You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Jun 20, 2024. It is now read-only.
I'm writing up a longer description, but as a place-holder:
If you run a program on Linux with a cgroups memory limit (and Kops runs Weave Net with a 200MB limit), and the data space of the program gets close to the limit (say 198MB against a 200MB limit), then Linux will be continuously paging the code of the program in from disk.
If you have Prometheus monitoring you can look at this metric:
I'm writing up a longer description, but as a place-holder:
If you run a program on Linux with a cgroups memory limit (and Kops runs Weave Net with a 200MB limit), and the data space of the program gets close to the limit (say 198MB against a 200MB limit), then Linux will be continuously paging the code of the program in from disk.
If you have Prometheus monitoring you can look at this metric:
and if the rate is hundreds or thousands of pages per second then you have a problem.
Suggested workaround is for the code to call
mlockall()
to tell Linux never to discard any pages.The text was updated successfully, but these errors were encountered: