VMware Cloud Community
s0ldad0
Contributor
Contributor
Jump to solution

4 x ESX hosts to acces 1 x LUN - datastore issues

What is the best way please?

I have 4 x ESX hosts.

2 will be apps cluster.

2 will be web server cluster

I have 1 x DELL nx1950 cluster with an MD3000 array

All is working now, but I want to know how to setup my 1.75TB partition for the 4 ESX hosts.

Currently when I add a 2nd ESX host to the Datastore it says: "Warning the current disk layout will be destroyed...all file systems and data will be lost permanently."

How do I have multiple hosts access 1 data store?

Should the only have their own datastore?

How does Vmotion then work etc in the cluster.

What size should the VM files be etc...

Do i separate my VM app servers from my VM web server farm?

Many thanks in advance.

Reply
0 Kudos
1 Solution

Accepted Solutions
jamieorth
Expert
Expert
Jump to solution

On your 2nd host you do not want to add storage - you need to rescan your HBA to present the existing LUN if you can't already see it. If you can you can "Refresh" your storage and the VMFS partition should show up.

View solution in original post

Reply
0 Kudos
6 Replies
jamieorth
Expert
Expert
Jump to solution

On your 2nd host you do not want to add storage - you need to rescan your HBA to present the existing LUN if you can't already see it. If you can you can "Refresh" your storage and the VMFS partition should show up.

Reply
0 Kudos
s0ldad0
Contributor
Contributor
Jump to solution

So i do a big giant Datastore for all VMs also?

Reply
0 Kudos
jamieorth
Expert
Expert
Jump to solution

Not necessarily - that is subject of great debate in these forums. I would say it depends on the application of the VM's in question. Are they high IO, web, File, etc? That would depend on how the storage is used / presented. I am not familiar with the Dell storage you have so I cannot specifically comment on it.

dpomeroy
Champion
Champion
Jump to solution

Assuming your 4 ESX servers are all VMotion compatible I would create one 4 host cluster and then you could create separate App and Web resource pools if you want to guarantee each type a certain level of resource availability. This gives you much more flexibility to deal with issues such as a host failure, ESX upgrades and patches, troubleshooting problems, and temporary performance issues.

I'm not a SAN guy so I cant give you step by step instructions on how to set it up, but basically you set up the SAN side so that all the ESX servers can access your LUN. On the ESX servers you just take the first server and create the VMFS volume, then on the rest rescan for new VMFS volumes. If you really want two separate clusters then I would create two separate LUNs out of that 1.75 TB (that is kind of big anyway), one for each cluster.

Don Pomeroy

VMware Communities User Moderator

doctormiru
Enthusiast
Enthusiast
Jump to solution

Hi

To have the other 3 nodes recognizing the datastore just add them to the cluster and perform a rescan on the HBAs.

As we have about 180 VMs running on 8 Clusters i can tell you about my experience. We are lucky today by having the following point kept in Mind.

- each LUN with a max. size of 500GB

- each LUN hosts not more than 20 VMs

- spread heavy disk I/O across multiple LUNs

- do not use small arrays for the LUNs (the better way is to have the VMFS volumes spread across different arrays)

rgds

drmiru

Reply
0 Kudos
Jae_Ellers
Virtuoso
Virtuoso
Jump to solution

I agree with the design suggestion of putting all the ESX hosts in one cluster. This helps your failover odds and spreads the load out on the systems. You can then just use resource pools to allocate & reserver adequate cpu & memory for these systems.

Typically the luns are sized in 300-800 GB increments. For your setup I'd probably split them into 3 or 4 equal sized luns plus one 20-30 GB repository for CD, DVD, and floppy images. Spread the high i/o around among the luns, make sure the luns are balanced with regards to the access paths so not all luns are accessing the san through the same hbas on the hosts and controllers (if active/active).

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

http://blog.mr-vm.com

http://www.vmprofessional.com

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

-=-=-=-=-=-=-=-=-=-=-=-=-=-=- http://blog.mr-vm.com http://www.vmprofessional.com -=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Reply
0 Kudos