We need to propagate the result from the cloud platform.
AppScale - self deployed
The infrastructure manager is not propagating the error up when making call to the cloud infrastructure. For example the error bot gets with out of resources, or the error we get on our private testbed of message expired, are not differentiate, so we are not able to retry or crash or inform the user properly.
The expire message error, was possibly due to the eucalyptus NCs default configuration which uses 'localtime' for the instances. Changing that to 'utc' seems to have limited those errors. To reproduce them again, we can revert it back.
- Target version changed from 2.6.0 to 2.7.0
- Target version deleted (
- Priority changed from Normal to Low
- Target version set to backlog
- Status changed from New to Rejected
Closing this for now: we are doing progress on an agent basis first.
Also available in: Atom