Multipathing Question

[edawg]

Hello-

I have question about multipathing settings on ESX 3.5.0 158874. I have four PowerEdge 2950's connecting in to an EMC Clariion CX3-20 (flair 26, I think) with two fibre switches. Each server has two Qlogic HBA's connecting into separate switches connecting into redundant storage processors. I am trying to use VRangerPro (backup software from Vizioncore) on my VCB backup server and can't get it to run the VM backups. The vendor has told me it is because I am running MRU multipathing and need to change it to fixed multipathing. VCB runs fine on the backup server when I don't use the VRanger software.

I have another ESX 3.5 cluster connecting into an IBM SAN, that is using fixed multipathing, with a VCB/VRanger setup that runs VM backups fine. My question is, Has anyone seen this problem before with VRangerPro where you have to have fixed multipathing to run VM backups? If you can't comment on the VRanger product can anyone comment on using fixed paths with a Clariion CX3-20 in conjuction with the ESX software and how going from MRU to fixed might affect my production environment?

For what it is worth EMC refers to my SAN as having active/active processors, but VMware refers to it as active/passive because it can't access the same LUN at the same time. At this point I am confused about what the affects would be for me moving from MRU to Fixed in my environment. Any thoughts are appreciated?

Regards,

Erik

[kjb007]

Vmware sees the SAN as active/passive, which is witnessed by the path being chosen as MRU. With MRU, if a LUN fails over to a secondary SP, then the server will continue to use the LUN on the path that it has failed over to, and will stay there. If the policy was fixed, then the esx server would attempt to use the original path that was used, which can cause disk thrashing where the SAN will try to failover the LUN back and forth between the SP's. This can lead to poor performance, at best, and LUN / SAN failure at worst. This is why this policy is important. I can't comment on VRanger directly because I've never used it, but as long as you are able to reach the LUN from the VCB proxy in the same manner as the ESX host is accessing it, then you should not have problems. Meaning, the LUN is presented to VCB the same as ESX, you should be good to go.

-KjB

VMware vExpert

[mcowger]

While EMC may CALL your array active/active (in the sense that its ALUA), its not really. A TRULY active/active array can access any LUN through any controller at any time without penalty. Higher boxes can do this (DMX, USP, 3PAR, etc). Your CX3 can't, and as such is correctly identified from a technical perspective by VMware as active/passive, regardless of EMC's silly marketing.

--Matt

VCP, vExpert, Unix Geek

[AndreTheGiant]

True: CX is active/passive on a single LUN. Is globally active/active, but on different LUNs.

Andre

**if you found this or any other answer useful please consider allocating points for helpful or correct answers

[AndreTheGiant]

VCB runs fine on the backup server when I don't use the VRanger software.

Have you try to use VCB on the server with VRanger? It does work? If no, it isn't a problem of VRanger.

Check the two machine to see if there are some difference: have you install PowerPath? Does the disks LUN appear one or twice in Disk Management?

Andre

**if you found this or any other answer useful please consider allocating points for helpful or correct answers

[edawg]

Thanks for the response. When I run VCB on the backup server with VRanger it works fine. This is the backup server that is connected into the EMC SAN using MRU. But, when I run Ranger on this same VCB backup server, the Ranger backup fails. When I run RangerPro on the backup server that is connected to my IBM SAN using fixed paths it works fine. The IBM SAN is not using powerpath, but the backup server that is having problems is using powerpath. I have included an attachment with two screen shots. The first shot is on the backup server that is having issues, showing the drives. The second shot is of the backup server that is going to the IBM SAN that is working. In disk management each lun being presented to the vmware cluster is only showing once as Healthy(unknown partition).

[AndreTheGiant · Accepted solution]

PowerPath is going to use more path at once and VCB loose the connection.

Usually I install PowerPath without the license key, so it works in simple failover.

You can also try to remove PowerPath.

Andre

**if you found this or any other answer useful please consider allocating points for helpful or correct answers

[edawg]

10-4

It sounds like I won't be able to get VRangerPro to work as long as I continue to use Powerpath in its current license mode. Since you have been so helpful, one more question. Is it correct to say that from the VMware side the multipathing (MRU vs. Fixed) is automatically chosen based on the storage processor type SAN I have...For example, true active/active SAN's will be automatically set as fixed by VMware, versus active/passive SANs which will be automatically set to MRU?

[AndreTheGiant]

Is it correct to say that from the VMware side the multipathing (MRU vs. Fixed) is automatically chosen based on the storage processor type SAN I have...

True.

Usually is not necessary change this configuration (in ESX 3.x).

In ESX4 maybe (if you want to use round robin or 3th part software).

Andre

**if you found this or any other answer useful please consider allocating points for helpful or correct answers