Running Virtual XMS allows us to benefit in some ways from the resiliency of VMware to keep the XMS alive.
But is there anything or anyone tried clustering the XMS?
Well unlike VMAX where the SP runs on the VMAX's power supplies the XMS sits external and so open to potential power issues outside of the XtremIO configuration itself.
Even if you lose the XMS ( from failure, power, restart whatever) it's not in the datapath so as long as the XIO remains online, data services will still be online and available to hosts as normal until the XMS recovers or is restored.
Yup. We have DRS under ESX but looking for equivalent to add resiliency for physical XMS.
Have seen an issue where power nocked out XMS and needed to get to Brick urgently, and couldn't ascertain state correctly. Customer just looking for as much additional protection as they can get....
We currently have the physical 1U server configured as our XMS server but are interested in running a virtualized XMS for a few reasons.
1). ESX HA. If the ESX host fails then the XMS VM moves to another host.
2). Multiple networks. It is my understanding that there is only one physical network connection from the 1U XMS physical host that was supplied with the cluster, but our ESX hosts have multiple redundant NICs.
3). We use Appsync for snapshotting EPIC and Clarity volumes. I believe the loss/crash of the XMS will stop Appsync from running scheduled jobs/commands against the array.
We have not yet migrated to the Virtual XMS (still in planning) and I am still looking for documentation to show how to transfer/recover the XMS on the new VM once it's functional. Does anyone know where that can be found?
You can find that in the Storage Array User Guide - there is a small section on XMS Migration. This will do the migration of the cluster, however anything XMS-related will be lost and have to be re-created from scratch. This include performance data, tags, reports, smtp, alerts, LDAP, etc. So this can be cumbersome depending on how you have things setup right now. From doing this at a couple sites and inquiring with EMC, there is no alternative that I have been told of to bring this information over with you from the source XMS. So the process is smooth to remove and add the cluster to the new vXMS, but the time will be spent in recreating everything. I would recommend you take a debug log file for backup so that you can apply the same tags, etc on the new vXMS quicker.
Also, if you have RP involved, this will impact that as well.
Wow. "Small section" on migration is an understatement. We're also have to change physical IP networks for all of our XIO controllers and the XMS so I'm sure this will get exciting. I have already pinged my SE for assistance. Thanks.
yeah - its a process that wants to be planned for sure.
For the Storage Controllers, there is a process documented on how to do this in the Install guide that can be followed by EMC or your partner.