fix(cluster): fix server outages due to cluster resource leakage

[Splaktar]

Add initWorker() to server which should clean up worker resources on both disconnect and exit.
Reference nodejs/node-v0.x-archive#9409
Update JSHint for server and fix some issues including removing unused requires.
Add key entries for frisbee keys.
Remove some newRelic cruft.

Fixes #54.

[Splaktar]

Please review the changes to server.js but do not merge this until I see if this actually fixes the issue.

This has been deployed to production, but I will need to see the server run for over 4 days straight without crashing before I can have any feeling that this might actually fix the resource leak.

[tasomaniac]

Looks good but I don't understand the part with the Frisbee client ids. They were used but not present? Could that be the reason that gcm registration was failing and we had to disable it in the app.

[Splaktar]

There was an exception thrown by the server last night due to config.keys.frisbee not being defined when trying to look up config.keys.frisbee.serverClientId. There was nothing in the code, readme, or documentation about this config.keys.frisbee configuration. So it's a new thing to me that was apparently undocumented and only setup on the deployment VM in OpenShift.

[Splaktar]

Still online, hoping it makes it through the weekend...

[tasomaniac]

👍

[Splaktar]

7 days now and the server is still up. Looks like this fix is working.

[tasomaniac]

👍

On Mon, Oct 5, 2015, 2:49 AM Michael Prentice [email protected]
wrote:

7 days now and the server is still up. Looks like this fix is working.

—
Reply to this email directly or view it on GitHub
#57 (comment).

[tasomaniac]

👍