UEM Configuration Share (Storage Ideas)

[JohnTwilley]

We are currently leveraging a multi-node Isilon cluster to provide the UEM Config share, and UEM profile shares.

We've had a few outages, which disrupts the UEM shares in VDI.  

I'm looking for other ideas for a simple redundant CIFs share option for our VDI environment.  I was wondering if you wouldn't mind sharing your configurations.

I was thinking of setting up Windows 2016 servers with DFS...but the little voice in my head is screaming NO at that idea...

I'd like to have it all self contained within vSphere environment if possible. for performance and control.

What are your great SMB solutions for UEM?

[VDINinja311]

JohnTwilley​

We have DFS-R setup between two Windows Server 2012 R2 file servers one at each Datacenter. This is hosting our VDI users' homeshares as well as UEMconfig and UEMprofile shares. We have a 10Gbps pipe between the two DC's so we are good there. We originally had the default setup of basically a round robin approach so it would hit either of the two servers but we started having issues as DFS-R wasn't able to keep up with the replication and was causing some data inconsistency issues between the two. It kept seeing multiple changes to files on the two different servers.

The few things we have learned since implementing:

1. Override the referral ordering so that one file server would be set as "first among all targets" that way all the desktops would read/write to that one and DFS-R wasn't getting mulitple changes to files on both servers.

2. Make sure to set the staging quota high. We started off small and have had to grow the staging quotas to now 192GB. We kept seeing where it would hit the highwater mark a lot and purge files.

3. We setup the Log Insight agent to capture certain DFS-R event logs so we know when replication is broken. This is just for peace of mind for us two admins.

4. We setup automatic replication reports via a powershell script and scheduled task on one of the file servers that will do a replication test and email you the report. This is also for peace of mind for us two admins when we were having issues, before we setup the referral ordering.

Hope this helps a bit if you go that route.

Otherwise no real issues since we setup the quota's to the correct size and the referral ordering.

-Jeremy

[jmatz135]

We have a multinode isilon as well.  While we haven't had outages there are a few things that seem to cause issues.  For instance if you have Windows 10 desktops in vdi you will probably need to turn off leasing, which is kind of the newer version of opportunistic locking, but isilon nodes don't know about leases on other nodes so this obviously will cause issues when the desktop hits a new node and tries to access a file that is leased on another node as it basically just borks.  You could also just just turn off opportunistic locking entirely on the shares for UEM but I would heavily test that first.

[JohnTwilley]

So here I am one year later...and just pulled a 40 hour shift due to login delays from UEM.  Our Isilon storage is just too saturated to provide timely response during logins for our VDI VMs. (shared environment)  Now I'll be a little more serious about choosing a SMB replacement.

Please let me know what your company is using to share SMB for UEM...and whether or not it's working for you.

John

[VDINinja311]

JohnTwilley​

Our deployment is still pretty small, ~275 daily linked clone sessions but for storage we run Nutanix on top of our Dell XC nodes. We have a 3 node infrastructure at each Data center that hosts all the server infrastructure for Horizon Connection Servers, App Volumes, Domain Controllers, GRID licensing, File servers, SQL Servers, and Print Servers, etc, etc. These Dell XC nodes are Hybrid SSD/HDD. ESXi is the Hypervisor but Nutanix runs the storage for our HCI. So far no issues scaling. Our biggest login delay is Group Policy that we are still fighting with our Info Sec to get trimmed down or migrate some settings into UEM to help with login times.

We still run DFS-R in the same configuration like I mentioned last year but we possibly might migrate over to Nutanix Files to host all UEMConfig / UEMProfiles and our Users Homeshares. Nutanix Files just runs SVM's (Storage Virtual Machines) in ESXi that host SMB/CIFS and it uses regular DFS on the front end to present to the desktops. That way we could get rid of DFS-R and it spreads the storage across as many nodes as we want instead of being pinned to the Windows File Servers (1 at each Data center for us).

Desktop side is also Nutanix running on top of Dell XC nodes.

Must be a big deployment?

[JohnTwilley]

Our deployment is around 2,000 concurrent users at the moment.  We are running the VMware Horizon environment on EMC XtremeIO SSD storage, which has been fantastic.

It's the SMB shares that has been a pain point, since it's not dedicated.   When we have issues with slowness, it probably due to something else in the environment hammering the nodes.

Have you had any replication issues with Nutanix's implementation of DFS across the SVM's?  That's the part of DFS that I'm worried about.

[VDINinja311]

We haven't actually POC'd Nutanix Files yet. We already own it with our current Nutanix Ultimate licensing and have been thinking about implementing it. Sorry don't have any real experience with it yet. So far DFS-R and Windows File Servers has worked out well for us. Most likely because of the great response/speed of shared storage of Nutanix across the server nodes.