We should track utility process stability metrics more accurately, in particular the stability of separate mojo services and other users of utility processes.
At a minimum we should track launches and crashes (and kills?).
Corresponding internal bug is b/64566828
as discussed with metrics team, the idea here is to have a set of sparse histograms that cover the following events:
- Launch
- Crash
- Killed
For utility processes, based on the hash of the process name. The hashes will be resolved server-side based on enums.xml additions. Hopefully we can keep this manageable.
Comment 1 by wfh@chromium.org
, May 24 2018