HPE simplivity physical capacity calculations

[prezidentj33]

Hello,

Goal: stand up new file servers to migrate to so we can decommission our old file servers. about 17TB of data total

Problem: Not sure if I follow through if I will run out of simplivity physical capacity or not. How is this calculated or is there some way I can figure out how much space will be used? I am afraid if I continue with replicating data that we will run out of physical capacity and things will go sideways quickly. 

We are doing some server migrations and trying to stand up servers along side the old to migrate easily. These are file servers with alot of drives attached and 10TB of storage. Currently have 2 file servers both with roughly 7TB and 10TB per box. 

Standing up new servers and replicating data (DFSR) to new file servers. Total of 4 or 5 new servers with data being spread across those as we are splitting up some of our data.  
Because of this that means 5 new servers with 84GB C drives that are from the same template and all of that data dispursed among those 5 new servers. So we are essentially doubling our space used for storage. 

Our HPE Simplivity shows Total Capacitty 38.6 TB. Capacity used is 33.0TB. Capacity Free is 5.6TB. 

The cluster shows we have 28.5TB free storage space. 

We have backups running for most VM's in simplivity as well. Cluster has 2 compute nodes and 3 compute/storage nodes. We are expanding this to 5 compute/storage nodes soon. 

Can anyone help in how I would figure out how much physical capacity is going to be used up in order to make sure I don't destroy our simplivity capacity?

Thank you!

[DaveOb]

This probably needs a deeper discussion then a forum can provide,it may be an idea to discuss the situation with your local HPE team / solution architect,their are potential pitfalls.

[BenH10]

I have already done this and can tell you exactly how to do this leveraging SimpliVity's Deduplication to make sure you don't take up extra space while having multiple copies of the same data.

WARNING!!! This is not in official documentation

SimpliVity stores data on the per VM and per node basis, meaning there is a copy of a vm on two nodes at any time. What you need to do is make sure the new file server (without the drives) is setup and running. Then login to an OVC through SSH, elevate to root, change your source to /var/tmp/build/bin/appdata, and run the command DSV-BALANCE-MANUAL. Look at where the copies of the VM are in you cluster and then using the excel sheet from the command, change the new server to be on the same nodes as your current server. You have to do this on a remote cluster if you do remote backups. Exept you are now having to use the rbm2.pl script (only support has) to move the data around. Our you can just call support and have them do all of this. 

Now that the Old File server and New File server have all of their data on the same nodes. Just copy the VMDK's from the Old File Server to the New. It will copy all of the data, no matter how much, in seconds. This is how you know you where successful as it didn't copy any data, but pointed existing data to a second VM. This is how you stand up a new version of a server without using any mopre physical space. 

[prezidentj33]

Hi BenH10,

This is a great answer. I really appreciate it. See below for the scenario. keep in mind these are using DFSR. 

For the other box I need to replace, that one is on the new cluster. would it be just as easy to vmotion the server to the same node as the current one is on? or is it not that simple? I have already created drives on this one and have all of the drives replicated except 1 drive which is about 400-500GB. I'm nervous to replicate that in worry physical space will be chewed up. Will that chew up 400-500GB of physical space then if I do not use your explained method?

Here is current status of my situation

OLDserver1 - on OLD cluster using approx 10TB total between all drives

NEWserver1 - on NEW cluster currently has 1 more drive to replicate with OLDserver1 which is 1.3TB then the ones I am mapping to this server will be done for the 4 drives I am using on this server.

NEWserver2 - on NEW cluster needs to replicate with 2 drives on OLDserver1 which equates to 3.5TB - This one has no drives replicated yet

OLDserver2 - on it's own NEW cluster - has 3TB data to replicate to NEWserver3

NEWserver3 - on NEW cluster needs to replicate 1 drive with 3.5TB of data with OLDserver2. 

I have one more subset of servers I need to replicate with but those are on another cluster and will use that approach on what I do here.  Not sure if I made things worse or if you can decipher my information haha. 

I guess i'm still not quite 100% how I will do this based off of my situation. 

[BenH10]

As I do not know how your SimpliVity nodes are physically configured and how you cluster them, I will not be able to tell you exactly how to do it. But the data has to already exist for you to not use more physical data. As in, if you are upgrading a VM with a new VM and they are in different clusters/datacenters, you will use at most 100% of the data in physical space. Assuming no efficiency, you will always get better due to their being similar OSs in the VMs. 

I recommend using the DSV-BALANCE-MANUAL, downloading the CSV file it creates and just reading through where all of your copies of the VM exist on the physical nodes. You can try to vMotion the VMs to be on the same node, but that does not necessarily mean the data is on the same physical host. The easiest way to see if a copy of data exists on a physical host, vMotion the VM over to it and see if you get the alarm "SimpliVity VM Data Access Not Optimized", this means you are running the compute of a VM on a different host than either of its copies of data. 

Always assume 0% efficiency when copying data!!!!!

When a node hits 100% capacity, it will go into "Read-Only" mode, this is not as safe as it sounds. "Read-Only" mode stops the node from sharing out storage over NFS, meaing your VMs will hard crash. As always, call support and they will webex in and help you with these moves. They would rather help move data then deal with a node in "Read-Only".

[SimonHorwood]

Have a look at this link, it will give you the basis of a script thast you can run daily to monitor your physical capacity on your clusters

http://www.vhersey.com/2017/03/17/quick-simplivity-omnistack-capacity-script/

Cheers

Simon