Comments on: High Availability change http://www.yellow-bricks.com/2008/07/29/high-availability-change/ Building blocks for virtualization... Thu, 18 Mar 2010 08:03:51 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: Alan http://www.yellow-bricks.com/2008/07/29/high-availability-change/comment-page-1/#comment-787 Alan Fri, 01 Aug 2008 12:05:07 +0000 http://www.yellow-bricks.com/?p=255#comment-787 I think it all depends on how confident you are with your environment some may have bad power, some may have bad network equipment. At least we have a choice to make now. I think it all depends on how confident you are with your environment some may have bad power, some may have bad network equipment. At least we have a choice to make now.

]]> By: Magnus http://www.yellow-bricks.com/2008/07/29/high-availability-change/comment-page-1/#comment-776 Magnus Thu, 31 Jul 2008 09:33:16 +0000 http://www.yellow-bricks.com/?p=255#comment-776 I think this is a good change because if you are in an environment where there is a risk that a junior network guy mess something up, you can end up with spanning tree problems for a few minutes. In that case you do not want _all_ ESX hosts to think they are isolated and _all_ VMs shut down. It's better that they are left powered on till the network is stable again. I think it outweights the other option where a single ESX actually becomes network isolated but keeps looks on the files so no other ESX can take over the VMs on that specific ESX. The probability for this happening is quite low (that all NICs fail), I rate the spanning tree problems more likely since I've seen it twice, not fun with 200 VMs :) On a real ESX or server crash, the file locks would be released as normal and other ESXes can take over. I think this is a good change because if you are in an environment where there is a risk that a junior network guy mess something up, you can end up with spanning tree problems for a few minutes. In that case you do not want _all_ ESX hosts to think they are isolated and _all_ VMs shut down. It’s better that they are left powered on till the network is stable again.

I think it outweights the other option where a single ESX actually becomes network isolated but keeps looks on the files so no other ESX can take over the VMs on that specific ESX. The probability for this happening is quite low (that all NICs fail), I rate the spanning tree problems more likely since I’ve seen it twice, not fun with 200 VMs :)

On a real ESX or server crash, the file locks would be released as normal and other ESXes can take over.

]]>