[pve-devel] training week : question about HA and service in error

[Alexandre DERUMIER]

Hi,

we are currently testing HA with last proxmox from no subscription repository,
and I think we have a bug for a specific case.

we defined an HA group with only 1 server : kvmformation1, with restricted.

vm106 is in the hagroup and run in kvmformation1.


when kvmformation1 is crashing, the vm ha state is going to "error"

Dec 8 12:18:59 kvmformation2 pve-ha-crm[5884]: recovering service 'vm:106' from fenced node 'kvmformation1' failed, no recovery node found 
Dec 8 12:18:59 kvmformation2 pve-ha-crm[5884]: service 'vm:106': state changed from 'fence' to 'error'


Then, when the kvmformation1 server is up again,

the vm is not restarted because of ha error state. (and we can't restart it manually with start button)


we need in this case, set "disable" state manually for this vm for reset HA state.
then reenable HA.


Is it a bug ?

Regards,


Alexandre