New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Report expected vs. actual samples count to highlight low CPU usage/availability to application #126
Comments
There's 2 problems here:
|
Other timers:
probably needs more thought. the risk is you are led to optimise for cpu when you should be looking at off cpu time (in which case you need a different tool). |
From what you say I take that the time will expire, but the SIGPROF will get swallowed. This matches my observations looking at HP logs. But I would need to experiment more to validate. |
This implies profiler stop sequence should be terminate timer, drain logger, log number of timer emitted signals. |
Rephrasing title to reflect discussion |
@ceeaspb reading the code more closely it looks like I misunderstood how the notification timer hangs together. In particular we rely on:
So if no CPU is used (user or system) no signal will be delivered. This does not mean we cannot follow through on this feature. An estimate of intended sample rate should be sufficient as an indicator. |
Low number of samples likely reflects low cpu usage or a short profiling duration, etc.
But by reporting the cpu cost as a percentage we could lose sight of this (low sample number).
This warning is presented in other sampling profilers (ie. typical / covers off a profiling risk).
How many samples is low? how long is a piece of string? TBC
The text was updated successfully, but these errors were encountered: