May 19, 2012

Installing and Configuring NetApp VASA Provider for vSphere

NetApp VASA Provider Installation and Configuration

Whoa, how did I miss this? Here is the Beta access for our VASA plugin for all you vSphere 5 folks out there! http://t.co/uDLVHkiM
@keith_aasen

As @keith_aasen pointed out on Twitter recently, the NetApp VASA Provider is currently in beta. I configured it in my lab over the weekend and wanted to outline some of the steps to get this running.

First off, if you are not familiar with VASA, it is a new feature if vSphere 5 that stands for vStorage APIs for Storage Awareness. Essentially rather than having to creatively name your datastores something like netapp_3140_prod_sas_600g_sm the array (or the provider) can tell vCenter what features it has – things like disk type (SSD, SATA, SAS, FC), replication so you know whether or not the volume is part of a SnapMirror or SnapVault relationship as well as the ability to tell if deduplication is enabled for the volume. If you want more information, there is a really good post here that has more details on how each vendor will implement VASA.

Requirements

  • NetApp controller running ONTAP 7.3.3 or later
  • A Windows 2008 x64 machine to install the VASA provider on (can be physical or virtual)
  • vCenter 5.0

Installation

This part is pretty straight forward, click next a few times and you are done. One thing to keep in mind is this should not be installed on the vCenter server. In my case I installed it on a standalone VM. If you install it on the vCenter server you will run into a port conflict and find that the VASA provider service will fail to stay running.

Port Conflicts

After you download the netappvp–1–0X2-winx64.exe file from here, double click it to start the install.

Install 1

In my case I’m fine with the default installation directory

Install 2

Click install

Install 3

And you’re done, leave the checkbox for Launch VASA Configuration to automatically launch the configuration window (alternatively, you can double click the shortcut on the desktop)

Install 4

Configuration

The first step is to enter the user account information to register the plugin in vCenter. Enter the account info and click Save. Then enter the vCenter information and click Register Provider.

Configuration 1

After you click Register Provider you should get a prompt telling you it was successful.

Configuration 2

Now we need to add the storage controllers.

Configuration 3

Wrap Up

The only thing left to see is within the vSphere client, if you go into VM Storage Profiles you will see some storage capabilities have been added:

Capabilities

And lastly, if I go into the datastore view I can see the capabilities of my current datastore:

Capabilities 2

Ironically it shows up as high performance even though it’s just the NetApp simulator but that is because the simulator sees it’s “disks” as SAS drives. It showed Replicated because I had setup a SnapMirror relationship in the background to make sure that it would detect it.

I think VASA will be a big benefit to VMware environments, but it still has some work to be done. I like the idea of being able to create a storage profile based on these capabilities that are sent up from the array provider (and being able to see those capabilities in the datastore list like the screenshot above), and more importantly, to ensure that the VM that I created on the “Gold” storage profile that is on high performance, replicated disk stays compliant with that profile by utilizing storage profiles.

Popularity: 11% [?]

Deploying Apple OS X virtual machines in vSphere 5

When I saw the announcement that OSX would be an officially supported guest OS for vSphere 5, I was pretty excited.  Being an Apple fan and a home lab guy, my first thought was I could run Snow Leopard as a VM and treat it as a centralized iTunes server.  It was announced that it would only run on the (now retired) Xserve hardware, but I figured for home lab purposes I could at least create the VM to test things out.

I quickly realized it wasn’t going to happen, my home lab does not contain any Xserve servers to use as vSphere hosts, they are whitebox ESXi hosts.  I was able to create the VM with a guest type of OS X 10.6 (64-bit) but as soon as I tried to power it on I was greeted with the following message:

Error when booting VM with guest type of OS X on non-Apple hardware

So there is some type of checking that happens to ensure that Apple virtual machines will only run on Apple hardware.  The other thing I found interesting is that virtual machines with a guest type of OSX don’t migrate off the host when going into maintenance mode.

Popularity: 6% [?]

Changes in vSphere 5 Licensing

As you may have seen already, VMware today officially announced changes to the vSphere 5 licensing.  After some of the fallout regarding the initial vRAM licensing, here is a screenshot that shows both the initial vRAM licensing as well as the new vRAM licensing entitlements:

The vRAM licenses are tied to a processor, but you are no longer limited on the number of cores per processor(nor physical RAM in the ESXi host).  Another new change announced today is that the vRAM entitlement is calculated based on a 12 month average, rather than a high water mark.  There is a monitoring tool built into vCenter 5 that will allow you to track this.

Screenshot from Licensing Reporting screen in vCenter 5

 

One thing to note with that, if you are using VM’s that have over 96GB RAM allocated – you only deduct 96GB from your vRAM pool.  However, to accurately report the licensing you will need a (free) utility for tracking vRAM usage initially.

In case you didn’t catch that….deploying a VM that is assigned anything over 96GB RAM counts the same as a VM that has exactly 96GB.  If you have a 1TB virtual machine, the cost for the VM would be more than 10x if you were not using the new licensing model.

Bring on the Monster VM!

Another concern people have had is how does this affect my VDI environment?  Will I be paying based on vRAM entitlements for this as well?  VMware has answered this with vSphere Desktop licensing.  The idea is you can only use it for desktop virtualization but you are only paying based on the total number of powered on desktops (unlimited vRAM entitlement).

 

 

Popularity: 3% [?]

Using VAAI with VMware and the NetApp 8.0.1 Simulator

2011-12-01 Update: I have a new post on using the latest version of the NetApp ONTAP Simulator (8.1) here, follow those instructions to deploy the newest  simulator in your environment.

This one is pretty straight forward, but I had a few people ask so I am making a post on it.  As you may know with vSphere 4.1 and NetApp ONTAP 8.0.1 VAAI is enabled by default on an iSCSI or FC connection.

Click for larger image

I’m using my simulator setup I described in my last post, and now you can see below I have added into my vSphere environment and is now being displayed within the NetApp Virtual Storage Console.  You can see even though I haven’t “configured” anything for it on the NetApp side, the VAAI capable column for TST-NA1 shows enabled while the other 2 arrays (a FAS3020 which is not running ONTAP 8.0.1 – also of note is this platform is not capable of running ONTAP 8.0.1, only ONTAP 7G) are displayed as not being VAAI capable.

Click for larger image

You can also verify this from the vSphere side (either by enabling Remote Tech Support (SSH) if using ESXi/using the vMA/Service Console if using ESX classic).  In my case I am using ESXi and chose to use the vSphere Management Assistant (vMA).  I ran vicfg-scsidevs -l and was able to see that my NetApp iSCSI LUN is supported for VAAI.  NetApp also has published a good Technical Report on using VAAI which goes into detail on how to view the VAAI statistics from the array side as well using the stats show vstorage command.  It does a good job of explaining what the counters are not only from the stats show command on the array but also from the esxtop output as well.

Popularity: 23% [?]

VM Alignment for NFS datastores with ESXi using the vMA

It’s been pretty well documented that using the NetApp mbrscan/mbralign tools aren’t currently supported in ESXi, the main two ways around this that I am aware of are:

  • Keep an ESX host around so you can still use the alignment tools from the Service Console
  • Mount the datastore (if using NFS) to a Linux host and run the alignment tools from there

Since I had already deployed the vSphere Management Assistant (vMA) I figured I would use that as my Linux VM.  In order to make this happen there are a couple of things we have to enable, first login via SSH to the vMA using the vi-admin user and run the following:

sudo /etc/init.d/portmap start
sudo /etc/init.d/nfslock start

Next I created a folder for mounting my NFS datastore (named na2_test_500g_sata)

sudo mkdir /mnt/na2_test_500g_sata

Note: You need to update your /etc/exports file on the NetApp to allow the vMA to mount the volume

sudo mount prd-na2:/vol/esx_test_500g_sata /mnt/na2_test_500g_sata

Now we need to copy the NetApp mbralign/mbrscan utility onto the vMA, I used a SFTP client to copy it onto my vMA appliance and then moved it to the /usr/bin directory

I won’t go into too much detail on mbrscan/mbralign as it’s use has been well documented, but the basics are:

To scan a VM for alignment:

sudo mbrscan /mnt/na2_test_500g_sata/delete_me/delete_me-flat.vmdk

And to fix alignment if necessary:

sudo mbralign /mnt/na2_test_500g_sata/delete_me/delete_me.vmdk

Note that once this process is done your VM will now be aligned, but it also left behind the original VMDK as a backup, once you confirmed the alignment was successful you will want to go back into the VM directory and remove the two -backup files.

Hopefully that can help for shops that are most Windows-based, I know NetApp (and I’m sure others) are working on tools to make this process even easier – and Windows Vista/7 and Server 2008 do not suffer from this misalignment issue by default.

Popularity: 20% [?]

Port Channels, ESXi 4.1 and the Management Network

Now that I have a few of our ESX hosts upgraded from ESX 4 to ESXi 4.1 I’ve noticed something about the Load Balancing policy on vSwitch0.  Like a lot of people we have multiple GbE connections that go into vSwitch0 (at least on these specific hosts), as usual when you first install ESXi it will set the Load Balancing policy to be based on the Virtual Port ID.

Next I configured the other two adapters that are on this system into vSwitch0
Now I can see in the properties of the vSwitch that vmnic0, vmnic1 and vmnic2 are all listed as active adapters.
But when I go to Management Network it only shows two of the adapters as active and the third adapter is listed as unused.
In ESX I would set the vSwitch to use a Load Balancing policy of IP Hash to work with our port channel we have configured on our switches and it would also change the Load Balancing policy on the included portgroups.
In ESXi it changes the Load Balancing policy for the Virtual Machine network to IP Hash but it does not change the Load Balancing policy for the Management Network, this stays configured as based on Virtual Port ID.
VM Network shows as IP Hash and has all 3 adapters configured as active same as the vSwitch configuration
However the Management Network still shows Virtual Port ID with 2 NICs active and 1 unused

Popularity: 17% [?]

NetApp Virtual Storage Console (VSC) for vSphere

NetApp released v1.0 of their VSC for VMware vSphere, VSC is a plugin for vCenter which allows anyone connecting with the VI Client to access it.  What is nice about the tool is it allows you to see an overview of your storage environment and ensure things are configured inline with NetApp’s recommended best practices and, if they aren’t, allows you to set the recommended values.

4072171914_c27283b945[1]

Click on the pictures for a larger view, you can see it adds a new tab called NetApp when the ESX server is selected.  In my environment we are using all NFS for storage to the ESX boxes, so in the next screen I show the ‘Storage Details – NAS’ option:

4071409815_03c2070da6[1]

What is nice about this view is I can quickly see I need to make a couple changes on the NetApp, the ucode settings should be set to on and the atime update should be set to on as well.

As mentioned before, there are also ESX specific settings that should be set to follow best practices, by right clicking on the ESX hosts listed in the NetApp tab (see 1st picture) there is an option to ‘Set Recommended Values’

4071409861_96e302a9c5_o[1]

Finally you can also set guest timeout settings, under the Tools option there is a download link for the .iso for Windows and Linux guests.  For some reason in my environment I couldn’t download the ISO from this page – if you have the same issue you can find it in the Program Files\NetApp\Virtual Storage Console\webapps\public folder.

Once you load the ISO in the VM, double clicking on the CD drive will ask you if you want to add the info in the windows_gos_timeout.reg file to the registry.

While I only have NFS mounted storage, you can also use the VSC to check things like:

  • Storage adapter timeout settings
  • Multipathing settings
  • Collecting diag info from fibre channel switches (or ESX hosts/NetApp controllers)

Also remember to check NetApp TR-3749 for best practices with NetApp and VMware vSphere.

Popularity: 10% [?]