Vmware ESX with EVA8000

[IvanForceville]

We have 6x DL380G5 servers running Vmware ESX3 connected to an EVA8000.

On the EVA we created 6 LUNs of 400GB and presented these to the 6 ESX Servers. No problems so far. All 6 ESX servers see the 6 LUNs.

We copied 2 additional LUNs (raw devices) from other EVA's (600GB and 1.9TB)to the EVA8000 and presented these 2 LUNs again to the 6 ESX servers

For some bizar reason only one of the ESX servers sees the two new LUNs. On the other ESX servers I can do a "rescan" but none of them sees the new LUNs.

Any ideas? We unpresented the LUNs and represented the LUNs several times but the issue remains.

[Uwe Zessin · Accepted solution]

A virtal disk must be presented at the same LUN address (VMware calls it the "LUN ID") to all VMware ESX hosts. Otherwise the VMFS will not be mounted.

You could check in the service console with "esxcfg-mpath -l" if the scan has at least configured the SCSI LUNs.

Or you could hunt the /proc nodes:
# grep "Id:" /proc/vmware/scsi/vmhba1/*

The first 128-bits show the LUN WWN which you can look up on virtual disk's properties page.

If the VMFS has been replicated via CA-EVA, it might be necessary to resignature the file system.

But please DO NOT use the "LVM.DisallowSnapshotLUN" - it is only used for storage arrays that cannot consistently map their storage to the same LUN address for multiple hosts (not a problem in the EVA).

[IvanForceville]

OK, It's getting stranger every minute.

First of all we are sure that the origin of the LUN (previously existed or new) has nothing to do with the problem.

When we present a new LUN to the ESX and do a CLI rescan on the ESX host or a GUI rescan in the VI the problem remains.

The 6 LUNs which were previously presented show up after a reboot on both vmhba1 en vmhba2.

Still, we presented a test lun yesterday to one of the esx servers and even after a reboot the LUN doesn't show up in de storage adapters window.

So, in my opinion it has something to do with the Qlogic HBA's which need a restart before the new LUNs occur.

There are many topics on the vmware communicuty forums describing the same problem (or related) and most of the problem could be worked around by rescans.

Unfortunately this is not the case with us and it's unacceptable to reboot all the esx servers every time a new lun is presented.

[IvanForceville]

This issue is resolved.

We rebooted all the ESX servers and now we can dynamically add/remove LUNs from our ESX servers.

Apparently the Qlogic HBA's don't like that you swap cables arround without rebooting the esx.
Our network engineer swapped some cables around, in test fase, and since then the Qlogic HBA's were acting very strange.

Anyhow, lesson learned.

[IvanForceville]

Close