graphite-dev team mailing list archive

[Jase Wells <question221830@xxxxxxxxxxxxxxxxxxxxx>]

New question #221830 on Graphite:
https://answers.launchpad.net/graphite/+question/221830

I am fetching JSON data from Graphite to show total page views, using summarize to get hourly or daily totals. I've noticed that sometimes the data fetched is way off, then it goes back to normal. To reproduce the problem, I created a test query string to fetch the same data twice:

http://graphite/render/?format=json&from=-1d&target=summarize(sum(stats.counters.prod.main.view.app.*.*.count),'1d')&target=summarize(sum(stats.counters.prod.main.view.app.*.*.count),'1d')

I would expect the two targets to produce the same result, and often it's the same or very close. (Maybe in the milliseconds it takes to calculate the first result, a few extra counts get added?) But then sometimes it's out of whack like this:

[{"target": "summarize(sumSeries(stats.counters.prod.main.view.app.*.*.count), \"1d\", \"sum\")", "datapoints": [[2167.6833333333311, 1360713600], [64.666666666666657, 1360800000]]}, {"target": "summarize(sumSeries(stats.counters.prod.main.view.app.*.*.count), \"1d\", \"sum\")", "datapoints": [[4390.0, 1360713600], [93.0, 1360800000]]}]

Am I wrong to think the same target should produce the same result? The one that's out of whack has a bunch of decimal places, which looks suspicious. Sometimes the lower value is the first target result, sometimes it's the second.

By the way, the relevant schema looks like this:
[stats]
pattern = ^stats\.
retentions = 10s:1d,1min:30d,1h:2y

And the relevant aggregation looks like this:
[sum]
pattern = \.count$
xFilesFactor = 0
aggregationMethod = sum

Any suggestions or clues what I should look into to prevent the odd summarize results?


-- 
You received this question notification because you are a member of
graphite-dev, which is an answer contact for Graphite.