Add collectors and background service for measurements #2239
Conversation
Instructions and example for changelogPlease add an entry to Example: ## Unreleased
- Add collectors and background service for measurements ([#2239](https://github.com/getsentry/sentry-java/pull/2239))If none of the above apply, you can opt out of this check by adding |
There was a problem hiding this comment.
I just gave this a quick glance as it's a big PR. @adinauer, what's the current state of this PR?
|
@philipphofmann the exact measurements we collect are still in flux but the code around it could be given a high level review regarding its concept. |
| Long clockTicks = parseClockTicks(stat); | ||
|
|
||
| // TODO 2-4 ms on first read, then cached 4 - 20 μs | ||
| List<Integer> maxFrequencies = CpuInfoUtils.getInstance().readMaxFrequencies(); |
There was a problem hiding this comment.
Doesn't seem to return useful data on my emulator, need to wait for new device to arrive to retest.
| && startRealtimeNanos != null | ||
| && startElapsedTimeMs != null) { | ||
| // TODO SystemClock.elapsedRealtimeNanos() + Process.getElapsedCpuTime() take 9 - 15 μs | ||
| Long diffRealtimeNanos = SystemClock.elapsedRealtimeNanos() - startRealtimeNanos; |
There was a problem hiding this comment.
This seems like our best bet as it's very fast and can be done synchronously. Probably also less risk of being taken away than reading /proc/self/stat
NOTE: The difference for Process.getElapsedCpuTime() might be higher than for SystemClock.elapsedRealtimeNanos() if multiple cores are under load.
Performance metrics 🚀
|
|
superseded by #2445 |
📜 Description
💡 Motivation and Context
💚 How did you test it?
📝 Checklist
🔮 Next steps