Comments on: How to avoid HA slot sizing issues with reservations? http://www.yellow-bricks.com/2009/11/06/how-to-avoid-ha-slot-sizing-issues-with-reservations/ Building blocks for virtualization... Thu, 18 Mar 2010 08:03:51 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: Cody http://www.yellow-bricks.com/2009/11/06/how-to-avoid-ha-slot-sizing-issues-with-reservations/comment-page-1/#comment-5025 Cody Sun, 08 Nov 2009 17:04:47 +0000 http://www.yellow-bricks.com/?p=4473#comment-5025 Good stuff. Ken, From the look of it, it seems the hosts resources get calculated into the "total amount of available resources" portion. The admission control mechanism would then use this in it's decision making. Example: Host 1 20Ghz 20GB Host 2 10Ghz 10GB Before host failure: ((30Ghz-(2Gz+1Ghz+256Mhz+4Ghz))/30Ghz) = 75 % available ((30-(1.1+.114+.626+3.2))/30) = ~83 % available Failing the 20Ghz node: ((10Ghz-(2Gz+1Ghz+256Mhz+4Ghz))/10Ghz) = 27 % available ((10-(1.1+.114+.626+3.2))/10) = ~49 % available As long as you don't violate the % and/or the host failures, I believe you should be OK. That is unless I've completely misread the above. Good stuff.

Ken,

From the look of it, it seems the hosts resources get calculated into the “total amount of available resources” portion. The admission control mechanism would then use this in it’s decision making.

Example:
Host 1
20Ghz
20GB

Host 2
10Ghz
10GB

Before host failure:
((30Ghz-(2Gz+1Ghz+256Mhz+4Ghz))/30Ghz) = 75 % available
((30-(1.1+.114+.626+3.2))/30) = ~83 % available

Failing the 20Ghz node:
((10Ghz-(2Gz+1Ghz+256Mhz+4Ghz))/10Ghz) = 27 % available
((10-(1.1+.114+.626+3.2))/10) = ~49 % available

As long as you don’t violate the % and/or the host failures, I believe you should be OK. That is unless I’ve completely misread the above.

]]> By: Ken Cline http://www.yellow-bricks.com/2009/11/06/how-to-avoid-ha-slot-sizing-issues-with-reservations/comment-page-1/#comment-5020 Ken Cline Sat, 07 Nov 2009 23:15:17 +0000 http://www.yellow-bricks.com/?p=4473#comment-5020 So...how does vSphere with percentages handle non-identical hosts in a cluster? Pre-vSphere, if you had a cluster with four hosts where three of the hosts had 10GHz of CPU and the fourth had 20GHz of CPU and you configured the cluster for 1 host failure, it calculated the reserve capacity based on the assumption that the largest host would fail (i.e. it would reserve 20GHz of capacity). How would this scenario play out with vSphere percentages? So…how does vSphere with percentages handle non-identical hosts in a cluster? Pre-vSphere, if you had a cluster with four hosts where three of the hosts had 10GHz of CPU and the fourth had 20GHz of CPU and you configured the cluster for 1 host failure, it calculated the reserve capacity based on the assumption that the largest host would fail (i.e. it would reserve 20GHz of capacity). How would this scenario play out with vSphere percentages?

]]>