Revision history [back]

Having the members of the scaled group scale themselves sounds like a fairly dangerous approach anyway. The intended way to do this is to push statistics from the members into Ceilometer (or another monitoring system of your choice) and have that generate the alarms that trigger the scaling policies.

Note that the workaround you're using is inadvisable on security grounds, as it requires you to put the stack owner's credentials on the deployed servers to allow them to access Heat.