Just seems strange. If I have centralized storage and I can migrate live vm’s, why wouldn’t I be able to take the risk to migrate a dead vm guest, as in, just start it on another available host and start it and consider it migrated. If the original vm host happens to come back up and the instance has already migrated, just leave it alone. Seems like that would go a long way for high availability. I assume i/o fencing becomes an issue. I’d prefer some kind of hook so if a VM host seems dead beyond, say, 5 minutes, i/o fence it by triggering a power off from the PDU and bring up the VM guests somewhere else.
-jeremy > On Jun 18, 2021, at 2:42 AM, Daan Hoogland <daan.hoogl...@gmail.com> wrote: > > :D > I'm sorry, when your bike is broken you have to walk. detection if the > exact situation is the issue here, good luck in your search! > > On Fri, Jun 18, 2021 at 11:31 AM Jeremy Hansen <jer...@skidrow.la> wrote: > >> I guess I’m looking for a recovery scenario where the dead vm host is not >> coming back, failed disk, caught on fire, and a reboot isn’t going to help. >> >> Thanks >> >>> On Jun 18, 2021, at 1:41 AM, Daan Hoogland <daan.hoogl...@gmail.com> >> wrote: >>> >>> Jemery, >>> If you don't have IPMI then ACS can not know for sure that the VM won't >>> come back. If it comes back the VM would be running twice and this must >> be >>> prevented at all costs. Maybe I am missing some functionality, and >> someone >>> else can give additional options. >>> >>>> On Fri, Jun 18, 2021 at 10:21 AM Jeremy Hansen <jer...@skidrow.la> >> wrote: >>>> >>>> I pasted that from the documentation. My end goal is if a VM host dies >>>> completely and I’m not available to fix it, I would like the VM guests >> that >>>> were running on the failed host to automatically migrate to an >> available VM >>>> host so the guest instances continue to run. Perhaps that’s not how it >>>> works. The hosts I’m using for testing do not have any kind of IPMI >>>> supported out of band management. They do have network enabled PDUs but >>>> let’s just say the VM host is gone completely. How do I get the VM >> guests >>>> that were running on the failed host back up and running without my >>>> intervention? I guess I wrongly assumed Cloudstack would handle this >> case >>>> by just starting the VMs on another available host machine after some >> kind >>>> of failed heartbeat threshold. >>>> >>>> Thanks >>>> -jeremy >>>> >>>>> On Jun 18, 2021, at 1:09 AM, Daan Hoogland <daan.hoogl...@gmail.com> >>>> wrote: >>>>> >>>>> Jeremy, >>>>> I don't fully understand your question. You say you are interested in >>>>> HostHA specifically but then you ask about restarting VMs when a host >>>> dies. >>>>> This would not be safe as we can't be sure a host really dies unless >> you >>>>> have HostHA enabled. Consequently you can't guarantee the VM won't >>>> suddenly >>>>> re-apear when the host is seen running again. So keep these things >>>>> separated. >>>>> HostHA is for rebooting suspect hosts, not for moving VMs around. I am >>>> not >>>>> aware of the connection between the two, that you seem to look for. >>>>> >>>>> >>>>> On Fri, Jun 11, 2021 at 11:03 AM Jeremy Hansen <jer...@skidrow.la> >>>> wrote: >>>>> >>>>>> >>>>>> I’m trying to play with HA. I’ve enabled it via the interface but the >>>> HA >>>>>> state is labeled as Ineligible. >>>>>> >>>>>> I’m specifically interested in this: >>>>>> >>>>>> HA for Hosts >>>>>> >>>>>> The user can specify a virtual machine as HA-enabled. By default, all >>>>>> virtual router VMs and Elastic Load Balancing VMs are automatically >>>>>> configured as HA-enabled. When an HA-enabled VM crashes, CloudStack >>>> detects >>>>>> the crash and restarts the VM automatically within the same >> Availability >>>>>> Zone. HA is never performed across different Availability Zones. >>>> CloudStack >>>>>> has a conservative policy towards restarting VMs and ensures that >> there >>>>>> will never be two instances of the same VM running at the same time. >> The >>>>>> Management Server attempts to start the VM on another Host in the same >>>>>> cluster. >>>>>> >>>>>> >>>>>> My assumption is if a VM Host dies, whatever guests that were running >> on >>>>>> that host would automatically move to an available VM host. Maybe I’m >>>>>> misinterpreting. >>>>>> >>>>>> Thanks >>>>>> -jeremy >>>>>> >>>>> >>>>> >>>>> -- >>>>> Daan >>>> >>>> >>> >>> -- >>> Daan >> >> > > -- > Daan
signature.asc
Description: Message signed with OpenPGP
