HP StoreVirtual VSA 2014

[barjolu]

Hi,

I'm installing HP StoreVirtual VSA 2014 and have a few questions.

We have three vsphere 5.5 hosts with local storage.

On every vsphere host we have created a datastore with the size 2 TB.

This datastores will be used to provide disk to StoreVirtual VSA, but how much should I provide?

Should all of the 2TB be provided to HP StoreVirtual VSA, or should we leave some free space? The datastore won't be used for any other virtual machines or anything else.

Another question is regarding placement of the VirtualCenter. Since the installation of VSA needs a up and running VC, it's located at a local storage. But, when VSA is installed and we have created our shared iSCSI SAN storage, is it best practice to move the VirtualCenter to the shared storage, or is it recommended to run it on separate storage?

//Johannes

[oikjn]

Thats more a personal decision.  You shouldn't really share disk resources with the storage drives of the VSA, so generally you should dedicate them 100% to the VSA or you will end up with random performance issues.  As for how much space to provision for the VSA... thats up to whatever you need.  Just keep in mind that all your VSA nodes should have the SAME amount of space presented as that if one node has less space presented then all the nodes will run with the amount of space that the smallest node has.

Can't talk about the virtual center control as I'm a hyper-v guy, but in general watch out for a chicken before the egg situation and think about what happens if the cluster is down, do you need that install to run the cluster?  If so, don't put it on the SAN... just like the FOM, don't think about putting the FOM on storage presented by the SAN either ;)

[barjolu]

Thanks!

Any vmware guys out there who can confirm it is safe to use 100% of a datastore as a disk in HP VSA? The vCenter will make nasty alarms per default, if a datastore is 100% occupied..

Another question is the size of the volume created in StoreVirtual. If I have a 3 TB available and wants to create a raid-10, should I create it as big as 1,5 TB or do I need to save some free space? In the EVA for example, you shouldn't allocate more than 85% of a disk group.

When presenting a volume to vmware hosts, should the hosts belong to a Server Cluster in StoreVirtual, or should the volume be presented separately to each host, or doesn't it matter? :)

Regarding the Virtual Center, I guess I can do some testing to see if it's ok to store in the Store Virtual SAN-storage or if it should be on a separate storage.

//Johannes

[oikjn]

windows will complain the same with a full drive.  To get around this, I just present the drive to the VM as a direct passthrough and then it doesn't matter.  In theory its also less overhead for the host, but I don't think its that much these days and I really just do it because it lets me present the most storage to the VSA.  Presumably esx has a way to present the drive as a passthrough too.

As for how much space to allocate.  NR10 requires TWO VSAs.  When you set a LUN size you set it for the presented sizse, but you have to mentally know that if its NR10 it could potentially fill to 2x that space so the SAN needs to say it has at least 2x the space compared to the size of the LUN you are about to create.

Suggestion 1 -  even if you aren't going to overprovision the SAN, use thin provisioning as it will make a difference with your snapshots (any snapshots of thick LUNs are thick as well for some crazy reason).

suggestion 2 - create one THICK LUN with something like 50GB of space and label it EMERGENCY RESERVE.  Keep it thick so the san takes that space out of the available space and then in the future, if you actually run out of space in the SAN you can delete it in 20 seconds and get your SAN back online without having to make a terrible decision on which production LUN you are going to delete to get the rest back up.

related to suggestion 2 and your question about how much space to leave.  The SAN will start warning you at something like 90% utilization, but its not a huge warning and it can be easily ignored.  The san will be happy to run with one thick LUN that is 99.9% of the available space, but the moment you get to 100% the SAN will turn off write access and everything becomes read-only and then all your servers will freak out ;)  

[barjolu]

Thanks again :)

Last question for now:

Since I haven't presented all disk (from vmware datastore) to the VSA, how do I extend the disk in VSA?

I have increased the size of the disk in vmware, but it seems that the VSA isn't aware of this. In the management console of the VSA the disk size is the same as before.

After a reboot of the VSA it recognizes the new disk size, but is that the correct way to do this? Can't find any documentation regarding this. Isn't it supported to increase the disk?

//Johannes

[oikjn]

not sure about increasing the same disk, but I know its supported to add additional disks.  Anyway, I've increased disks on VSAs just like you said and it has worked, but if this is a new VSA and its going into production and this isn't the ONLY VSA (ask deleting it doesn't delete the management group and wouldn't wipe out any settings you already have configured for the group).  Then just to be safe I would just destroy the VSA and recreate it as it takes all of 15 minutes to build a new VSA.

[fourg]

VMware guy here,

Yes we set our VSA's to use 100% of the local storage and we put all VM's including vCenter on the VSA storage.  Doing so provides redundancy (RAID 6 on hosts + 2 way mirroring on VSA cluster) so in the event one host goes offline we can either have VMware's fault tolerance setup to start critical VM's up on another host or we can manually attach the VM to another host and power it back up without risk of losing any data.

The VMware host does not need to be in vCenter in order for you to present LUNs/Volumes to it from a VSA cluster.  Each hosts connection to the VSA cluster is independent and done with iSCSI targets.