In Distributed configuration, when the master slave is stopped (expected situation or due to a crash) and restarted,
unlike slave nodes which it synchronizes its "global registry state" knowledge before starting any of its local installed
components/Service assembly, it does not. This means that if a component is installed in the master node and have to
sends messages to endpoints located on some slave nodes that currently running then it will automatically failed because
the message receiver endpoint is unknown.
The problem can be reproduce with Logica use-case : petals-se-notification and petals-se-kpi component are supposed to be
deployed one node - the "slave" one - whereas other components of the use-case are deployed on a different node
- the "master" one. Then simulate crash on master node --> Notification producer can not send registerPublisher() request because
the Notification broker endpoint is unknown.
Bug
Resolved
Critical