prometheus: add some extra buckets for upload metrics #591
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
✅ All tests successful. No failed tests found. @@ Coverage Diff @@
## main #591 +/- ##
=======================================
Coverage 97.55% 97.55%
=======================================
Files 425 425
Lines 35593 35593
=======================================
Hits 34724 34724
Misses 869 869
Flags with carried forward coverage won't be shown. Click here to find out more.
|
Codecov ReportAll modified and coverable lines are covered by tests ✅ ✅ All tests successful. No failed tests found. @@ Coverage Diff @@
## main #591 +/- ##
=======================================
Coverage 97.55% 97.55%
=======================================
Files 425 425
Lines 35593 35593
=======================================
Hits 34724 34724
Misses 869 869
Flags with carried forward coverage won't be shown. Click here to find out more.
|
Codecov ReportAll modified and coverable lines are covered by tests ✅
✅ All tests successful. No failed tests found. @@ Coverage Diff @@
## main #591 +/- ##
=======================================
Coverage 97.55% 97.55%
=======================================
Files 425 425
Lines 35593 35593
=======================================
Hits 34724 34724
Misses 869 869
Flags with carried forward coverage won't be shown. Click here to find out more.
|
Codecov ReportAll modified and coverable lines are covered by tests ✅
✅ All tests successful. No failed tests found. Additional details and impacted files@@ Coverage Diff @@
## main #591 +/- ##
=======================================
Coverage 97.60% 97.60%
=======================================
Files 460 460
Lines 36799 36799
=======================================
Hits 35916 35916
Misses 883 883
Flags with carried forward coverage won't be shown. Click here to find out more.
This change has been scanned for critical changes. Learn more |
| # The 0.98 bucket is to stop Prometheus from interpolating values much | ||
| # lower than 1 in its histogram_quantile function. | ||
| buckets=[0.98, 1, 5, 10, 20, 50, 100], |
There was a problem hiding this comment.
This is really weird tbh. Will this ever emit metrics <1? In which case, we would probably want to early-return much earlier? Also, this metric would probably benefit from a bunch more smaller buckets, maybe even all the way from 1..=10?
There was a problem hiding this comment.
agree this is very weird. i don't think we expect any raw uploads to contain 0 raw reports, but i don't know whether it can happen anyway. that's part of why i'm introducing this weird bucket
the reason for it is: the true median may be 1.0, but all prometheus can say for sure is that it's in the bucket between 0.0 and 1.0. based on the relative size of the bucket, it will interpolate a value in that range to call the median
by adding a bucket at 0.98, the range for the estimated median is limited to 0.98-1.0. so it'll look nicer on a graph, but also, we can ask prometheus how many values lower than 0.98 were observed to know whether any 0-report uploads are coming in. 0.99 or higher would probably work but i am being overly cautious of floating point
There was a problem hiding this comment.
more granular buckets in the single digits is a good idea, i'll rebalance the buckets a bit
(click on VPN) https://l.codecov.dev/S0nAAE is a dash with some new upload size / report size metrics. it has some quirks:
histogram_quantileinterpolates values <1 for the number of reports in an uploadi added some buckets to help with those quirks