Is aws cloudfront requests average the right value from infrastructure cloudwatch metrics connector?


We use the infrastructure AWS cloudwatch metrics connector through grafana cloud to import some cloudwatch metrics in a prometheus format:

It works pretty well but one of the metrics that we want to use seems to display a wrong value. It’s the “aws_cloudfront_requests_average” metrics. In my humble opinion, the right metrics would be “aws_cloudfront_requests_sum”. If I query directly cloudwatch, I have to use sum instead of average if I want to display the right request count of a cloudfront distribution.

Can anyone tell me if they have the same problem or if I am using it incorrectly? I’m looking for the github repository to open an issue but can’t find it if anyone has a hint that would be nice.

Thanks a lot!


1 Like

Seems to be the same for the metric “aws_cloudfront_bytes_downloaded_average”. Rate metric are correctly handle (in percent) with average function but gauge of count are not. Or I missunderstand the way to use it. The trends of the graph is good but the value is not.


Hi @juliensimontwicpics let me see if I can shed some light on what’s going on here,

At the moment we only support pulling the average statistic for all metrics. We are actively working to allow you to pick which metric + statistics you would like the integration to pull.

I think this is a similar situation as you mentioned in the first post. If you are attempting to compare data between the integration and cloudwatch you will need to ensure the statistics are the same. The other piece to note is the choice of Period for CloudWatch. We use a period of 5 minutes to match the scrape interval. The data points should match up if you make these adjustments to your CloudWatch options with the only difference being the timestamp.

Hopefully that helps and happy to answer any further questions.

Hey Kyleeckhart,

Thanks for this insight. Yes it help me understand better the behavior of this integration.

Best regards,