StoreVirtual Storage
Showing results for 
Search instead for 
Did you mean: 

Graceful shutdown of LeftHand P4500 G2 Stretched Cluster (10 nodes)

Occasional Visitor

Graceful shutdown of LeftHand P4500 G2 Stretched Cluster (10 nodes)

Hello Community

Just wanted to run something by your guys & girls.  I have a 10 node stretched cluster between two datacentres.  This management group has the correct number of managers & FOM.  I wanted to shutdown the entire management group and the 3 clusters inside it so I can perform a fibre upgrade on the backend switches.  Replacing this fibre will break the connection between the two datacentres so best practice will be to shutdown.

The enviroment only hosts volumes with virtual server data contained in them.  I was speaking to a HP support engineer who said I should do the following:

1. Shutdown all the VM guests.

2. shutdown or places hosts in maintenance mode.

3. In CMC unassign servers from each volume from each cluster in management group.

4. Shutdown the SAN.

5. Shutdown/Perform maintenance on the network LAN/SAN switches.

6. Power on network switches.

7. Power on SAN.

8. Turn maintenance mode off for VM hosts or power on hosts.

9. Turn on VM guests.

My only questions to the above is for point 3, aslong as all guests ahve been shutdown that should break the iSCSI connections so do I need to remove access to the volumes in CMC? Point 4 is there an order required for shutdown and startup or as we have shutdown all guests no writes will be made to the Volumes so we can shutdown and startup in any order?


Thanks, DPGentry

Occasional Advisor

Re: Graceful shutdown of LeftHand P4500 G2 Stretched Cluster (10 nodes)


if you shutdown your hypervisor hosts I think you can skip step 3. Maintenance Mode is not a good idea as powering down the SAN then leads to an APD situation. To shutdown the Management Group gracefully, you can right click the management group and should down all the nodes including FOM at the same time so that no resync is needed after power on. All clusters are shut down then.