This post is more than 5 years old
10 Posts
0
179285
Compellent free space reclamation
Hi guys, I'm installing a new storage Dell Compellent SCv2020 with vSphere 5.5 U3b for an existing customer
I have deep knowledge of Dell PowerVault storages but I'm new with Dell Compellent. I've read that making snapshots for backups or running Storage vMotion can affect free space on the storage due to native LUN thin provisioning: communities.vmware.com/.../436628;tstart=0
The suggestion is to manually run or schedule the command "esxcli storage vmfs unmap -l MyDatastore" that invokes the SCSI UNMAP primitive
I've tested a Storage vMotion and indeed the space on the source LUN remain allocated until manually "unmapped". I've not yet tested how backup snapshots affect the free space on the storage because I've not yet migrated any VM on the new infrastructure (by the way I will use Veeam Backup v8 that heavily rely on VMware snapshots)
Anyone here can share ideas and feedback? Many thanks guys
Marco
mluethye
26 Posts
0
December 30th, 2015 09:00
You are spot on, use the unmap tool to reclaim space on the LUN. Compellent thin provisions everything on the back end and depending on your environment, you should probably be thin provisioning your LUNs on the VMware end. Over time as you delete servers in VMware and vmotion from datastore to datastore that space will not be reclaimed on Compellent b/c the storage is ignorant of the fact the data is no longer being used. The 'fix' for this until there is complete integration between VMware and Dell is to run that unmap tool.
What we started doing every quarter or so is to run the unmap tool on all LUNs that are thin provisioned (in our environment that is all volumes). I've tested unmap multiple times and I have had no issues with running the tool while servers are running on that volume. I would suggest running the tool on a server that doesn't have a lot of workload because the CPU does spike during the unmap period (datastore reads will spike as well).
As for backups and snapshots I have no issues with space regarding those. Replays will freeze blocks of data for reference, but once that replay is expired the space is recouped. Regarding backups it depends on what you are using for backups - we use CommVault which will take a snap of the LUN, copy the data off then release the snap (so for a short period of time there is space eaten up during the backup but it takes only a few minutes to copy off the data).
mnovelli
10 Posts
0
December 31st, 2015 00:00
Many thanks for your feedback. Did you script the UNMAP tool or you run it manually? I've tested a couple of scripts in PowerCLI but they time-out after some minutes
mluethye
26 Posts
0
December 31st, 2015 07:00
I do not script them. I did play with that idea early on with PowerCli (there is a command for unmap) but it's tricky b/c of the timeout. When I was reading up on this there is a default timer on PowerCli commands of like 15 minutes. You have to change that value and from what i remember there is a max timeout of like 30 minutes, which is okay for smaller volumes but if you have anything above 4tb it may take around 30 minutes to complete. I also would prefer to run the tool one volume at a time, which get's tricky with scripting b/c you have to trigger when the script has completed the process - something you cannot assume when it's done it's done b/c of that timeout issue.
Sorry long story short - yes I looked into scripting it and started down that road but quickly realized it was far too complicated for what we needed to accomplish. We have about 30ish volumes, i can easily just knock that out in a week doing a half dozen or more a day. If you worked for a major corporation that had hundreds of volumes I can absolutely see the value in scripting, for us it wasn't worth the time cost of developing the script.
If you figure out an easy way to script it shoot me a message - i would love to automate it with PowerCli.
mnovelli
10 Posts
0
January 2nd, 2016 08:00
Hi, I've found a super cool script here: communities.vmware.com/.../2483410
The script is named vmware_unmap_datastore.ps1.zip - 3.1kbit
Be careful to run set-executionpolicy unrestricted before running the script and remember to create a C:\Temp folder before running the script. I'm using it in interactive mode, it is super fast due to advanced calculation of block size to use with UNMAP command. Default esxcli command on my 1 TB LUNs took about 4 - 5 minutes, with this script just a few seconds
Be careful to point the script to you vCenter server and to have at least vSphere 5.5. The script will automagically detect thin-provisioned LUN at storage level, so don't worry if you have a mix of thin and thick LUN on different storages. It will UNMAP only thin-provisioned LUNs
Marco
mossaab.ouaddari
1 Message
0
January 19th, 2018 07:00
Dear,
We faced the same space reclamation issue with Dell Compellent but with Solaris 11 installed in a physical Dell blade. The filesystem is ZFS. Do we have to manually u,map and map again the LUNs to free space? we are already running out of space and we are seriously considering migrating to REHL.
Your help is very much appreciated.
Mossaab