C3000,Lefthand and V-Sphere

[Uwe Zessin]

Yes, that is what I meant. No offense meant, but many people seem not to realize that VMware ESX does not create ressources 'out of thin air' and expect they can put 100% load on their servers and HA still working.

It's been some time that I had a discussion with a customer who did this:
- a server with 16GB memory
- a number of VMs which took a good portion of the memory

Then he created another VM, gave it a 12GB(!) memory reservation and started this VM.
Ooops...

[DThomson_1]

You almost gave me a heart attack Uwe, hahahahah. I have ESX clusters running already in the environment but none with only two hosts, i just assumed it would work with two. Good lesson learnt in assuming :)

Thank you for your input so far, its very much appreciated.

[Steven Clementi]

I think the moral of thre story is simply that you need to take into consideration the fact that in order to "failover" to one or the other servers... you need to have the appropiate resources available.

In a 2 node cluster... you need to double up, if not more, on your resources in order to survive a single node failure.

For clusters with more nodes, then the question you need to ask yourself is how many nodes do we want to be able to lose in order to survive...

To answer the original question...

"My concern: Will HA/DRS and VMotion work if one of the servers fail ?"

It depends on which server fails. If the node with the SB40 attached fails.... you are simply SoL or DitW (Dead in the Water). Best option is getting a second SB40 (which may have been mentioned, but I refuse to look back right now).

In a properly configured 2 node cluster.. yes, sure.. DRS/HA/vMotion will/should work just fine.


Steven