Comments on: vSphere HA design consideration https://magander.se/vsphere-ha-design-consideration/ Wed, 01 Feb 2017 08:54:55 +0000 hourly 1 https://wordpress.org/?v=6.9.4 By: magander3 https://magander.se/vsphere-ha-design-consideration/#comment-95 Wed, 01 May 2013 07:40:11 +0000 http://magander.se/?p=957#comment-95 In reply to Bozo Popovic.

Hi Bozo,
it is a valid question, thanks for bringing it up.The cases i have seen so far includes failure of a dual port CNA, administration mistake making one ESXi host losing CNA connectivity (for both 2xSingle port CNA and 1xDual port CNA) so i agree this will probably not happen that often.
No matter if it will happen often or not i think you should make the customer aware of what can happen when certain decisions are made.

Thanks
Magnus

]]> By: Bozo Popovic https://magander.se/vsphere-ha-design-consideration/#comment-94 Tue, 30 Apr 2013 18:11:52 +0000 http://magander.se/?p=957#comment-94 Hi,

i did not get your design right i assume? I now understand what are you trying to tell, you are using FCoE for SAN traffic, which if right will make the virtual machines fail after CNA failure, but network connectivity as it is other adapters will not and thus hosts will have network connectivity and VM’s will not re-start.

I know this may be an overkill question but does it make real life sense to assume both on-board (or mezzanine) FCoE controllers (i assume each server has them = no SPoF) fail and nothing bad happens to the rest of the motherboard = 2 NIC uplinks in order for this scenario to happen? Also what about designing a solution with VM-Monitoring for HA events?

If one can predict such cases are valid than hats off to your work.

Keep em coming!
Thanks,
Bozo

]]> By: magander3 https://magander.se/vsphere-ha-design-consideration/#comment-93 Fri, 26 Apr 2013 20:40:22 +0000 http://magander.se/?p=957#comment-93 Hi, the blog post covers the single ESXi host perspective. If the entire, in my case Nexus 5K, infrastructure goes down no ESXi host will be able to run any virtual machine and vSphere HA does not provide any additional availability.
If one ESXi host loses FCoE connectivity using the CNA and the vSphere HA heartbeat runs over the CNA, which is my suggestion, the vSphere HA isolation response will be triggered. If the vSphere HA isolation response is configured to “Shut down” the affected virtual machines running on the ESXi host losing CNA connectivity will be started on another ESXi host. If storage connectivity goes down the ESXi host will not trigger APD but losing the vSphere HA heartbeat will trigger the isolation response.

Thanks.

//Magnus

]]> By: popovicbozo https://magander.se/vsphere-ha-design-consideration/#comment-92 Fri, 26 Apr 2013 16:37:39 +0000 http://magander.se/?p=957#comment-92 Hi, thanks for sharing. I am trying to understand why would HA be helpful in a case where FCoE traffic is disrupted. I am assuming you don’t have any single point of failure, that is, assumption is that FCoE traffic has no single point of failure.
But in case all your FCoE paths go down to particular storage, then your design would not provide any benefits with HA as it is assumed that APD of FCoE on single host affects all the hosts. This can also happen in case where we have FC traffic and that is a classic APD and hard-shut of all vm’s as a result.

]]>