Dangling NATS Consumer Causes Client Not to Receive Result Event After Abnormal Exit

[anhthii]

Dangling NATS Consumer Causes Client Not to Receive Result Event After Abnormal Exit

Description

When the mpcium-client sends an event (e.g., keygen request) to the MPCIUM cluster, it may fail to receive the corresponding result (e.g., mpc_keygen_result) if a previous session exited abnormally. This results in the NATS consumer (mpc_keygen_result) still showing Waiting Pulls: >0, even though the client is no longer running.

As a result, new instances of the client may not receive responses due to lingering/dangling pull requests.

---

Investigation

• Stream: mpc
• Consumer: mpc_keygen_result
• nats consumer info showed:
  • Waiting Pulls: 3 of maximum 512
  • Unprocessed Messages: 0
• No result was delivered to the new client instance.

After running:

nats consumer rm mpc mpc_keygen_result

Suggestions

Documentation:

• Add this as a "Troubleshooting" entry in the docs.

• Explain how to check Waiting Pulls and clean up dangling consumers.

Code Improvements:

• Auto-detect and clean up orphaned consumers on client start.

• Consider using ephemeral consumers or appending a unique instance ID to the consumer name.

• Implement a client shutdown hook to remove the consumer cleanly.