-
Suggestion
-
Resolution: Fixed
-
None
- configurable grace period between when the server detects that the heartbeat has stopped and when the instance is shutdown.
- backend functionality to terminate arbitrary instance exists already
[BAM-3264] System (server) will shutdown EC2 instance when agent heartbeat stopped.
No there's no functionality to restart, but there's a use case there for debugging when things goes wrong
So that's why we give them a time out
What is the reason behind delaying instance shutdown if an agent is killed and unregistered - after that it cannot get back online and I have not seen any functionality to start an elastic agent on a running instance.
To clarify:
- When heartbeat period expires, the agent is killed immediately & the instance killing wait period begins
- The instance gets killed after the wait period
The shutdown should be triggerred happen when an agent has stopped. The trigger should wait the timeout in ElasticConfiguration before it actually shuts down the agent.
Current behavior is that when remote (or elastic) agent does not send heartbeat for bamboo.agent.heartbeatTimeoutSeconds then LocalAgentManager.stopAgent is called. This method for elastic agents performs deregistration of the agent.
Shall this behavior be retained or shall I change it so it will not be unregistered immediately after heartbeat delay?
moving to M5 since there is no doubt review work required here.