Monitoring with Graphite by Jason Dixon

Author:Jason Dixon
Language: eng
Format: epub, mobi, pdf
Publisher: O'Reilly Media
Published: 2017-03-27T04:00:00+00:00

Figure 6-13. Discovering series outliers with mostDeviant()

While mostDeviant() definitely has value for weeding through noisy metrics, it can resemble a blunt force object at times. Sometimes you need a little more surgical precision with your filters. Often, I’m faced with the prospect of deciphering a graph where the scale has been thrown out of whack due to an accidental or anomalous datapoint. It doesn’t always have to be a malignant metric; sometimes it’s just a value that’s so obtuse that it gets in the way of an otherwise straightforward troubleshooting operation.

In these cases, it helps to be able to isolate and ignore those values. Graphite has a handful of functions for excluding values above or below a given numeric threshold. Depending on the function, the threshold can represent either a static value or a percentile. These are the permutatively named: removeAbovePercentile(), removeBelowPercentile(), removeAboveValue(), and removeBelowValue().

Figure 6-14 demonstrates the effect of a single “bad” datapoint. This particular metric measures the run time of a batch job, usually hovering between 900-1000ms. A single anomaly was identified and resolved, but its impact is felt long afterwards. The team responsible for this app can’t easily tell if there’s a 10–20% performance regression (the sort of thing you might look for during normal development cycles) due to a spike that happened hours in the past.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.