I've powered off a host of a KVM cluster to simulate the server failure and I'm experiencing the "Agent state cannot be determined, do nothing" loop.
How I can tell the manager that the host is died and to start the HA procedure? There is some field on the db I can update? 2013/7/29 Kirk Jantzer <kirk.jant...@gmail.com> > Great information, thanks so much for sharing!! > > > On Mon, Jul 29, 2013 at 4:28 AM, Ryan Lei <ryan...@cht.com.tw> wrote: > > > FYR, I have done some similar tests several weeks ago to test the host HA > > functionality. > > > > CS 4.0.2 + XenServer 6.0.2: Works as expected. HA-enabled VMs (including > > System VMs) were automatically restarted on HA-dedicated hosts. > > CS 4.1.0 + XCP 1.6: No HA thing happened at all much like what you > > described. The states of VMs and hosts were just like before, but > > inaccessible. CloudStack just couldn't seem to detect the host down! > > CS 4.1.1 (git) + XenServer 6.1: Works as expected. Just like the first > > case. > > > > I didn't have time to try other combinations, and the logs are gone, but > I > > guess the problem more likely has to to with XCP support than CS 4.1.0. > > > > > > > ------------------------------------------------------------------------------------------- > > Yu-Heng (Ryan) Lei, Associate Reasearcher > > Chunghwa Telecom Laboratories / Cloud Computing Laboratory > > ryan...@cht.com.tw< > > > https://email.cht.com.tw/owa/redir.aspx?C=-wE1FEC3G0SWYpVkiWo8SsDdf3ZqO9AIuAPTzRnFYCUi-z4YljtI_hyVKkNHfn9F1Bn-vUWJnQ4.&URL=mailto%3aryanlei%40cht.com.tw > > > > > or > > ryanlei750...@gmail.com > > > > > > > > On Thu, Jul 25, 2013 at 2:07 AM, Kirk Jantzer <kirk.jant...@gmail.com > > >wrote: > > > > > Can someone "explain like I'm 5" how HA in CS should work? I have 4.1 > > setup > > > with XCP hosts. To simulate a host failure, I've hard powered off a > host > > > through iDRAC and CS doesn't seem to know about it, at all -- the > > instances > > > still show as running, but I cannot connect to them, and the host shows > > as > > > up in the infrastructure section. > > > > > > The instances, including the SSVM, that were on there are "gone" and > have > > > not been restarted on any of the other hosts. > > > > > > -- > > > Regards, > > > > > > Kirk Jantzer > > > http://about.met/kirkjantzer > > > > > > > > > -- > Regards, > > Kirk Jantzer > c: (678) 561-5475 > http://about.met/kirkjantzer >
