Skip to content

Incorrect CGroup v1 Resource Detection #1683

@talbii

Description

@talbii

Let's assume that cat /proc/1/cgroup outputs:

...
5:cpu,cpuacct:/some_group
...
3:memory:/some_other_group
...

moreover, assume we are running under some container (e.g., Kube). Then it might be the case that the CGroup /memory/some_group doesn't exist, causing us to skip over reading memory values. On the other hand, the container's limits are also present in /memory/memory.limit_in_bytes (for v1 only!) (in fact, this is the system's memory limit so it applies globally).

The same thing of course can happen with CPU values.


This issue was raised by a Discord member, link here.

Metadata

Metadata

Assignees

Labels

MANAGEDbugSomething isn't working

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions