"Optimize" a CSV volume

Question

I have a H-V 2012 cluster with 3 nodes.  I use a CSV volume to store the VMs.  I have all the latest patches installed. I have an Equallogic SAN (PS4000) with the latest firmware on it providing the LUN for the CSV.  Everything in my environment is
supposed to support re-thinning (or unmapping, or whatever the right term is) the LUN.  I have about 500 GB of unused space on the 2TB volume, and the volume was thin provisioned.  A restore of some very large vhd files from backup caused the
thin provisioned volume to grow and to use almost the entire volume at one point but the corrupt VHDs have since been deleted.  Now I have "dirty" blocks in the LUN that I want to reclaim into free space on the SAN. This all happens, apparently,
when Server 2012 performs an "Optimize" on the disks.  In my environment this is scheduled to happen once a week.  It did apparently do something this last week, because my volume utilization on the SAN went from 96% to 91%.  Not even
close to reclaiming all dirty blocks, but it's a start I guess.  So now I went in to the "Defragment and Optimize Drives" utility and told it to commence a manual optimization.  Nothing happens and event viewer give me this error:
The volume VMStorage1 (C:\ClusterStorage\Volume1) was not optimized because an error was encountered: CSVFS failed operation as volume is not in redirected mode. (0x8007174F)
So my questions are these:
Shouldn't it put the CSV in redirect mode if it needs to do this in order to optimize the drive automatically?
If it can't do this automatically, how did it return 5% of the CSV SAN volume to free space last week?
Can I put the volume in redirect mode manually and do the optimize manually? Redirect mode is not supposed to be necessary in 2012 CSV any more- at least not for backup.  Why here?
Will my environment re-thin, Unmap, whatever?  It appears it MIGHT.  Does it take several iterations (ie weeks)?
Can anyone explain this incredibly vague and cloaked process from a Windows server 2012 perspective?
Thank you for any help!
DML
DLovitt

Answer 1

Hi,
In CSVv2.0, every effort was made to expand the number of scenarios that would use Direct I/O over Redirected I/O. 
Direct I/O delivers faster performance with lower network overhead. Emphasis is on using Direct I/O for all types of file open actions. 
Direct I/O uses buffered reads and writes which means it can take advantage of the Windows Cache Manager. As an example, Direct I/O results in better virtual machine creation times and improved copy performance. In CSVv1.0, to get the highest performance
during a copy operation, the destination node had to be the Coordinator node for the destination CSV volume. 
CSVv2.0 uses a new algorithm for determine what types of I/O are redirected. 
Oplocks are used as a distributed locking mechanism to determine if I/O can go via a direct path.
All of the optimizations and performance improvements are for naught if the file system cannot remain available for the applications to use. 
The new file system check and repair capability goes a long way towards ensuring file system availability. 
The new file system health-checking model coupled with new functionality in chkdsk helps in this area. 
In addition, CSV volumes can take advantage of these new capabilities. 
Thanks.
Kevin Ni
Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread.