Hi all, I want to ask if any of you have ever encountered this process and if the steps my client followed were correct or was there another way to make it simpler.
Objective:
To change their Fiber SAN, a VMAX to a POWERMAX, so what they did was the following:
1. Removed all VM's from inventory.
2. Removed the zones from the fiber switches for the VMAX, and when the DataStores became inaccessible, they removed them.
1. Cloned/replicated the LUNs at the box/SAN level from the VMAX to the PMAX.
2. Created the new zones.
At that point, on the ESXI's they did a Re-scan of the storage and the ESXi's could see the LUNs, but not the DataStores.
They tried to register the Datastore from the vCenter, and the vCenter wizard asked them if they wanted to keep the signature or create a new signature, they indicated to keep the signature, at the end of the wizard, the DataStore could be seen only in the ESXi in which the registration had been done, but not by any other ESXi of the cluster.
When trying to register it on another host from the same wizard, it indicated that it already existed. Despite storage rescanning, the other hosts in the cluster could not see the DS.
That's where I come in and ask them to run an:
esxcfg-volume -l
And then an
esxcfg-volume -r <volume label>
This for each LUN found
After that a re-scan from vCenter and from there all the DS's could be seen from all the ESXi's, the next thing was to rename each DS to the original name and register the VM's.
The most complicated thing, and this is where I want to see if something in the process was incorrect, was that they have several VM's with many disks in different DS's, so they had to:
1. register the VM's
2. Edit them, to remove the original disks, which of course they did not recognize because the ID of the DS's had changed.
3. Re-add each existing disk, taking care of the order and the original controller.
In the end it turned out fine, but I want to know if it was the best process, or could several steps have been avoided if ESXi could have recognized the LUN's as original?
Thanks!
I probably would resignature the volume normally, but other than that my process would be very similar.
Thanks Duncan, so, if i'm understanding well, the process of register the vm's disks manually had to be done either way?
Personally, I feel that a resignature is the safest method. If you don't resignature and do a "force mount" you have to do that same procedure on every host. As you already unregistered all VMs, it doesn't make a difference, as you will need to re-register them anyway at this point. Personally I feel that a signature is the safest option combined with re-register.
PS: A faster method of doing this of course would be to use something like Site Recovery Manager to manage the process of re-registration etc for you. You could set it up temporarily, do the migration, and remove it again even.
How much data in total or why was a simple svMotion not the solution?
When having a PowerMax we are not speaking about 10TB for sure.
More than 50 TB, Dell Consultant suggested doing it by cloning LUNs from SAN to SAN, I entered in the project just when they were not able to "mount" back the cloned LUNs
50TB you can migrate with a svMotion over a long weekend.