It's been a while since I've updated the blog but things have been really busy at home and at work so I apologize for not getting this out sooner.
Design Workshop
I recently attended VMware's Design Workshop for vSphere 4.1 and I have to say I was surprised at how open the course was. I really expected more structured learning but it was very interactive and I really enjoyed the course. I was also surprised at how many different engineers had so many different interpretations on the design of their environments (we did scenario based designs with mock requirements).
Most everyone did a UCS (Cisco) architecture while I (more knowledgeable of Dell) went with Dell blades due to their density and back plane capabilities with the M1000e chassis. However, as I read more about UCS and it's capabilities with the fabric connect switching (to include future capabilities like FCoE) I am thinking that UCS is really being designed with virtualization in mind (from the start).
All in all, the class was great and I have some great reference materials now on the design aspect of vSphere and I can leverage that and my data center administration exam studying and experience to develop a good platform (from start to finish) on future deliveries.
VMworld 2011
So I am lucky enough to have my employer fund my trip to VMworld 2011 in Las Vegas, NV this year and I am really excited to be able to attend. I've had the privilege to meet with VMware to discuss road maps on future virtualization/cloud computing deliveries and let's just say that there will be a lot of announcements on a lot of product lines at VMworld this year. If you haven't made plans, I would high recommend that you try to attend this year. It's not cheap but well worth the investment. More information can be found here...
http://www.vmworld.com
Understanding HA and DRS
Now onto some technical discussion about HA and DRS. When you define this setting in your cluster, you assume some default settings that can be modified that pertain to certain things like restart priority, aggressiveness of your automation settings (priority levels determine level of automation), and recommendation settings.
DRS is relatively easy as you set how conservative or aggressive you want the resource utilization to be and what priority settings you want to set on VMs. Higher priority VMs may have a higher migration threshold setting so in the event of resource utilization increasing on a certain host, you don't run into performance issues on the individual VMs. Distributed Power Management can be incorporated here as well if you wish to save power on underutilized hosts. This coupled with resource pool utilization, you should have well balanced hosts throughout your cluster(s).
HA is tricky and I have a funny example that shows how some people really need to read on how HA works before assuming it's a truly automated solution. As with anything you have to assume that there are points of failure that will force certain redundancies to fail. In this example, I'm talking power... Never assume that if your data center loses power that your VMs will automatically restart just because your hosts restart... With restart priorities defined in your HA cluster, you have to have a host online to facilitate those requests. Additionally, that host has to detect host failures and respond by restarting VM's on available hosts. HA elects up to 5 primary nodes per cluster (in the order they were added) and if all hosts fail due to a power event, no primary node is available to facilitate VM restart priorities. Therefore, when power is restored and hosts restarted, they are simply restarted and the VMs will have to manually be powered on. This is assumed that vCenter is virtualized as well since HA events are facilitated and coordinated via vCenter. There is a great read on this on YellowBricks and I highly recommend you read this and check out Duncan Epping and Frank Denneman's book on the HA and DRS Technical Deepdive
Book - http://www.amazon.com/gp/product/1456301446?ie=UTF8&tag=yellowbricks-20&linkCode=as2&camp=1789&creative=9325&creativeASIN=1456301446
HA Deep Dive - http://www.yellow-bricks.com/vmware-high-availability-deepdiv/
DRS Deep Dive - http://www.yellow-bricks.com/2009/10/21/drs-deepdive/
No comments:
Post a Comment