We have a meteor application serving ~10’000 users. From time to time, we experience the server to become unresponsive (hang / freeze). This means it does not directly return a HTTP 5xx error-code but clients can no more establish a websocket connection.
The server is deployed using ‘mup’. If the server becomes unresponsive, I can connect via ssh and show the logs. However, this is not very helpful to find the cause of a freeze.
As a workaround, I have to restart meteor (mup stop, mup start) to get everything back online. This is very unsatisfying.
What techniques do you recommend to debug such hangs? How can I connect a local debugger to productive meteor / mup environment? Is there any documentation to start with?