wishlist: maintenance mode for nodes

[robb1e]

Using Xen, when I restart the node controller and wait for a short while (e.g. updgrading code), the cloud controller marks the instance as terminated. When the node controller comes back online, it finds the instance and then terminates it.

It'd be great if the node controller could tell the cloud controller that it was going into a maintenance mode for a short while and for the cloud controller to assume all is well until the node controller comes back online. During the maintenance window, the cluster controller could buffer incoming requests and pass it down stream when available.

This could also be of benefit if there was a network issue and the node controller was unavailable for a short while, although the node wouldn't be able to inform the cloud that it was going to be unavailable. Perhaps this could be achieved with an 'unknown' state for the instance.

Cheers

[Andy Grimm (agrimm)]

This issue is now being tracked upstream at http://eucalyptus.atlassian.net/browse/EUCA-2643

Please watch that issue for further updates.