It's true that before implementing a software solution we must take a look in Hardware Compatibility Guides, but if we already have the hardware components and they aren't in HCL, the term engineering comes to play. :)
My company acquired Vsphere 4.1 and we had two EVAs: EVA 8000 and EVA 5000. With the first of them there were no problems. It's fully supported (Active/Active array with ALUA compliance) . The SAPT recognized it as VMW_SAPT_ALUA. But, always there is a but, EVA 5000 VCS 3.11 is not supported. It is recognized it as VMW_SAPT_DEFAULT_AA (Active/Active). As you see at the fig.1.
fig.1
The problem came when I did a virtual disk failover between controllers from Command View. All virtual machines inside that datastore hung up. The datastore was missing.
I began to look for a solution, but nothing. In some blogs the recommendation was to upgrade to VCS 4.x, which support Active/Active mode, but, another but, to do that, I should upgrade Command View from version 6 to 9 because version 6 doesn't support VCS 4.x. The problem is the license for the upgrade: license=$$. In others blogs the recommendation was to add rules for the desired SATP but they weren't specific to EVA 5000.
After a period of trial-error adding rules, the output of esxcli --server x.x.x.x corestorage device list gave me the information needed:
Vendor: COMPAQ
Model: HSV110 (C)COMPAQ
With this information, the addrule command is:
esxcli nmp satp addrule --vendor="COMPAQ" --model="HSV110 (C)COMPAQ" --claim-option="tpgs_off" --satp="VMW_SATP_DEFAULT_AP" --description="Support Mod for EVA5000"
tpgs_off means that the array is not ALUA compliant.
Reboot the host and as you see in fig.2 you will have the two host port of the non managing controller marked as "Stand By".
fig.2
Now, it is time to test the configuration. In Command View change the owner controller of the disk to the other. If all was successful, you must see the failover condition:
fig.3
Maybe, this configuration isn't supported, but, it works...
No comments:
Post a Comment