We've tried to use this tool to export some metrics to our Prometheus. What we noticed immediately was that the Memory Usage data reported by our prometheus was completely different than the data reported by AWS Container Insights. We've investigated the issue and learned that the differences are caused by the way AWS calculates the usage of the memory inside of the container.
With this users of esc_exporter would be able to decide how they want to present the data. We could do it in the same way AWS does (by substracting cache value), show both on a graf etc.
Hello,
We've tried to use this tool to export some metrics to our Prometheus. What we noticed immediately was that the Memory Usage data reported by our prometheus was completely different than the data reported by AWS Container Insights. We've investigated the issue and learned that the differences are caused by the way AWS calculates the usage of the memory inside of the container.
As you can see in the link below, AWS calculates the usage as the difference between usage and stats.cache (It was changed in 2016): https://github.com/aws/amazon-ecs-agent/commit/f2019aee016a1be1e340f06147be24367d6f96a6
Could we add this cache metric to the list of exposed metrics. I'd propose naming it as:
With this users of esc_exporter would be able to decide how they want to present the data. We could do it in the same way AWS does (by substracting cache value), show both on a graf etc.