While working on an upgrade to vSphere 5.0U1 on a Cisco UCS environment, where the ESX hosts boot from SAN, I noticed one of the hosts was not registered correctly on the EMC VNX, as it showed up as unmanaged. Because the ESX hosts boot from SAN, the host has to be registered before it can auto register, and when it was registered manually the host was not able to update the registration.
I tried several things, like power off hosts, deregister initiators, power on host, reregister boot initiators and add registered host to storage group again, and reboot host, with and without restarting the VNX management servers, before I found a solution that did work, which is kind of dirty if you ask me.
The solution is to put the ESX host in maintenance mode, enter engineering mode on the VNX, and deregister all logged in initiators for that host. Next, do a rescan on the host (which will not work because the deregistration removed the host from the Storage Group) to let the host register at the VNX, and add the newly registerred host to the correct Storage Group on the VNX again. Do an other rescan on the ESX host now to see your Datastores again.
IMPORTANT: Reboot your host after you did this, since a rescan of the Datastores does not Rescan your boot disk if you use Boot from SAN
Please remember this is could be a tricky operation since even though the host is in maintenance mode, you still yank the mounted LUNs from the host while it is running, which could cause an APD condition for instance, and the ESX host can probably only survive without its boot LUN for a limited amount of time. Unmounting the Datastores on the host in ESX 5 before deregistering could be a cleaner alternative, but have not tested that.
Duco, what version of VNX OE are you using? 31? 32 (Inyo)?
The VNX where I ran in to this issue is still running a 31 version.
I wonder if it is VNX OE version specific (even down to the patch level). I’ve seen some stuff happen on old FLARE 29 CX4’s that don’t happen under VNX OE 31. Probably due to EMC fixing issues along the way (and creating new ones!)
That could be very true. I am pretty sure I have seen manually registered host being updated by an ESX host in the past. One of the other things I noticed in this case was that the host properties on the VNX still showed ESX 4.1 and that was only updated after I deregistered with the procedure I mentioned in my post.
Nice Post Duco.
The issue is still there with boot SAN on FLARE 32 VNX.
Your suggestion did the trick. Now I am going to open a ticket with EMC to get more information about that.
Reboot one SP, or wait a few hours.
Initiators will register correctly.
De-register means after entering into the engineering mode it show the initators with checkbox. so need to uncheck the initators ?
I recently encountered what I thought was this issue only to discover the hosts has not been named or pointed to a domain controller / DNS. Once that was done, the initiators auto-registered beautifully. I implemented a boot from SAN configuration with Cisco UCS. The single “temp” initiator per host I manually registered for the UCS blade, were automatically updated by the array once the hosts were installed with ESXi and assigned hostnames.
Thx a lot dude 🙂
We encountered the same problem. After configuring a proper Hostname and Domain in “TCP/IP Configuration” -> “DNS Configuration” the auto-registration worked just fine.
Is there anything inherently wrong with just leaving hosts manually registered?