No Median CPU Usage in my track results

[bernitorres]

Hello,

I created 3 different tracks to compare the results of some mapping changes. Out of these 3 tracks, only one of these doesn't report the Median CPU Usage. The metric is not reported at all, not even with a blank value.

What could be the reason that CPU usage is in the other 2 track results but not this one?

I am indexing 10 million documents
esrally 0.5.3

esrally race --track=flat --distribution-version=1.7.3

Lap	Metric	Operation	Value	Unit
All	Indexing time		0.19725	min
All	Refresh time		0.0711167	min
All	Median CPU usage		97.9	%
All	Total Young Gen GC		0.635	s
All	Total Old Gen GC		0.068	s
All	Index size		0.0413944	GB
All	Totally written		0.116398	GB
All	Heap used for segments		0.352222	MB
All	Segment count		22
All	Min Throughput	index-append	1365.1	docs/s
All	Median Throughput	index-append	1365.1	docs/s
All	Max Throughput	index-append	1365.1	docs/s
All	100th percentile latency	index-append	7116.46	ms
All	100th percentile service time	index-append	7116.46	ms
All	error rate	index-append	0	%

.....

[danielmitterdorfer]

Hi @bernitorres,

Based on the metrics I see here (e.g. index size is just 40 MB) I guess that your benchmark does not take very long and sometimes even finishes within the warmup period. Rally will not include any warmup samples in the final report and that may be the reason why you do not see any CPU usage results.

If you store your metrics in a dedicated Elasticsearch metrics store, it will contain all raw samples and you could run a query to check this. You could also (temporarily) set the warmup time period of the bulk-index operation in your track to 0 to check my assumption. However, I suggest to index more documents as you're basically only measuring Elasticsearch's warmup behavior.

Daniel