Bug 1368426

Created attachment 1192105 [details]
patch

Description of problem:

We have observed a regression in Docker 1.10.3 (used with OpenShift 3.2.1.x) wrt Docker 1.8.2 (with OpenShift 3.1): the `docker stats` command shows zero-values for metrics on containers that share the network of another container.  This is particularly visible in the context of OpenShift since all containers of a pods share the network namespace of the openshift3/ose-pod container.


This is a known bug, fixed in Docker 1.12.  See:

- Bug: https://github.com/docker/docker/issues/21848
- Fix: https://github.com/docker/docker/pull/21904


Version-Release number of selected component (if applicable):

All systems running up to date RHEL 7.2 / Docker 1.10.  This is limited to dev/test currently.


How reproducible:
The behaviour is repeatable.

Steps to Reproduce:

Test case taken from #21848:

```
[root@eric-rhel72 ~]# export PAUSE_ID=$(docker run -d -m 10M kubernetes/pause)
[root@eric-rhel72 ~]# CHILD_ID=$(docker run -d -m 100M --net container:$PAUSE_ID kubernetes/pause)
[root@eric-rhel72 ~]# docker stats $PAUSE_ID $CHILD_ID
CONTAINER                                                          CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O
2383fd90a8bca7788f32e16e72e4d868693880692fd37e62c5ed2426f43e601a   0.00%               2.367 MB / 10.49 MB   22.58%              648 B / 648 B       0 B / 0 B
7ede4312d597935e4a18e6078c20872e322cb4cb48447479662d41758906f8a5   0.00%               0 B / 0 B             0.00%               0 B / 0 B           0 B / 0 B
[root@eric-rhel72 ~]# rpm -qi docker
Name        : docker
Version     : 1.10.3
Release     : 46.el7.10
Architecture: x86_64
Install Date: Tue 16 Aug 2016 12:25:17 PM UTC
Group       : Unspecified
Size        : 46245001
License     : ASL 2.0
Signature   : RSA/SHA256, Fri 29 Jul 2016 07:26:38 PM UTC, Key ID 199e2f91fd431d51
Source RPM  : docker-1.10.3-46.el7.10.src.rpm
Build Date  : Fri 29 Jul 2016 05:46:27 PM UTC
Build Host  : x86-039.build.eng.bos.redhat.com
[...]
```

Additional info:

Fix:

Adapting and applying the patch applied to docker https://github.com/docker/docker/pull/21904, and building + installing the RPMs shows the patch resolves the issue:

```
[root@eric-rhel72 ~]# export PAUSE_ID=$(docker run -d -m 10M kubernetes/pause)
[root@eric-rhel72 ~]# CHILD_ID=$(docker run -d -m 100M --net container:$PAUSE_ID kubernetes/pause)
[root@eric-rhel72 ~]# docker stats $PAUSE_ID $CHILD_ID
CONTAINER                                                          CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O
1ddf815fb6b4f8418c1eca027b1e7a0f36c6b9e1f710642d70507d665a0aea37   0.00%               290.8 kB / 10.49 MB   2.77%               648 B / 648 B       0 B / 0 B
2d6425dabe73fb1724580ba74959591c4969c7735f19993389291f8ee23f650f   0.00%               290.8 kB / 104.9 MB   0.28%               648 B / 648 B       0 B / 0 B
[root@eric-rhel72 ~]# rpm -qi docker
Name        : docker
Version     : 1.10.3
Release     : 46.el7.10.patch
Architecture: x86_64
Install Date: Tue 16 Aug 2016 12:20:47 PM UTC
Group       : Unspecified
Size        : 46273796
License     : ASL 2.0
Signature   : (none)
Source RPM  : docker-1.10.3-46.el7.10.patch.src.rpm
Build Date  : Tue 16 Aug 2016 09:52:43 AM UTC
Build Host  : eric-rhel72.novalocal
[...]
```