FastBoot should emit metrics

[kratiahuja]

This issue is an enhancement for future purposes.

Post FastBoot 1.0, we should look at emmitting some useful metrics like server side latency, memory used by the node processes so that apps using FastBoot can listen to these metrics and create useful operational dashboard.

This probably will need an RFC at some point IMO.

[sdhull]

notes from discussion in fastboot meeting:

• would tooling be low-overhead and on all the time?
• would there be some debugging mode that you can turn on just to debug?
• are there specific metrics you want?
• time from when the visit request started to when the promise was resolved
• when rehydration starts, how long does it take?
• we're not going to do better than existing node.js tools re: memory leaks & stuff
• express & node have tooling
• what kind of unique-to-ember things can we emit?
• did a rerender get scheduled?
• did you use a tag that is known to have issues with rehydration? (eg <select> [aside: wait, for real?])

@ryanto has some middlewares for some tooling he might be willing to share...?

[rondale-sc]

I believe in the rehydration builder there is a path that happens when a node is skipped or removed (when the rehydration builder can't decipher something so it throws it out). It would be great to see a counter of that accessible in some sort of debug mode to determine how effective rehydration is.