UNSOLVED
SS
Spencer Simpson
13 Posts
0
1019
July 18th, 2017 09:00
Feature request - separate (manual) NAS subsystem startup/shutdown
I'd like to break an unnecessary dependency loop (without buying new hardware).
I have a Unity 300 that has both iSCSI-based VMWare datastores and a CIFS NAS server. The NAS server is joined to an Active Directory domain. The domain controllers are virtual machines that use the VMWare datastores. The NAS server is on the same (default) VLan as the management interface. iSCSI runs on a different VLAN.
This may not be the most optimal configuration but that's the way it is. It works great most of the time, except on the rare occasions I have to shut down the whole system and later bring it back up (in this case, the Unity had to be moved within the rack)..
Because the domain controllers/DNS servers (being VMs) have to be shut down first, the NAS server immediately starts emitting events that it can't reach them. This continues until the storage system is shut down.
When starting everything back up, I can't start the domain controllers until I start the virtual hosts, and I can't start those until the VMWare datastores are available. So first, the Unity has to start. The iSCSI LUNs did not become available to the virtual environment for an hour and a half! (that's something I already have a support case open for).
All the while the NAS server (which started right up) was emitting the same events about not reaching the domain controllers.
It would help a lot if I could manually shut down just the NAS server before bringing the rest of the storage system down, and later start it up manually after the virtual environment is back up.
There is no user commands available to shutdown the NAS Server or Luns. But from what you described, this seems like annoyance than an issue during physical movement of unity array or any maintenance which could be a planned activity.
I think this setup do not have proper DR configured as maintenance of unity going to bring down every thing (not just DNS or DC). You should get one more Unity and configure as a DR that may help to keep alive the ESX and VMs in addition to the DR scenario.
AndreD Dell
Moderator
•
239 Posts
693
1
Posted July 18th, 2017 23:00
Not sure I understand this correctly.
The DC is on a block or file datastore?
Did you figure out why the datastore/LUN takes an hour and half to become available? That doesn't sound at all normal.
So, given your example, all you are trying to do is avoid the NAS server errors until the DC comes online? I can't see any other benefits here..
So, as you know, this is not exactly best practices, but I understand you have limited resources. It's hard to give advice in this case, because that too would probably escape best practices, example:
Did you consider putting the DC on local storage on the ESXi host side (with some VMware redundancy otions, like running the DC on two hosts for HA), maybe? This way it would not be dependent on the same array where the NAS servers are.
Andre @ Dell EMC