Lefthand - Overprovisioned cluster how to reduce the size

[luumanio]

Hi,

I have already read lots of posts related to overprovisioned volumes on Lefhands but I still can't understand a few things:)

I have a situation with my Lefthand where my cluster reports the following status:

"normal, overporvisioned and 81%full"

I have three volumes in my cluster (2x LH P4500, each 6.55TB RAW and 5.31TB usable space)

ds1 reported size 1.7TB, consumed space 2.53TB, provisioning Thin - VMWare datastore

ds2 reported size 1.7TB, consumed space 2.78TB, provisioning Thin - VMWare datastore

fs1 reported size 2.5TB, consumed space 3.34TB, provisioning Thin - iSCSI folum for file server

I've tried to change Thin provisioning to Full on fs1 and as soon as I did it I got red warning message in CMC saying that my cluster running out of space @95%. Accordingly to CMC I still have 2TB available, please see attachment. FYI, I don't have any snapshots in my cluster

Is there any way to reduce a space in my case? 

I know you will tell, buy additional SAN but I just don't undestand one thing. When I check the datastore size in VMWare for ds1 and ds2 datastores I see capacity 1.7TB and provisioned 1.41TB and 1.96TB, where CMS reports consumed space way bigger than on VMWare.

It just doesn't make sense for me but hope it makes for you and you can help me undestand this:)

Thank you for all your help in advance!

[oikjn]

That summary page on CMC is always tough to read.

I assume you are using best practice and are using NR10 for each LUN. If that is the case, consumed space will be 2x the size of the volume (as much as it is expanded). So... if VMWare says your drive is expanded to 1TB, it will consume 2TB of space on the SAN.

DO NOT use thick provisioning.

Adding up your SAN consumed space totals to 8.65TB which means all your drives have expanded to 4.3TB. If you don't have 4.3TB worth of data there is potential for you to save some space. If you have 4TB worth of data or pretty close to that, there isn't really anything you can do other than write a check to HP and buy one more node to add to your cluster.

If you are only have something like 3TB of actual data on those LUNs there is the potential that you can save space, but unfortunately there is no way with this SAN to reclaim space on an expanded thin volume, your only option is to migrate the data off the LUN, delete it and create a new thin LUN. If you don't want to go through the trouble of setting up the initiators again, you can just shrink the reported size of the LUN and then re-expand it, just keep in mind that any data on that LUN would be corrupted so you would have to reinitialize the volume after that process completes.

Lack of pace reclaimation is a major weakness IMO for this SAN and I"ve voiced that concern w/ HP, but as always, the more people who complain about that missing feature the more likely they are to do something about it.

[luumanio]

Thank you for your advice.

I think I have another idea, my VMs are split between two datastores ds1 and ds2 and since I still have 2TB free I was thinking to increase ds1 size from 1,7TB to 2TB.

I still have about 700GB free on ds1 so increasing datastore size by 300GB gives me enough space to move all VMs from ds2 to ds1. Once I move all machines to biger datastore I can decommision empty ds2 datastore and reclaim 1.7TB this way I should be able to get rid of the overprovision alert on my SAN, but I guess I will still need extra LH node in the future.

What do you think about this idea?

[oikjn]

if you migrate everything off a lun and then shrink and then expand the lun you can reclaim space.

Warnings about being overprovisioning aren't a problem and there is no need to do anything there.  The problem is when your thin volumes expand more than you want and then your AVAILABLE space gets too small.

You can have 20 thin 1TB on your SAN with 100MB each on them without having a problem on your san...  you will have a problem if suddenly they decide to expand out to 1TB each.

Watch your PROVISIONED space.  Assuming all your LUNS are NR10, provisioned space should roughly be equal to 2x your actual data stored on the cluster.  If your provisioned space is way more than that you can free up some provisioned space by deleting LUNs or shrinking them manually.  If you just delete all the data off a LUN and leave it as is on the san it will continue to consume the same amount of space it did when you had data last on it.

You definitely WANT to keep your san overprovisioned.  The reason is because storage is $$ and every volume wants an amount of free space which IS needed for most things to run correctly, but when you add up all the free space on all your volumes it can become a significant waste.  Instead, if you use thin provisioning you can shrink the effective amount of free space significantly and subsaquently save you a lot of money.

[Server-Support]

So how do we get the warning before 95% by email ?

[oikjn]

not sure about that one.  I think they only set that one up as a critical point and anything lower is something you just need to monitor.  Otherwise you will be flooded by warnings. 

If you really want, setup a dummy thick lun to make up the difference between the 95% warning space by default and the % warning you want and then when it gets to the fake 95% you will get your warning and you can delete or shrink the fake LUN to free up some space before things get bad.  Either way, if you are going to overprovision at all or use snapshots, I suggest you use a dummy thick LUN of some size to allow you to quickly regain some space and some breathing room in case something suddenly takes over your SAN storage with little notice (or you are like most people and you can't dedicate 100% of your time to monitoring the storage system)..