Closed ngrewe closed 3 years ago
Hey @ngrewe,
thank you for your contribution! Would it be possible for you to add a specific test? I'm not feeling well merging this without good test coverage.
Cheers,
Thanks Lukas! I've added a test, which was also very instructive because it revealed that the problem is only triggered under some very unusual circumstances (very high setting for serialize_precision
in php.ini, values that happen to fall through the cracks of IEEE 754).
We also noticed that all hell breaks loose if the PHP process doing the collection has a different value for serialize_precision
as the one observing the metric (essentially the collector process will see most values as 0 because the generated string for the buckets does not agree with what the observing process wrote to Redis). Not sure whether defending against that is worthwhile though…
Hi @LKaemmerling,
just checking: Is there anything else that you need me to do here?
Thanks,
Niels
Hey @ngrewe,
sorry, i completely missed it while being on vacation and then the easter holiday.
Everything is fine for me, looks good! Thank you! Good job!
Hi folks,
I noticed that the LUA script for updating histograms currently does not work in many cases because it expected
HINCRBYFLOAT
to return the increment instead of the new value of field in the hash. It would also fail if the value returned byHINCRBYFLOAT
differed in formatting from the input value. This PR fixes the issue.Thanks,
Niels