HPE Storage Users Group
https://3parug.com/

UNMAP - > RCG .r VLUN disk not shrinking? / Total reserved
https://3parug.com/viewtopic.php?f=18&t=3260
Page 2 of 2

Author:  MBILC [ Wed Aug 07, 2019 3:43 pm ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

Good to know.

When i check the usage of the snapshots for those RCG's it shows them using 0 space , the _LG volumes right? But i do see under Snp it shows usage...

Image

In order to delete the snapshots, would i just go into Restore Points (in the GUI), select the one i want and choose delete?, example _vm_001_LB

We are hoping to get Patch 149 (includes 144) installed possible tomorrow or Friday on our 2nd 3PAR system, lets see if that helps free some more space up!

Author:  MammaGutt [ Wed Aug 07, 2019 4:26 pm ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

They show 0 and a * per snapshot. The reason is simple, if you have 20 snapshots, what sense does the size per snap matter? If you remove snap #10, you will most likely not get that space back because some/most/all changes here will be consolidated into some of the other snapshots. So everything is consolidated in the base volume.

I haven’t played alot with snapshots and SSMC, but what you are saying sounds like the logical place to remove them.

As for reclaiming space faster...the higher percentage of free space per volume usually improves reclaim. As for converting to vmfs6 my experience is to either clean out full datastores and delete/re-create or clean out full, run unmap, reformat datastore and storage vmotion back. The first one gives instant reclaim, the second gave me very good reclaim speed (~100+GB per day)


Also remember that for VMFS6 you want your vmdks to be thin for auto unmap, rather than eager thick zero as you would with vmfs5 for best zero detection reclaim.

Author:  MBILC [ Wed Aug 07, 2019 4:40 pm ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

Your going to have to send me a bill for all of this! More info than HPE support so far.

I had thought of the vMotion, what I have done instead as we changed entire naming scheme and everything was just built out new Virtual Volumes, presented them, new AO's and new CPG's, basically did a Spring cleaning of the entire system, but due to all of these old datastores space got tight and the slow releasing of this space I basically hit a wall, along with several VM's using RDM's cause of SQL and Windows Clustering.

It has been fun to say the least but I am about 80% of the way there.

We are doing all thin on the Vmware side, and all thick on the 3par backend.

Will keep you posted on the removal of the snapshots and also if patch 144/149 helps.

Author:  MammaGutt [ Thu Aug 08, 2019 12:29 am ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

Why thick on 3PAR? That is basically throwing money out the window. With thick a 6TB volume will always allocate 6TB and unmap will basically make no sense as there isno ppoint in trying to shrink.

Author:  MBILC [ Thu Aug 08, 2019 9:26 am ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

That was a decision made by someone else. The thought was they did not want to have to worry about the back end storage and over allocating space and deal with it instead on the vSphere side (more familiar). My suggestion was to then use VVOL's instead but that is not something they wish to try at the moment.

Future plans is to move a lot of this data into Microsoft services and cloud so total space in the end will shrink down by about 30-40TB initially once some projects are completed. Basically these 3PAR systems will be run until they can't any more and by then the data foot print should be small enough to then possibly migrate to another option.

Author:  MammaGutt [ Thu Aug 08, 2019 12:05 pm ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

Okay, for those that decided that. Ask them the following questions:

1. With AO and thick provisioned volumes, at which tier will all the unused space be?

2. With AO and thick provisioned volumes, at which tier will all new VMs and new writes to existing VMs hit?

3. With AO and thin provisioned volumes, at which tier do all new writes and new VMs hit?

Author:  3parker [ Fri Aug 09, 2019 9:14 am ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

Does running an UNMAP blow up the snapspace with the unmapped blocks if you have a snapshot on a datastore volume?

I want to do some housekeeping on our datastores but we run RMC with a small snapshot chain so we can take incremental backups. I suspect I'll need to remove all the RMC snapshots after running an UNMAP to free up the snapspace, and I think this will force RMC to take a full backup of each datastore.

Cheers,
Adrian

Author:  MammaGutt [ Fri Aug 09, 2019 12:27 pm ]
Post subject:  Re: UNMAP - > RCG .r VLUN disk not shrinking? / Total reserv

It depends.

The snapshot will contain all blocks required to "reset" the volume to that point in time. The blocks deleted when using unmap usually contains some data so the snapshot should grow.... However I'm not sure if it will grow instant or when the volume actually shrinks.

Page 2 of 2 All times are UTC - 5 hours
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/