Today I will write about a great subject which means a lot to me.
I read so often that people will use NFS as a cheap and easy shared storage for VMware ESX.
Conceptional! If you want to test something in a test environment, you need cheap storage or what ever you can use NFS. But why so much pepole will use NFS on Linux? There is not one benefit against NFS on ZFS with Solaris 10/Nevada. So lets use a great protocol from the same vendor who develop a very great filesystem and operating system. Lets checkout NFS and ZFS with Solaris...

VMware ESX, Solaris 10/Nevada with ZFS. The perfect NFS...

First, I will show you the benefits of NFS:

• You can use NFS fore free
• You dont need new hardware or something like FC-Switches
• Easy administration
• Use your infrastructure
• ...
  

Now I will show you some benefits of ZFS:

• Filesystem and Volumemanger in one system
• Easy administration with only 2 commands - #zpool and #zfs
• advanced raid level and functions
• snapshots
• automatic checksum over all data
• 128 Bit
• automatic shrinking and growing volumes
• ...

So, I think we are ready to start with this great couple.

The first you need to do is to download the free Solaris OS from Sun. You can download Solaris 10 at www.sun.com. Its also possible to use the upcoming Solaris in a Developer Edition. Solaris Nevada is available at www.opensolaris.org but there is no possibility to buy support at the moment.

Dont be scared ahead the installation of Solaris! A lot of pepole wirte me that they don´t want to install Solaris but it is so easy with the new versions like Windows or Linux. Just test it! You need only a fews clicks to complete the installation.

After you installed Solaris we can create a ZFS Pool.

In the Bash (right click the desktop and then terminal) you can initiate all hdd´s with "#disk" and show all disks with "#format". You need a minimum of 2 hdd´s in the system because you can´t install ZFS volumes on the system disk.

We can create now a pool with for example 4 hdd´s. Just set up this command:

"#zpool create zfspool raidz c0d1 c1d1 c2t0d0 c2t1d0"

What happend now? We create a ZFS Pool with the 4 HDD´s c0d1 c1d1 c2t0d0 c2t1d0. The ZFS Pool is mounted to /zfspool. The pool is directly a RAIDZ. This is a sepcail RAID level for ZFS. It´s like Raid 5 with better functions because there isn´t a write break between „date-write“ and „parity-write! That´s it! Easy going or what do you think?
With the command "#zpool" list it is possible to see your ZFS Pool.

In this ZFS Pool it is possible to create a lot of ZFS Volumes. This is a great option. Because all volumes shrink and grow automatic you must just watch the size of the pool. The volumes mange themself.

"#zfs create zfspool/VMware_ESX_NFS"

With this command we create the new Volume in the pool.



Now you have a full functional ZFS Volume with RAIDZ and you only use 2 commands. I think this is very great!
In this example you see the automatic growing and shrinkingoption. I copy a file to the filesystem and delte it again and the Volume has the same size as before.

Now we are ready to share the ZFS Volume over NFS to our ESX Server.

Firt we need to setup the vmkernel at our ESX Server. Over the vmkernel the ESX connect to the NFS Server.
Connect with the VI Client to the ESX Server or Virtucal Center and configure this option.

After you have done this, we can configure the ZFS Volume that it is possible that the ESX Server can connect to ths NFS volume.

#share –F nfs –o rw,root=192.168.36.204 /zfspool/esx_nfs_datastore

Now the ESX Host can access the NFS datastore. If you have more then one ESX host you can add more IP adresses to this command.

This is just a short view for the options with ZFS and NFS. I hope you read this article and test this great couple. Solaris, NFS and ZFSis a greap option for a cheap, fast and save storage system on a ESX Server.

Feedback is welcome.

Wish you all guys a nice day,

Thomas