When the XCPNG host restart, it restarts running directly, instead of being in maintenance mode

psafont

@olivierlambert XCP-ng will include an option to keep hosts disabled after reboot. This will be part of the next update for xapi. (25.33.1) This was contributed upstream by Andriy and is used by adding an extra parameter to disable: xe host-disable uuid=$HOST auto-enable=false.

olivierlambert

Ah nice!

oliv77

Hi,

@olivierlambert, @psafont @pilow Perfect, thank you a lot !!

olivierlambert

@psafont do you know if it's already available since then in XCP-ng 8.3? If yes, worth checking XO team to also making this option visible in the UI

psafont

@olivierlambert said:

@psafont do you know if it's already available since then in XCP-ng 8.3? If yes, worth checking XO team to also making this option visible in the UI

Yes, we're shipping xapi 26.1.3, which includes this enhancement.

olivierlambert

Thanks! I will make sure the Maintenance mode button in XO extends with this new parameter.

Pilow

@olivierlambert having done a RPU yesterday on 3 hosts, I still have the "bug" where some VMs with "auto start" switched ON, but halted on purpose, reboot when a host is rebooted.

we do stop some VMs during RPU to lessen the migrate times, but at every host reboot, they DO START UP

annoying. not critical, but annoying.

stormi

@Pilow Are these VMs HA-protected? If yes, maybe when HA is re-enabled after the RPU it believes that the VM is down and must be started.

There's a pool-wide setting tell HA not to restart a VM that was manually stopped:

xe pool-param-set uuid=<your pool uuid> ha-reboot-vm-on-internal-shutdown=false

Pilow

@stormi no, we do not use HA, it's disabled on the POOL and on the VMs

Pilow

I did verify and yes

    ha-reboot-vm-on-internal-shutdown ( RW): true

it is enabled on our Pool. but no HA

                           ha-enabled ( RO): false

stormi

So, if I understand correctly, you would like that a VM stopped on purpose but still marked as autostart on be not started during a RPU. I'm not sure how the system would know that the VM was stopped on purpose, after the reboot.

Pilow

@stormi indeed.

but the host restarting is by design empty of all VMs because of evacuate process

the stopped VMs are on other hosts. so strange to see them booting when restarted host is coming online.

ps : as I wrote it, I understood my "error". halted VMs are on no host at all, just attached to the pool.

Pilow

perhaps "in the context of a proceeding RPU, do not start halted VMs" ?

but I can imagine corner cases where this is not wanted.

some chicken & egg problem.