vSphere Design Workshop Day 2 Takeaway

Quote of the day: “The 3 most common lies: I Love you, The check is in the mail, and that will be fixed in the next revision”

Today started out with us talking about HA Slot Size, by default this is 256Mhz and 256MB. However it can be set manually by editing some advanced options. More important is to remember that whenever you make a reservation that size becomes the slot size (whatever the largest cpu or ram is the new slot size)

Why is this important? Because if your slot size is 256MB of ram, and you have 100VM’s over 4 hosts you will have an HA reservation of almost 6.5GB that cannot be used and is reserved for HA. Same goes for Mhz

Another thing to remember about HA is that there is (by default) 5 primary hosts that control HA. If working with blades you MUST make sure that those 5 primary blades are not in the same chassis… if they are and you encounter an enclosure problem HA will break and NO VM’s will be restarted.

Resource pools are only used when a resource is in contention… otherwise they are ignored.
Resource pools are mostly miss understood and certain VMware guru’s recommend NOT to use them at all. This is because if you have say three pools – High with 2000 shares, Medium with 1000 shares, and Low with 500 shares and you have 16GB in contention. And lets assume you have 10 VM’s inside of the High pool, 25 in the medium pool, and 100 in the low pool you would then have this much memory per pool:

Pool Shares % of ram Ram for pool VM’s Ram per VM
High 2000 57% 9.12GB 10 912MB
Medium 1000 28.5% 4.56GB 25 182.4MB
Low 500 14.5% 2.32GB 100 23.2MB

As you can see this REALLY short sides your VM’s therefore it would be better to set reservations on a per VM basis that way you know the critical VM’s are going to get what they need.

DPM – a VMware Technical marketing guy mentioned that while right now DPM is almost scary because it can only power down hosts, and then it will cause cooling issues and other ill effects in the datacenter, to keep you eye on DPM because they have some really neat things on the timeline…. Stuff like being able to control A/C units from the cluster and even an entire zone of a datacenter so that when all of those hosts are powered off the A/C and switches and PDU’s can also be shutdown so that REAL power savings is realized.

Also with DPM, strick admissions control MUST be enabled if DPM is enabled otherwise DPM will violate HA and could power down all hosts but one.

Exchange 2010 has the ability to move archive data to a different datastore … enabling an administrator to put that datastore on SATA … or low end storage and no take up 15k sas drive space.

More to come tomorrow!

Share This Post

Leave a Reply