Currently, the node cannot recover form server-manager failures on the fly. When such failures happen, the node has to be restarted and all existing input must be processed again in order to restore the state and resume operation.
✔️ Solution
As a further step to #541, we shall add support for session management on the node, based on snapshots saved by leveraging an existing server-manager feature which saves state between epochs.
This would allow the node to implement proper error handling, and the node would:
keep operating after a failure
recover its state from the most recent snapshot, avoid unnecessary reprocessing of inputs
📈 Subtasks
[ ] Implement stop on inspect if it detects a tainted session
[ ] Add Redis connection to advance and inspect
[ ] Advance must receive path to snapshot and start from it, if it exists
[ ] Implement inspect restart when new session is received from Redis
📚 Context
Currently, the node cannot recover form server-manager failures on the fly. When such failures happen, the node has to be restarted and all existing input must be processed again in order to restore the state and resume operation.
✔️ Solution
As a further step to #541, we shall add support for session management on the node, based on snapshots saved by leveraging an existing server-manager feature which saves state between epochs.
This would allow the node to implement proper error handling, and the node would:
📈 Subtasks