Unexpected 1 hours vs 2 hour graph

[gtorrance (Greg T)]

Hi all,

I noticed some unusual results when comparing the graphs for node garbage collection for a 1 hour period vs a 2 hour period. The two graphs are below, and are taken at essentially the same time.

I'm confused why the 2-hour graph shows spikes almost to 40,000ms, whereas the 1 hour graph only shows spikes up to 30,000ms. Why would this be? I could understand a reverse situation -- where a 1 hour graph is able to show a greater level of detail (that maybe gets lost from a 2 hour graph). But this is the other way around (i.e. seemingly greater detail over the larger time period).

Any thoughts?

Thanks,
Greg

[Marius_Dragomir]

The size of the buckets(the steps in the date histogram on the X axis) when doing the aggregation is different between the "Last 1h" and "Last 2h" graph. And in the case of GC Duration, which does a sum for that bucket, you will see larger values when the bucket is larger.
That 10 seconds in the top might have led you to think otherwise, but it's not the size of the step in the graph, but how often the graph refreshes.

[gtorrance (Greg T)]

Thanks for the reply, Marius! That makes sense.

Greg