Failures when starting VMs are not detected

[tcrivat]

If a VM fails to boot for any reason (misconfiguration, cloud reaching its max. capacity etc.), ConPaaS fails to detect this and waits indefinitely for the VM to become ready (which obviously never happens). The service remains in an inconsistent state and cannot be stopped or deleted by any means (I tried this using both the graphical front-end and the command line client). Consequently, an application containing such a service cannot be deleted.

This is not only annoying to the user, but also hampers testing: before reaching a working configuration, the account becomes cluttered with lots of test applications that cannot be removed.

[gpierre42]

This is a bit more subtle than that. I believe this problem happens only when a manager VM fails to start. Managers should normally have a timeout mechanism in case an agent VM fails to start.

It don't think it can be too difficult to implement the same mechanism in the director as well? Maybe I am underestimating the difficulty...