EVA5000 & VI3.5

[nmiguelrf76]

Hi,

Is anyone there running ESX 3.5u2 or u3 on the EVA 5000? If yes what is the VCS you are using? Currently testing it on EVA3000 with VCS 4.004 and we do not encounter any issues.

PS - It is not officially supported by HPm but you can find info on the documentation Fibre SAN Guide for 3.5u2 & u3 on how to present vdisks from EVA 3000/5000 active/passive and active/active.

http://www.vmware.com/pdf/vi3_35/esx_3/r35u2/vi3_35_25_u2_san_cfg.pdf (page 64)

Regards,

Nuno

[GLP]

We are currently running ESX3.5u2 on a EVA 5000, VCS 3110. We are encountering some issues that seem to point back to the EVA 5000. VMs will go to a PXE boot, resolve Power off and power back on. Within the Guest OS Windows 2003 disk errors are being reported. Currently working through the issue with VMWare and HP.
Here is the output from esxcfg-mpath from a LUN on the EVA 5000:
Disk vmhba2:1:2 vml.0200020000600508b400103e910000800002240000485356313130 /dev/sdn (1024000MB) has 2 paths and policy of Most Recently Used FC 66:1.1 50060b0000e5492a:50060b0000e5492b<->50001fe15005d27c:50001fe15005d270 vmhba2:1:2 On preferred FC 66:2.1 50060b0000e548ca:50060b0000e548cb<->50001fe15005d27d:50001fe15005d270 vmhba4:1:2 On active
LUNs from the EVA 5000 are only being presented from one SP.
Here is the output from esxcfg-mpath from a LUN on the EVA 8100:
EVA8100
Disk vmhba2:3:1 /dev/sdt (1024000MB) has 8 paths and policy of Fixed
FC 66:1.1 50060b0000e5492a<->50001fe1501084e9 vmhba2:3:1 On active preferred
FC 66:1.1 50060b0000e5492a<->50001fe1501084ed vmhba2:4:1 On
FC 66:1.1 50060b0000e5492a<->50001fe1501084eb vmhba2:5:1 On
FC 66:1.1 50060b0000e5492a<->50001fe1501084ef vmhba2:6:1 On
FC 66:2.1 50060b0000e548ca<->50001fe1501084e8 vmhba4:3:1 On
FC 66:2.1 50060b0000e548ca<->50001fe1501084ec vmhba4:4:1 On
FC 66:2.1 50060b0000e548ca<->50001fe1501084ea vmhba4:5:1 On
FC 66:2.1 50060b0000e548ca<->50001fe1501084ee vmhba4:6:1 On
This LUN sees all 8 paths on both SP that is presenting disks
I do not think we are in a position to update the VCS. We have not configured the mode for custom, I am waiting to hear back from HP and VMWare for any recommendations. We are not in a position to abandon the EVA 5000, we have a considerable amount of Production VMs running.

Any input from anyone else would be greatly appreciated.

[Adam Garsha]

You need 4.x for active/active presentations. I've got a 3.5u2 test environment on EVA5K VCS 4.004 and I've had no problems (but it is a pretty small environment). Production is on EVA8.1K's and I've stayed away from presenting/zoning 5K's to prod clusters since not listed in vmware's HCL (not that I think it wouldn't work, I just don't want my neck sticking out)

[nmiguelrf76]

Hi,

We have have it running with ESX3.5u3 without a glitch on EVA3000 with VCS4.004

I just find strange that on the official documentation of VMware they mention the EVA3K & 5K... as a block storage provider there should not be any interoperability issues.

On the host type we use VMware, otherwise we would use custom as defined on the SAN Guide.

Regards,

Nuno

[Steven Clementi]

My Demo environment is running ESX 3.5 on p-Class blades and an EVA500 running 4.007 I believe, or perhaps 4.004. I did not have a problem presenting the luns to the servers.

I have no problems with the servers, than again I have 2 vm's running at the moment which consist of a backup AD/DNS server and a vmotion demo OS.

Steven

[GLP]

Steven, could you SSH to 1 of your HOST and run esxcfg-mpath -l and include the output from the EVA5000 LUN. We have not updated the VCS to 4 as of yest and I would like to know if 4 will resolve the issue that only LUNs are presented from 1 SP.

The output from a ESX 3.5u2 HOST attached to a EVA5000:

Disk vmhba2:1:8 /dev/sdl (1024000MB) has 2 paths and policy of Most Recently Used
FC 66:1.1 50060b0000e5492a<->50001fe15005d27c vmhba2:1:8 On active preferred
FC 66:2.1 50060b0000e548ca<->50001fe15005d27d vmhba4:1:8 On

Thanks

[Uwe Zessin]

GLP, that looks like a configuration error. Even an active/passive EVA5000 _does_ present a virtual disk through the non-managing controller and you see 4 paths with 'esxcfg-mpath'.

In your case, adapter 'vmhba2' has access to Controller-B and 'vmhba4' has access to a different port on Controller-B, but there is no path to Controller-A.

[GLP]

What version of ESX are you running and can you provide the output for a disk mapping on a ESX 3.5 Hosts

esxcfg-mpath -l

Thanks

[Steven Clementi]

Output attached... I show 4 paths.

The first thing I would check is your zoning to make sure you are presenting controller A to your server(s).


Steven

[GLP]

You Guys Rock
We are not able to update the VCS to 4 at this time. We are making a modification to the SAN and presenting all LUNs from a common controller, we are currently configured to atlernate controllers. There is documentation that suggests that in the configuration we are currently in we could experience thrashing in http://www.vmware.com/pdf/vi3_35/esx_3/r35u2/vi3_35_25_u2_san_cfg.pdf

In the current configuration I am experiencing instability: Ex Today I was adding a RDM to a VM, in my notes when I attached it, it reflected /vmfs/devices/disks/vmhba1:0:6:0. Later when I went to detach it, it was presented on /vmfs/devices/disks/vmhba2:0:6:0, which implies that that a path failover occured or thrashing from what I understand from researching. It is currently /vmfs/devices/disks/vmhba1:0:6:0. I will update the thread with the results of the change.

Thanks

[GLP]

Steven thanks for the output from the mpath. You are running with VCS 4. We can update at this time, but in the near future I understand the upgrade is planned. I feel confident the change this evening will eliminate some issues, but we will continue to be in an active / passive configuration and not sure if ESX 3.5 will see both SP. We are applying the recommended custom connection type next weekend. Didn't want to risk 2 changes on the production SAN presented to VMWare.

Thanks