Comment 9 for bug 1720816

I believe I am seeing the same issue where Magnum (and Heat) have resources in the "CREATE_IN_PROGRESS" state for a long time, and eventually timeout. I'm using devstack (stable/pike) and the Mesos COE - but I suspect this is seen with other COE's

After some debug, I"ve identified an issue where OS::Heat::SoftwareDeploymentGroup and OS::Heat::SoftwareDeployment (at least) are failing to signal their completion using heat-config-notify. Specifically, the POST is failing with 403 error codes.

It appears (root cause still unknown) that the HTTP request parameters (req.params) passed to EC2Token._authorize() contains more data than the original HEAT signature was generated with. As a result the Keystone ec2tokens request fails.

The additional data in the req.params is the date from the /var/lib/heat-config/deployed/*.notify.json. I'm not yet sure how it ends up in req.params (more investigation is required).

Attached is a temporary patch that removes these additional keys from req.params (the JSON of notofy.json is the key, and the associated value is '')

I'm keen to see if this patch addresses the concerns seen by others on this issue.