Which Features to Enable. -Real world practical advice.

[Kumatech]

Hello,

We have a client with a VMware I3 environment with all the bells and whistles... There are 3 nodes. We have the option of using VMotion, DRS, HA etc. but should we?

If you use these HA options? what are the recommended configuration settings (REAL WORLD) not out of the VMware text book...

The core of what I am looking for here is, real world practical advise on whether we should enable HA and if so, what is the typical resource or trigger values we should consider setting?

I appreciate the help, thanks!!!

[Troy_Clavell]

My mindset is if you paid for the features, use them. Enable HA, DRS and vmotion. It will help you with load balancing your cluster and enable VM's to be restarted on other hosts in the cluster if you have a failure of another ESX Host. Remember all these features are dependent on shared storage.

For our environment we have HA enabled with all of the defaults, which includes "leave vm's powered on" in case of an isolation response. We do not use VMM. For DRS we have it set to fully automated with 3 star recommendations.

hope this helps.

[AntonVZhbankov]

I have all the features also, HA, DRS and VMotion.

There was no possibility to see if HA worths its money. Fortunately.

I use VMotion almost every day and DRS in semi-automatic mode.

---

VMware vExpert '2009

http://blog.vadmin.ru

[Rob_Bohmann1]

Hello,

We have a client with a VMware I3 environment with all the bells and whistles... There are 3 nodes. We have the option of using VMotion, DRS, HA etc. but should we?

DRS/VMotion allows you to do maintenance activities without affecting the availability of the business applications. So I think you should use that feature. It also allows you to have a "Virtual Admin" - the DRS algorithm, watching the environment and redistributing the load 24 x 7 so again I think you should use that feature. You can use DRS affinity rules to keep together dependent servers to maximize performance and anti-affinity rules with HA to separate machines to maintain a higher likelihood of availability if someone pulls the wrong cable or cord, if there is a hardware failure or a server crash, etc. HA also minimizes downtime and allows the "Virtual Admin" to restart your servers @ 2 am, preventing you from getting that call possibly.

As someone already noted, your client paid for them, they should use their benefits. Its like paying for insurance and coming out to see a tree has fallen on your car/house and saying I do not know if I should call my insurance agent -what are they going to say...

As far as real world settings, every environment is different. Every software vendor makes defaults that will work well for "most" environments. Start there and do a little experimentation to understand how HA and DRS behave. For DRS the sensitivity defaults at level 3 - right in the middle of the road. This works well for a lot of environments and results in relatively few vmotion activities until the load on a host becomes significant. I started with manual, moved to semi-auto, and then fully-auto to understand how it behaves. Some clusters worked best at level 3, some at 4 and some at 2. If you are not at auto level, then maintenance mode will not move VM's off a host when you want to patch a host for example. Over time you may determine that you need to adjust these as the environment changes.

To clarify your last sentence, HA will only kick in when a host becomes "isolated" - loses its mgmt-heartbeat network connection with the other hosts, or crashes/power failure etc., it does not trigger at a specific resource usage level - that is DRS. Unless you are beyond your capacity (Not N+1 from a ESX host perspective), then HA is something you should consider.

Finally, how I approached this was I started with DRS and played with that for a while before using HA. If you want to use HA, it is important to have redundant network connections so you do not have a false failover event. There are options/workarounds (das.isolation.address/leave machines running) to prevent this scenario from happening. Build your cluster, build a few test machines and play with all the settings to see how they work. Understand the business needs and performance characteristics of the machines that will be running on the hosts.

I leave you with this- defaults are there for a reason, its the place you start from, not necessarily finish.

[cainics]

You should use the VMotion, DRS and HA features. With all the bells and whistles also configure Virtual center if the client has the licenses to do so.

The real world setttings for DRS is fully automated. For HA it entirely depends on how far you want to go. Turn on HA for sure and set the isolation response. If you want the virtual machines to turn off during any isolation of a esx host then set the function to off. As a safety net you can leave the virtual machines to remain on.

DRS and HA will ensure that your client sees the benefits of virtualization and commends you for the good work. In reality, it is actually an appreciation of VMWare but when clients see it initially they like you implementing the feature and helping them out with the unknown. VMotioning of servers will work more regularly with DRS. VMotion comes real handy when you want to perform maintenance on a ESX host and then you enter maintenance mode on it. VMotion will automatically move the virtual machines running on that ESX host to the others that are running in normal mode.

DRS triggering can be set to aggressive, medium or low. Usually, set it to medium so that there is no VMotioning required all the time but it also depends on how many virtual servers you run per esx host. When you create a VM cluster also create resource pools - production and test/dev. Set the resource shares to high for production, normal or low for test/dev. Depending on how critical your production pool is you can further set CPU and memory reservation for the pools as well.

So use the merits of these features and get some kudos too. VI3 is a boon for system administrators and IT infrastructure consultants. It's something that we always craved for our windows servers and now you have a chance to use it.

Hope this helps. I don't know how far you want the answers to go so i have attempted to answer as briefly as possible with the required details.

-Anil

"Dazzle them with your intelligence or baffle them with your bullshit"

We have a client with a VMware I3 environment with all the bells and

whistles... There are 3 nodes. We have the option of using VMotion,

DRS, HA etc. but should we?

If you use these HA options? what are the recommended configuration settings (REAL WORLD) not out of the VMware text book...

The core of what I am looking for here is, real world practical advise

on whether we should enable HA and if so, what is the typical resource

or trigger values we should consider setting?

[Texiwill]

Hello,

If you have shared storage for your VMs then use them all. I have had HA protect me when my UPS recently failed. Vmotion I use daily. DRS alleviates contention on my hosts.

You paid for them, enable them. Plus you get EVC, DPM, and SVMotion on top of all the others.

Best regards,
Edward L. Haletky
VMware Communities User Moderator, VMware vExpert 2009
====
Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.
Blue Gears and SearchVMware Pro Blogs -- Top Virtualization Security Links -- Virtualization Security Round Table Podcast