Shut down metrics server in on_stopped so it only stops when main puma process stops and ignores when workers are stopped. This allows us to maintain a running metrics server when workers restart or crash.
To replicate issue, start up puma server with multiple workers, kill one of the workers, and notice that the metrics endpoint no longer works.
Shut down metrics server in on_stopped so it only stops when main puma process stops and ignores when workers are stopped. This allows us to maintain a running metrics server when workers restart or crash.
To replicate issue, start up puma server with multiple workers, kill one of the workers, and notice that the metrics endpoint no longer works.