with some screenshots from the Flex client.

[JWolberg82]

Have you tried setting it via command line on the node to see if that sticks?  As HP said, you shouldn't have to do that, but I've never seen the CLI fail.

VMware Knowledge Base

[nrinaldi]

I'll try and update you. Thanks!

[nrinaldi]

UPDATE: So, I tried switching to MRU and then back to RR using this command line: esxcli storage nmp device set --device naa.600c0ff0003bb76f41d56f5c01000000 --psp VMW_PSP_MRU and esxcli storage nmp device set --device naa.600c0ff0003bb76f41d56f5c01000000 --psp VMW_PSP_RR, but it still doesn't use both interfaces.

I also tried unmounting and remounting the datastorage, but still no luck.

[JWolberg82]

I don't think this is a VMware issue then.  Are you sure both controllers are active on your storage or does that particular unit run Active/Passive?  I know I have some older Dell DAS units that are dual controller but not active/active.

[nrinaldi]

That's what I'm trying to know, but at least from HPE support, they say both controllers are in Active/Active mode.

I'll let you know if anything new happens. Thanks for your time!

[nrinaldi]

I called HPE, and they showed me a shell command where I can see that both of my Storage Controllers are Active-Active:

HPE MSA Storage MSA 2050 SAN

System Name: x

System Location: x

Version: VL270R001-01

# show controllers

Controllers

-----------

Controller ID: A

Serial Number: x

Hardware Version: 1.0

CPLD Version: 1.9

MAC Address: x

WWNN: x

IP Address: x

IP Subnet Mask: x

IP Gateway: 1x

Disks: 8

Virtual Pools: 1

Disk Groups: 1

System Cache Memory (MB): 8192

Host Ports: 4

Disk Channels: 2

Disk Bus Type: SAS

Status: Operational

Failed Over to This Controller: No

Fail Over Reason: Not applicable

Multi-core: Disabled

Health: OK

Health Reason:

Health Recommendation:

Position: Top

Phy Isolation: Enabled

Controller Redundancy Mode: Active-Active ULP

Controller Redundancy Status: Redundant

Controllers

-----------

Controller ID: B

Serial Number: x

Hardware Version: 1.0

CPLD Version: 1.9

MAC Address: x

WWNN: x

IP Address: x

IP Subnet Mask: x

IP Gateway: x

Disks: 8

Virtual Pools: 1

Disk Groups: 1

System Cache Memory (MB): 8192

Host Ports: 4

Disk Channels: 2

Disk Bus Type: SAS

I don't know what else I can do...

Thanks.

[JWolberg82]

That may be the case but only one controller can "own" the vDisk at a time.  I bet if you found the proper CLI to show more in depth stats on the virtual pool/disk group that you'd see it's owned by Controller A.  It sounds to me like their Active/Active is really Active/Passive and that you aren't going to be able to use true round robin.

[JWolberg82]

https://h20195.www2.hpe.com/V2/GetPDF.aspx/4AA4-7060ENW.pdf

Some snips from best practices for MSA SANs:

"Unified LUN Presentation (ULP) failover If a controller unit fails on a dual controller HPE MSA Storage Array, Storage Pool ownership transfers from the failing controller to the secondary, or backup, controller in the array."

" Because both controllers can access data for any of the specified virtual volumes in both Storage Pools, a preferred path will be identified following the Asymmetric Logical Unit Access (ALUA) specifications. This is part of the Reported Target Port Group (RTPG) protocol which identifies the virtual volume’s pool and then identifies the optimal path to the Virtual Volume. The owning associated pool controller always performs the I/O to disk for these requests.   "

"Storage pools are associated with either of the two controllers. By default, Pool A is assigned to Controller A and Pool B is assigned to Controller B. When creating Virtual Disk Groups, the SMU attempts to distribute disk group creation evenly across these two Storage Pools. Since both controllers are active, at least one Virtual Disk Group must be assigned to each pool to balance the controllers.   "

So it sounds like some sort of ghetto balancing and not true active/active.  I'd say that the behavior you are seeing is normal.

[nrinaldi]

When I wrote to the HPE Support Team, they sent me that file. Starting in the page 28, you can see that they recommend using Round Robin, but at the screenshots you can find in the file, you can see that the RR isn't working (Because just one path is in (I/O) mode).

This is very weird. I'm very confused. I also have this file, but there's nothing useful in it: https://h20195.www2.hpe.com/v2/getpdf.aspx/A00015961ENW.pdf  (Starting at page 27)

[JWolberg82]

"ULP and vSphere Hypervisors such as VMware vSphere use ALUA to communicate with backend storage arrays. ALUA provides multipathing (two or more storage networking paths) to the same LUN on a storage array and marks one path “Active” and the other “Passive.” The status of the paths may be changed either manually by the user or programmatically by the array.  

   Multipath Considerations for vSphere To maintain a constant connection between a vSphere host and storage, ESX software supports multipathing. To take advantage of this feature, the ESX host requires multiple FC, iSCSI, or SAS adapters and the HPE MSA virtual volumes need to be mapped to these adapters. This can be accomplished easily on the HPE MSA Storage by creating a host definition as outlined in the previous section and associating the World Wide Names (WWNs) of the multiple interfaces (HBA ports) on the host server to this new host object. When mapping a Virtual Volume to the host object in the SMU, all the path mappings are automatically created to support multipath to the host. To do this in the CLI an entry for each path would need to be created or use the Host/Host Groups with wildcards. As recommended in the previous section, HPE recommends configuring the HPE MSA Storage to use a Host Group for a vSphere cluster and use the cluster object when mapping Virtual Volumes. This will create all the mappings to all the adapters to support multipath in the cluster in one step. VMware vSphere supports an active/active multipath environment to maintain a constant connection between the ESX host and the HPE MSA Storage Array. The latest version of vSphere offers 3 path policies: “Fixed,” “Most Recently Used,” and “Round Robin.” HPE recommends using the “Round Robin” preferred selection path (PSP) policy for best performance and load balancing on the HPE MSA Storage. By default, VMware ESX systems use only one path from the host to a given volume at any time. This is defined by the path selection policy call MRU path. If the path actively being used by the VMware ESX system fails, the server selects another of the available paths. Path failover is the detection of a failed path by the built-in ESX multipathing mechanism which switches to another path by using MPIO software, VMware Native Multipathing (NMP), and the HPE MSA firmware. "

The key in the above paragraph here being that HPE recommends you use RR as the preferred policy, but then the MSA firmware does some magic on the backend when it comes to determining which to actually use.  I can understand the confusion because it does say active/active in the CLI, but it's not true active/active as the industry definition would state.

---------------------------------------------------------------------------------------------------------

Was it helpful? Let us know by completing this short survey here.

[nrinaldi]

I see what you say. Thank you for your time!

Nicholas.