Hyper-V 2016

[Joker8784]

When using Drivepool I get oplocks on the new Hyper-v read-only files. 

 

 VMCX files are binary and cannot be read

http://windowsitpro.com/hyper-v/reading-binary-vmcx-file-windows-2016-hyper-v

 

Are there any known workarounds to this ? 

 

Thanks in advance!

[Christopher (Drashna)]

Stop the VM in question, duplicate the data, and then let it run. 

 

Though, was  the pool/folder duplicated when you created the specific VM in question? 

[Joker8784]

Even when the servers off, those particular file do not duplicate because Windows cannot read them. It appears that hyper V creates binary files so they cannot be altered or copied to prevent corruption of vm's

[Christopher (Drashna)]

Windows can still read the files.  The change in the files was from XML to a binary format, so that it's not HUMAN readable.  You can still open the file in a hex editor and view the files just fine.  But they are gibberish when trying to figure out what it means.  

 

That's what the article is saying.  And that microsoft did this to prevent people from manually editing the files, because they could cause serious issues and corrupt the files. 

 

 

 

The "oplock request is denied" means that the file cannot be locked by our service, which generally means that the files are still in use. meaning that the Hyper-V service is likely accessing the files still. 

 

It may be a good idea to run "resmon" on the system, open the "disk" tab and see what is actually accessing the files. 

 

Otherwise, enable tracing, reproduce the issue, and then upload the logs:

http://wiki.covecube.com/StableBit_DrivePool_2.x_Log_Collection

 

But this will likely tell us the same thing (that the files are still open). 

 

You can run "net stop vmms" to stop HyperV, and then restart it by running "net start vmms" after the files have been duplicated. 

[Joker8784]

I'm going to try to rebalance tonight to see if that works because even with it off it is not de-duplicating.

 

SIDE NOTE:
VM's do not seem to be caching through SSD's

When I run IO Meter on the physical server i'm getting 800+ score on read performance but when I run within a VM on the same drive i'm getting 140 ish which is +/- 10% of what I was getting before I turned on the SSD Caching.

 

For a file server i'd imagine this thing would fly; for hosting VM's do you have any tweaks or recommendations ? 

[Christopher (Drashna)]

Just a heads up, if you're using the SSD Optimizer here, you'll want to set up a file placement rule to ensure that the files are kept on the SSD(s), otherwise, they will get moved off (sooner or later, and that may actually be part of the cause of the issue above). 

 

And you'll want to uncheck the "unless the drive is being emptied" option in the main balancer page. 

http://stablebit.com/Support/DrivePool/2.X/Manual?Section=Balancing%20Settings