Closed khelle closed 8 years ago
Supervision and routing now supports passing hashes. Once a container enters the failed state it registers an unique failure hash, which then passes to supervisor. After that, only commands having the same hash are executed, while other are rejected. It solves this problem, but makes developer responsible for passing hashes.
When container fails and escalates the problem to its parent supervision system, it doesn't care what instructions it get and from. It makes it accept instructions that might not be connected to solving the issue. What's more it allows executing the commands that should not be allowed in failed state.