ESX3.5 Cluster & MD3000i -- Both servers see iSCSI Targets, Only one server can use partition

Solution 1:

I feel your pain....I've done battle with ESX and iSCSI multiple times over the past year.

I'm not certain, but you might be running into an issue due to the size of the resultant datastore. There's a 2TB limit to an iSCSI LUN, which is fine since you have split it into two 1.6 TB LUNs.

I wonder if epi3 can't load the datastore because it believes it to be an invalid size.

Have you tried loading each lun as it's own datastore to see if the hosts can see them correctly that way?

Solution 2:

Seems like it's allowed iSCSI access but no read/write... Has this been done?

Select “Yes: This host will share access to the same virtual disks with other hosts”

(from http://www.dell.com/downloads/global/solutions/pvault_esx_storage_deployment_guide_v1.pdf)

EDIT: To eliminate ESX as being the issue, can you put the second ESX in a separate host group, and assign that hostgroup a lun? Also, I saw some old posts where if the initiator name was longer than 31 characters, the ESX box would not connect. From what I see on your screenshots, and assuming they fixed that, you should be okay. Just thought it was worth mentioning here.

Solution 3:

Its not a great answer, but we solved the problem.

It seems our "EPI2" server was freaking out in some manner that refused to share its storage.

Once i removed EPI2 from the cluster and rescanned using EPI1 (ESX4.1) and EPI3 (ESX3.5) both found and mounted the storage properly.

Since EPI2 had caused these problems we decided to migrate all virtuals off it and upgrade it to 4.1.

Since upgrading, we've had no problems, all 3 ESX boxes see the storage and share it properly.

Thanks everyone for your help.