缺少一个docker容器的cgroup路径

Question

I use cadvisor to get CPU usage of my docker container, but it's always very small(about 0.001), which should be wrong because it's a production worker container and it should be more than 10% from htop outside the docker. And there's no processes in the container. But it has four processes

I found the cgroup of the container is strange like:

$ cat /proc/14996/cgroup
10:pids:/docker/4d7976bc5814e85cc793cc4926c09c150a0e8bb453b8bc01856e8ac0ef17a34a
9:perf_event:/docker/4d7976bc5814e85cc793cc4926c09c150a0e8bb453b8bc01856e8ac0ef17a34a
8:net_cls,net_prio:/docker/4d7976bc5814e85cc793cc4926c09c150a0e8bb453b8bc01856e8ac0ef17a34a
7:freezer:/docker/4d7976bc5814e85cc793cc4926c09c150a0e8bb453b8bc01856e8ac0ef17a34a
6:devices:/
5:memory:/
4:blkio:/
3:cpu,cpuacct:/
2:cpuset:/docker/4d7976bc5814e85cc793cc4926c09c150a0e8bb453b8bc01856e8ac0ef17a34a
1:name=systemd:/system.slice/docker.service/docker/4d7976bc5814e85cc793cc4926c09c150a0e8bb453b8bc01856e8ac0ef17a34a

I have no idea what to do for this problem, but I can provide more information if necessary.

Answer 1

I assume you are super in checking stats for the right container. What OS we are talking about? In the docs they explain some extra configurations required for RHEL for example.

https://github.com/google/cadvisor/blob/master/docs/running.md

Have you specified any concrete run options? Like --cpu-period=100000 --cpu-q --cpuset-cpus 2 ... ?

You can cross check your status with the docker stats command.

CONTAINER           CPU %               MEM USAGE / LIMIT      MEM %               NET I/O               BLOCK I/O           PIDS
e21c0d937cb9        0.12%               8.27 MiB / 7.624 GiB   0.11%               4.386 kB / 2.192 kB   7.593 MB / 0 B      2

Here I have 0.12% which means it's about one tenth of a percent. In cAdvisor that would be 0.0012 .

If those don't match it might indicate an issue.