This message was deleted.
# microsoft-fslogix
s
This message was deleted.
👍 8
l
Defo requires a blog post!
t
Congrats! Would love to hear how you set up your file servers, storage, etc. Especially with all the problems we just had with our deployment.
👍 1
n
FSL is very cool, however I'm shy'ing away from it for now due to lack of support from MS. I had an issue, took them weeks to contact me. Back and forth with India via email. Any product in prod I need to be able to get support right away and they don't have it (like the old company did)
n
What problems did you have, @Tim Riegler? Based on recommendations I've seen both here and elsewhere, we're certainly not in the "ideal" position. Our server team balked at SOFS and Microsoft clustering in general and would not budge one bit. As a result, I ended up with two Server 2016 boxes (4 vCPU, 16GB RAM) front ended by DFS (no replication), one share for each container type, and all hosted on our EMC VMAX flash storage with compression enabled and dedup disabled. Time will tell if that holds up. Our average profile is 250MB right now, but I've got a few users with 2-4GB profiles, and it all looks like OneNote is what's bloating them. They're definitely heavier OneNote users than everyone who tested this in QA. I guess I could always redirect that via the XML, but I don't know if I want to go down that road.
t
All sorts of performance problems. We are mostly sure it had to do with a VMware bug based on our version of ESXi that the file servers are on. Although it may also have had to do with our file server configuration. We've had ODFC in prod for several years without any problems. For new 2016 desktop deployment, we built new 2016 core file servers, nvme disk adapters (because backed by AFA, so why not), and ReFS on the volume (along with VMFS6). All of that was different than previous prod (2012 R2, not core, pvscsi, NTFS on VMFS5). Once our new profile and ODFC (separate VMs) hit load, they stopped accepting network IO to the storage. Also had all sorts of performance problems with it (laggy Citrix sessions). Ultimately we patched ESXi, modified the file servers to use pvscsi and NTFS, and things are much better. Not sure on our current profile sizes, have to do some digging there...
n
Ouch. I have no view into our ESXi environment because of how compartmentalized our teams are. That can be both a good and bad thing, at times.
After poking into some of the larger containers, it's pretty clear that I'm going to be shrinking these down on a normal basis. I'd love for Microsoft to figure out a way to do that automatically.
t
I know that @David Ott has a compact tool (here: http://www.citrixirc.com/?p=1052) that you can script an run automatically.
n
Yeah, I'm already running a script on my own every night. I'd just love to see Microsoft built that into their own utility so that I don't have to rely on something like this, where locked files would be common, etc.
n
They have no incentive to do that as it uses more space on Azure for which they can charge more...
n
It's Friday, don't kill my hope!
n
@Tim Riegler i never knew you could add nvme to a vm, i'm actually spinning up a vm with that at the moment to check it out, i also have a AFA, does it really make a difference in performance?
t
Supposedly. Couldn't really say, as we broke our environment when implementing, and that was one of the changes we reverted, just in case. I don't actually think that was the problem, and I didn't want to take any chances.
@Nick Panaccio the @David Ott script checks for locked files and handles them well. If locked files are that common, you could literally just schedule it every 8 hours or so, regardless of who's logged in.
n
so you are NOT running an nvme adapter within the vm, correct?
t
@Nick Casagrande the idea with the NVMe is that it is much less SCSI commands from the VM to the host, so should theoretically improve performance. I'm not entirely sure that's the case when pointing to an AFA though, as at some point there are still SCSI commands to the array.
Not any more. We rolled out with it in the VM, and pulled it after our world blew up.
n
gotcha
t
Just doing the basic VMware pvSCSI
n
im still not sure if pvscsi is better than lsi sas which is still the default from vmw when creating vms
t
It's the default because you can install a Windows OS directly to it, which you cannot with the pvSCSI. Everything I've read (which is A LOT) says to use the pvSCSI adapter for performance. The caveat is that the OS has to be installed (or you install the pvSCSI device during install, which is a pain), and then you have to add the second adapter, blah blah blah. In short, it's a pain during initial install. Once you have a template, no problem. Sources: https://www.amazon.com/VMware-vSphere-Host-Resources-Deep-ebook/dp/B0746H6DFC/ref=sr_1_2?keywords=vsphere+host&qid=1571406284&sr=8-2 and https://www.amazon.com/Virtualizing-SQL-Server-VMware-Technology/dp/0321927753/ref=sr_1_1?keywords=vsphere+sql&qid=1571406297&sr=8-1
n
@Tim Riegler I know I can do these things (and actually already do it), but would still want Microsoft to add the feature themselves. If they don't, I'll keep doing what I'm doing. It's definitely not a hill I'm willing to die on.
t
Completely agree there! We did see quite a performance improvement for our SQL and File Server workloads moving to pvSCSI, that's for sure.