This message was deleted.
# _general
s
This message was deleted.
🧵 3
j
Well we lost that fight a long time ago. During their tests they were not satisfy with the dr options available with Nutanix Files.
d
Well I can’t relate ;) But why using ABS and not just attaching local vdisk to the file server? However I don’t think the issues are Nutanix related.
o
Neither, i'm pretty sure it's related to the default SMB settings I posted above, we're testing out some more optimized settings on a DEV filer and will report back
c
What performance issues are you seeing? We went a similar path pre 3.6 release because of some issues we were having. With Windows Failover cluster and a General file Server role, with storage presented by ABS over iSCSI. How many vdisks are you using from ABS? Initially we found we hit the limit of a single vdisk around 4500 IOPS and then moved to using many vdisks and used a storage pool to aggregate the storage and get a larger amount of iops. We've actually recently migrated away from this solution because we found we hit the limit of the virtual nic's driver with the lack of support for RSS on AHV currently and we were seeing cpu core 0 at 100% which caused significant file server availability and performance issues. We recently moved the workload back to Nutanix Files on 3.6.1 and honestly all of our problems went away they simply are just not struggling and perform nothing other than consistent no matter the load.
❤️ 1
o
Thank you for your response, Chris. I will provide disk numbers tomorrow. I will add something , which I didn’t mention in my original post. We are only seeing the poor file copy performance on the SMB share that’s got continuous availability enabled.
c
That is the one thing we did not do, since it's not supported with Citrix UPM we did not enable continuous availability. Are you using separate virtual NIC's for SMB vs iSCSI traffic?
o
We don't have separate vNICs on the filers , no I did some reading over the weekend from Microsoft articles, and I believe enabling CA on the file share would only be useful if we had chosen SOFS for the file server type, we did not. We've got a case opened with MS, and will confirm the impact of disabling it with active users. I did a failover from the pri > secondary server last night, and noted a 45 second unavailability of the clustered share, I'm assuming it will be even longer with CA disabled on the SMB share
We escalated to Microsoft via a case, and the culprit was was having "CONTINUOUS AVAILABILITY" enabled on a-non SOFS type file server. They recommend only turning CA on when you've got specific work-loads on a SOFS type file server: SQL, Hyper-V VHDXs. As our filer is used for FSLogix as well as folder redirection , the setting is working against us. With it on, we would get 40-60 MB / sec on file copies, with it turned off as of this AM, we're back to 600 MB/sec
👍 1