This post is more than 5 years old
1 Rookie
•
51 Posts
1
3512
When i delete the data from the host, in backend symmetrix vmax the volume doesnot reflect the space freed up
Hi folks,
I have a windows 2008 R2 host which is connected to VMAX. the lun size is 200GB. i have data of 180 GB in it.
When i delete data from the LUN, for eg: 100GB in the host. however the space utilized in the backend thin device shows almost 90% used only.
Does the tracks in the backend will be 0'ed ??
If i use space reclamation the 0 tracks only freed up right?
Why does the space is still showing as 90% used?
Cheers,
Paul
M_Salem
213 Posts
4
May 18th, 2014 14:00
Hi Paul,
Thanks for contacting EMC Community.
What you are seeing is normal. When you delete the data from LUN, It will not zero out the deleted blocks on the array. With ZSR (Zero Space Reclamation) you cannot reclaim previously written but deleted tracks unless it is all zeros (which is rare to happen) or marked NWBH tracks, so the way we use to reclaim dirty tracks in windows OS is as follows:
If you are talking about w2k12, it supports SCSI T10 TRIM, which is almost the same like UNMAP for ESX5.x so it will do the same behaviour. (It will set dirty tracks as NWBH and will be deallocated and returned to the pool)
For W2K8, There was an awesome tool called (Storreclaim.exe) however, it went EOSL in Q1 2013. This tool was using WRITE_SAME SCSI T10 so it will mark the previously written but deleted tracks as NWBH (fill them with zeros) and then you can run (start reclaim on tdev xxx) from SE host side. .
@semc,Thin Reclamation Utility (TRU) is used for z/OS Mainframe not open systems
http://www.emc.com/collateral/hardware/white-papers/h10688-z-os-virt-prov-best-practices-wp.pdf
@Dynamox, Yep, sDelete will do the same behaviour as with Storreclaim but you have to be careful when using it as you sDelete will fill up the whole free space on TDEV with zeros so you will see written and allocated tracks reaches 100% so no space for active applications using this tdev. (Personally, I won't recommend it, running unsupported storreclaim will be my first preference!)
Hope that helps
Mohammed Salem
SEMC1
9 Posts
1
May 18th, 2014 07:00
The Thing is that Operating level differs than the Storage level so it does not reflect
I read once that there is an option called Trim which should be active for Host so that storage reclaims the deleted Data automatically
there was a tool EMC had called Storereclaim which can be run to do the same but EMC stopped that tool and removed the Documentation for so i guess it is replaced , i will check for the new one and let you know
there is Thin Reclamation Utility
never tested but i read it is compatible with the different Storages
SEMC1
9 Posts
0
May 18th, 2014 07:00
and migrating Luns will help removing the deleted space too
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
May 18th, 2014 12:00
didn't sdelete allow you to write a bunch of 0 to free space, allowing reclaim to get them back ?
M_Salem
213 Posts
1
May 18th, 2014 15:00
For w2k12, Kindly find page 31 of Virtual Provisioning Space Reclamation whitepaper: :
http://www.emc.com/collateral/software/white-papers/h6730-virtual-provisioning-space-reclamation-wp.pdf
It is worth mentioning that UNMAP/TRIM is supported by ucode 5876.159 and later. Another thing you should be aware of (For UNMAP with ESX not sure if it applies to TRIM with 2012, I believe it is the same but haven't tested it myself before) is when the extents are persistently allocated, That will block UNMAP from deallocating the tracks so as a workaround to this, if your TDEV was originally created with Persistently Allocated tracks, You can unset this persistent allocation by running a new command was introduced with SE v7.4 to unset the persistent allocation on the TDEV.
The command is: symdev -sid xxx unset -persistent
Hope that Helps
Mohammed Salem