stoutat
Contributor
Contributor

How should I go about doing HA/Clustering?

I have 2 hosts on ESX 3.5. My plan is to bring 2 web servers into production, one on each host. I need to make these as fault-tolerant as possible. I was thinking of using HA/DRS at the ESX level, but should I use NLB or MSCS at the guest OS level? These server won't receive a ton of traffic, probably a couple of thousand hits a day is all, but we're running on an old hunk of junk right now that has issues with IIS services simply stopping and not restarting. What happens on one of these super duper highly available web servers if services are unavailable? Should I consider some type of DNS failover?

I'll give thanks in advance - I need some guidance here.

0 Kudos
2 Replies
rDale
Enthusiast
Enthusiast

HA will not give you failover unless the actual host fails but is a good thing.

NLB will not give you service level failover but will allow you to load balance resulting in both servers being loaded.

MSCS will give you service level failover and if active active cluster then also a degree of load balance.

MSCS will also mean that due to the current standards for virtual cluster you will not be able to vmotion your vm onto another host due to local disk storage.

If you can live with the lack of failover when a Operating system service fail, then go with HA/NLB. You can always setup event trigger to fire when IIS fails that reboot the machine or restart the service.

stoutat
Contributor
Contributor

Sounds like HA/NLB is the way to go then. I just read about the MSCS limitation with local storage, which is not what I want to do with VMWare. Plus, I don't really want to buy an enterprise license, whereas NLB comes with standard. Thanks for the info - anyone elses input would be great!

0 Kudos