When using heatmaps in Grafana, non-uniform bins cause visual artifacts that can be quite misleading, or hide actual features of significance. The important feature is that the bin distribution should be smooth, i.e. either linear, logarithmic, polynomial, etc.
We have a few hand crafted bin sets in our metrics that are not very uniform (neither linear nor logarithmic). This PR fixes one of those that was causing gfr confusion in mlab-sandbox development.
metrics.FileSizeHistogram will now have a smooth bin distribution, and easier interpretation. It uses 7 bins per decade, to approximate the density in the previous bin set.
NOTE: When grafana heatmaps cover a period with multiple bin sets, the display will be almost meaningless. Changing the time window to an interval with a single bin set resolves the problem.
When using heatmaps in Grafana, non-uniform bins cause visual artifacts that can be quite misleading, or hide actual features of significance. The important feature is that the bin distribution should be smooth, i.e. either linear, logarithmic, polynomial, etc.
We have a few hand crafted bin sets in our metrics that are not very uniform (neither linear nor logarithmic). This PR fixes one of those that was causing gfr confusion in mlab-sandbox development.
metrics.FileSizeHistogram will now have a smooth bin distribution, and easier interpretation. It uses 7 bins per decade, to approximate the density in the previous bin set.
NOTE: When grafana heatmaps cover a period with multiple bin sets, the display will be almost meaningless. Changing the time window to an interval with a single bin set resolves the problem.
This change is![Reviewable](https://reviewable.io/review_button.svg)