ESX 3.5 connected to SAN - what configuration required

[nkv]

Hi Everyone,

We have a 10 host and want to devide this 10 host into 2 cluster (Clutser A and Cluster B) - with each cluster having 5 hosts. We want these clusters to be connected to LUN on SAN. Each cluster will be connected to different LUN but they will be going through a common fabric switch and the LUNs will be on a common storage. Need help in understanding some of the questions I need to check with SAN team.

Question 1: What configuration will be required on the fabric switch so that host of one cluster dont see LUN of other cluster (zoning i believe is required)

Question 2: What configuration will be required on the Storage side so that host of one cluster dont see LUN of other cluster (masking i believe is required)

Question 3: What configuration is required on LUN so that all the hosts in a cluster can see it and LUN works fine for vmotion

Question 4: I believe WWN number of HBA card is required. I did not understand the relation why WWN number is required for LUN. Please help in understanding where WWN number will be required and where the configuration will be performed.

Question 5: What is the simple way we can understand a active/active array and active/passive aray

Any info will be much appriciated

Thanks

nkv

[PaulSvirin]

Regarding 5) - do you need an explanation of "A-A", "A-P"? Is this the one you were looking for: ?

Other questions - you can read about LUN masking here:

---

Paul Svirin

StarWind Software developer ( http://www.starwindsoftware.com )

[Erik_Zandboer]

Hi,

Question1 + 2:

zoning is always required (best practice at least). You would zone each ESX node to each storage processor. This in itself does not separate the stores both ESX nodes see. That is done on the storage itself: You will have to create two host-groups (cluster A and cluster B), and then use masking to show all LUNs for cluster-A to the host group named cluster-A and the LUNs for the cluster B to the host-group named Cluster B.

Question 3:

LUNs require no specific config. They must be carved out of RAID sets. Make sure you get enough spindles under the LUNs for size and performance. One requirement: LUN size should never exceed 2TB.

Question 4:

When zoning on the FC switches, you need the WWNs of the HBAs. That is what zoning normally acts on. WWNs are like MAC addressen in the IP world. If you have ESX already running, the HBA drivers should be active, and your FC switches should already be seeing the WWNs. Normally you would create logical names for the HBAs, then assign the proper WWNs to them so you can zone using logical names instead of WWNs.

Question 5:

An array is active/passive or active/active. Some arrays claims to be active/active while they're not (really). The essence is, that in smaller arrays each LUN is assigned to a single storage processor. You need to configure ESX path failover to "most recently used", so that all ESX nodes will access that LUN via the preferred controller. If you do not (and use a fixed policy), one ESX node will try to access the LUN via storage processor A, while another ESX node will attempt this via storage processor B. The result is called "path trashing", where the SAN will move LUN constantly between the Storage processors (SPs), effectively killing all performance.

On the other hand there are true active/active arrays. These arrays are able to truly deliver all LUNs via all SPs simultaniously. In that case you can set path policy of ESX to FIXED, because it does not matter over which SP each ESX node accesses a LUN.

Visit my blog at http://www.vmdamentals.com