The Mean Time Between Failure (MTBF) is a very useful metric that can be used to asses the stability of an application. It is defined as the uptime of all runs of an app, successful (terminating normally), or unsuccessful (ending up in a crash), divided by the number of runs.
This metric was the core metric displayed by a custom crash tool that I worked on in the past (before discovering Sentry !!), it was plotted over time so we could see if the app was getting more stable or the opposite. For a desktop app, we started around less than one hour, and ended up at more than 30 hours if my memory serves me.
The subtleties are:
-
Right now ravens probably don’t send anything for “good runs”. This kind of code can be triggered when the object that is used to track crashes goes out of scope.
-
On iOS/Android, people hardly close their app.
-
When an app is in the “background” (as it can be on Mobile), should we count this time as part of the app runtime or not. It’s easier not to.
Does anyone here who’s using the Mobile SDK think that it’s a useful feature ?
This metric can be tracked independently of sentry, but since it is very crash related it would be beneficial to have it as a core sentry thing. This could also be plotted in the weekly email reports, or could be added to the “Overview” view in sentry as another chart.