p4300 san space issue

[mattit4]

Hi

I have a site with 3 x p4300 SAN 3.6TB storage devices which we have allocated storage to various hyper-v 2008 r2 servers directly (ie via iscsi within the instance and not to vhds).   

I can cleared down one servers disk usage to 100gb recently however the SAN usage still shows the disk as 1.17TB (600GB configured as Thin in Raid 10)

We are running v10.0 build 1895 (I think 10.5 is available)

Is there a way to relase the unused space without reformatting the disk? 

P.S. this thread has been moved from Storage Area Networks (SAN) (Enterprise) to HP StoreVirtual / HP LeftHand Storage - Hp Forum Moderator

[Emilo]

I can cleared down one servers disk usage to 100gb recently however the SAN usage still shows the disk as 1.17TB (600GB configured as Thin in Raid 10)

How long ago did you clear up the space?

It should start to release some of the space back.

Just remember though that the LH product works at the block lever and not at the file level. So even though a block maybe free on the files sytem side (windows os) the LH block has no way of knowing that the block contains file system deleted data it just knows it was used. You should give it some time and should release some that space back. If that doesnt work then your only option is to delete the volume (migrate the data) and the re-create that will re-allocate the space back.

[mattit4]

I cleared the space around a week ago, I also defraged and shrinked the partition in windows (to 250gb)

I tried creating a new partition on the remaining space and formatted then removed the partition again.

Been getting warnings of 95% usage on the san.  I'll try to migrate the data over the weekend and delete the old one.  I thought this would have been a a fuction of the san.

[Emilo]

Doing the defrag was a bad idea.

It 'touched' every block, so that made the san think it was used block. Remember its block level.

Look at this document it should give you some details.

Defragmentation

See pages 152

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c03035314/c03035314.pdf

[oikjn]

there is no space reclaimation function currently available for VirtualStore's thin provisioned volumes.  If you create a thin LUN and it expands, it will NEVER contract.  You could format the partition in windows and then disconnect it from the server and it still won't go down.  Your only current solution would be to migrate your data to another thin LUN now that you have expanded that other one.

This is a major issue I see with the HP solution, but it is managable for me since I'm mostly using my luns as CSV stores which I can migrate my VMs from an old expanded LUN to a new compacted one when it baloons too much or I assume the LUN will expand to 100% capacity and plan for that.

The only "space reclaimation" on thin provisioned LUNs you will see is when you delete a snapshot.

As a general rule, when I'm using thin LUNs, I make sure to make a 50GB full LUN called "emergency reserve space" so I can delete quickly IF I ever run into a situation where a runaway LUN expands to take up all my free space on the san.

Sorry for the bad news, but the only thing you can do now is migrate the data off, delete the LUN and then recreate one.  If you really don't want to do that, you can migrate the data off, disconnect the LUN from the server and then shrink the LUN volumle size in CMC and then expand it once the space consumtion shows it as low as you want and then reconnect it back to the server.

[mattit4]

oh well - at least that will keep me busy at the weekend.

We will eventually delete the drive soon so I may as well just migrate it to a virtual disk on the direct storage of one of the servers.

thanks for the assistance.

[oikjn]

just keep in mind that the data will continue to grow depending on how that drive is accessed.  Some of my LUNs expand like crazy even when the actual growth rate on the drive shows nothing while others don't grow at all... it all depends on how the server accresses the drive and some programs are better optimized than others.

[mattit4]

I managed to recover 600gb from one LUN over the weekend, initially it was configured for 600GB which at raid 10 ment it occupied 1.2TB of the SAN.

Method:

1. Backup Disk - I used the Sysinternals utility disk2vhd to convert the disk to a vhd onto the host server.

2. Reduce/Resize the Windows partition a smaller size - ie 300GB

3. Take the disk offline in Windows disk manager and disconnect in iSCSI Initiator

4. Resize the LUN in the HP Lefthand CMC - accept the warnings and process takes around 5 minutes depending upon the size.

5. Refresh the iSCSI initiator and reconnect the disk.

6. Check the drive is available and test the data.

This went ok for me and now I have 87% usage instead of 98% so I can snapshot again, I also tested the backup vhd before resizing the LUN. 

Hope this helps others.

[oikjn]

just curious, if these disks are all for hyper-v servers, why not migrate the VMs stored on the drive into another drive, delete the expanded one and recreate it and then move the VMs back to the new drive?  This is what I do and it requires little to no downtime and works easily as long as you keep enough free space on your san to temporarily move the files off your old LUN and onto a new LUN.

Obviously, nothing here is ideal and will continue to be a problem until HP impliments space reclaimation for the SAN.