Right now my 2 VMware ESXi have local storage and i`m planing to move it all to iSCSI and will be nice to boot the ESXi from the iSCSI... but i`m not sure yet because a FlashDrive is more easy to manage on a little home-lab....
You know, it seems like ESX/ESXi just continue to get more and more stateless all the time and the product is quickly moving toward being truly and fully decoupled from the hardware itself. Consider the vSphere Enterprise Plus distributed virtual switch as another step in this direction. No more per-host networ configuration is necessarily needed. Of course, it's not fully decoupled, but is getting close.
Lose a host due to hardware failure? No problem... just pull the blade and replace it.
Posts: 36
Joined: 29.Jun.2010
From: VA
Status: offline
Host profiles is another great example of what Scott's saying. Rebuiling an ESX/ESXi host has always been relatively fast and painless process, but attaching a host profile to a cluster makes the re-configuration process even more pain-free and less error prone.
I also like the move to the ESXi format and getting away from the bulky COS, even though Tech Support Mode is now officially supported, sorta, so it's not like you are going completely COS free. I wasn't sold right away, but it's won me over. With the vast majority of the configuration being pushed to the ESX/ESXi box, significantly less touch-time is being required when new hosts are brought into the datacenter.
If you're looking for a stateless set-it-and-forget-it computing environment, take a look at the joint venture between EMC, Cisco and VMware called the Virtual Computing Environment (VCE) coalition. One item that's come out of this coalition is the VBlock, which is a strictly configured hardware environment that includes gear from all three companies. I attended a VBlock session a Gestalt IT Tech Field Day event in May and was blown away. The Cisco servers that the coalition is using are truly stateless and can be swapped out with ease. The entire solution - as long as it meets VBlock standards - is supported via one telephone number regardless of which component - storage hardware, hypervisor, networking hardware, servers - fails.
It's an incredibly cool concept and takes abstraction to a whole new level.