- AppSignal for Ruby gem versions
1.x- most recent.
- AppSignal for Elixir package versions
0.0.x- most recent.
- Systems: Containerized systems such as Docker and Heroku (LXC).
Host metrics reported by AppSignal in the host metrics feature are incorrect, reporting the data for the container's host system instead.
AppSignal uses a library called probes-rs, created by AppSignal, to report host metrics for the host on which the AppSignal agent is running. It does this by reading the
/proc virtual filesystem. This directory however, always reports data from the parent host rather than the container's constraint resources. This is why the AppSignal agent is reporting the wrong metrics.
- In the AppSignal host metrics feature the host metrics do not match the container's restraints, reporting resources many times its allowed amount.
No complete solution is available at this time.
Memory host metrics were fixed in Ruby gem
2.8.0 and Elixir package
1.9.0. The other metrics were be improved in the future.
Let us know which metrics you are missing, are inaccurate or if you have a problem, at email@example.com.
Note that in our newer releases the AppSignal agent reads the container runtime metrics if a container is detected. If no limits for your container's memory, CPU, etc. have been configured, the container will report the maximum possible value. A container without swap (support) will report a 0 value. For more information on how to limit your container's memory see the Docker or Kubernetes documentation.
Let us know if it's reporting the memory statistics accurately for your apps or if it doesn't. In which case it would help us a lot if you could send us the
appsignal.log file and some information about your app's container setup so we can reproduce the issue.
None available at this time.