You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think we should add another callback to the installer which is triggered when the installer machine is going down.
This callback should
Close/cleanup the installRequest on the machine-service
Be relayed to the manager for the machine so that it can perform post-install things. It would be ok to simply reply 501 Not implemented as well.
Alternatively, the playbook would not reboot the machine by calling reboot or something traditional, but rather call on a tateru.deploy.finish module which performs basically the same steps. In that case, this would not be a callback but a normal endpoint on the machine service and be called from the client. This solution seems cleaner implementation wise, but requires all managers to implement another endpoint.
Why is this needed? In an ideal world this would not be needed. But for instance the proxmox installer has a problem: The change of bootorder requires a real shutdown of the VM. So either we make the ansible playbook do things directly against the proxmox cluster - in that case why do we have the manager in the first place, we could do the reconfiguration and the start (bootInstalleRequest) in Ansible without any managers... And if it is not needed for a certain manager, than simply return 501 for the endpoint and be done.
This lets tateru handle all the problems with the "hardware" and ansible with the configuration of the operating system.
The text was updated successfully, but these errors were encountered:
I think we should add another callback to the installer which is triggered when the installer machine is going down.
This callback should
501 Not implemented
as well.Alternatively, the playbook would not reboot the machine by calling reboot or something traditional, but rather call on a tateru.deploy.finish module which performs basically the same steps. In that case, this would not be a callback but a normal endpoint on the machine service and be called from the client. This solution seems cleaner implementation wise, but requires all managers to implement another endpoint.
Why is this needed? In an ideal world this would not be needed. But for instance the proxmox installer has a problem: The change of bootorder requires a real shutdown of the VM. So either we make the ansible playbook do things directly against the proxmox cluster - in that case why do we have the manager in the first place, we could do the reconfiguration and the start (bootInstalleRequest) in Ansible without any managers... And if it is not needed for a certain manager, than simply return
501
for the endpoint and be done.This lets tateru handle all the problems with the "hardware" and ansible with the configuration of the operating system.
The text was updated successfully, but these errors were encountered: