Windows 2012 R2, Fail Over Cluster - Nimble Reported Storage

[oromero51]

We have been enjoying the performance and features of the NibleStorage. Very happy with the choice we made for our System Center Virtual Machine Managers > Failover Cluster > Hyper-V setup.

We do have a, I guess for lack of better term, "a feature request" that we considered very important. We did have a system outage because of this.

Not fully understanding what was happening, I did open a ticket with support team, and explained the situation.  After a an email exchange, and a couple of calls, I did come to understand what was happening and do get the point of the Support Engineers, but I don't think that they understand mine. I am hoping to find someone that can

In a nutshell, the issue is that since the Servers(Windows 2012 R2 Nimble Windows integration kit installed) using the Nimble Storage are not aware of the compression, it does not see the space saved as space available for use.

For example, if a have a LUN of 1.5 TB. The host does see it as a 1.5 TB Disk. However, if the sum of the files on this LUN equals 1.45 TB, although because of the compression(Nimble) the saved space can amount to 600GB, the Server still thinks that it only had 50 GB free. And what is worse, when it gets to 1,499 GB the Server will report it as full and will stop writing to it. It happened to us.

Here is an example;

Do you see the problem? Anyone having this issue? How are you dealing with it? Do we just over provision the LUNs?

Regards,

Oscar Romero

[thoule26]

We are aware of this problem too and we simply over provision the LUNs.  Their engineers are aware of this and it is simply the nature of the beast. 

[oromero51]

Talbert, thank you for sharing how you are dealing with this issues. This gives me some confidence that I am on the right track. When I open the support request on 9/23, the Engineers did not seem to be aware of the issue, or at least they did not own to it. So I was not provided any guidance.

Before I can leave this as the correct answer, are you aware of any limitations? For instance, can you over provision, past the amount of storage you actually have? 

Say that I have 15 TB on the CWS300, and  for easy math, say that I have a need for 15 LUNs. So I decide to create each LUN with a Size of 1 TB,   can I add a 16^th LUN and size it 1TB if the actual free storage is 1TB?

Can you think of any other side effects?

[thoule26]

Hi Oscar,

As long as the space being actually used is not over what you have, you can over provision.  Obviously, you need to stay cognizant of amount you're using so you don't fill up your array.  Besides that, the fact Nimble zero's out any unused space, you're good to go. 

[marktheblue45 · Accepted solution]

The issue you are observing is the same behaviour that you see with VMware datastores on Nimble Storage. Classic example is with Datastores or even In-Guest iSCSI presented volumes. We have a number of SQL DB only volumes with average compression of roughly 3:1Datastore is 1 TB in size on VMware and the Volume on Nimble is 1TB also. The Nimble array reports it is using only 340GB with compression but the datastore is full 920GB (almost full). So I have to increase the Datastore size and volume size.

On a positive note you are really only consuming the space after the compression. Sure it can be a little confusing at times. At least with Nimble it's really easy to know how mush free space is available or indeed consumed in one quick glance on the array GUI.

Mark.

[marktheblue45]

Don't forget the "White Space" issue observed with Thin Provisioning. However if it was a white space issue you would see that the Nimble side would be showing more space consumed than the VM if Compression was actually 1:1

[jliu79]

I don't really see it as a problem. I also have a power shell script to alert me if the free space on the CSV volumes is below 10%.

[oromero51]

You are correct Jason Liu, but once you know about it.  However, we did not knew about it. Had to learn the hard way. And when the issue arise, and contacted support it was not acknowledged.

regards,

Oscar Romero

[oromero51]

This is the powershell script that we setup to monitor the free space as seen by the Host. You can run it on any member of the cluster;

$username = "Place user doamain\user    here. has to have at least read access to cluster"

$password = "password goes here"

$secpasswd = ConvertTo-SecureString $password -AsPlainText -Force

$credential = New-Object System.Management.Automation.PSCredential ($username, $secpasswd)

$Param1 = "type the cluster name here"

$Param2 = "50"

$Text = "NOTICE"

$Arguments = @($Param1, $Param2)

$Job = Invoke-Command -ComputerName $env:COMPUTERNAME -Credential $credential -ArgumentList $Arguments -ScriptBlock {"‪C:\Windows\MyMonitorFolder\CSV\CSV_20.ps1"}

If ($alert -gt 0){Write-Host $Text": One or more drives have exceeded the threshold value of" $SpaceThreshold "% remaining!"}

Write-Host "Cluster:"$ClusterName

$objs | ft -auto Name,Path,@{ Label = "Size(GB)" ; Expression = { "{0:N2}" -f ($_.Size/1024/1024/1024) } },@{ Label = "FreeSpace(GB)" ; Expression = { "{0:N2}" -f ($_.FreeSpace/1024/1024/1024) } },@{ Label = "UsedSpace(GB)" ; Expression = { "{0:N2}" -f ($_.UsedSpace/1024/1024/1024) } },@{ Label = "PercentFree" ; Expression = { "{0:N2}" -f ($_.PercentFree) } }

return $alert