/etc/hostname no longer present on OpenShift workers

[thanusiak]

Testing agent-resources-openshift.yaml
with:
Server Version: 4.10.18

I got the following error:

Events:
  Type     Reason          Age              From               Message
  ----     ------          ----             ----               -------
  Normal   Scheduled       8s               default-scheduler  Successfully assigned ibm-observe/logdna-agent-test-znlr2 to worker12.test.com
  Normal   AddedInterface  7s               multus             Add eth0 [10.254.60.102/22] from openshift-sdn
  Normal   Pulled          5s               kubelet            Successfully pulled image "icr.io/ext/logdna-agent:3.5.2" in 1.319768025s
  Normal   Pulling         4s (x2 over 6s)  kubelet            Pulling image "icr.io/ext/logdna-agent:3.5.2"
  Warning  Failed          3s (x2 over 5s)  kubelet            Error: Cannot mount /etc/hostname: path does not exist and will cause issues as a directory
  Normal   Pulled          3s               kubelet            Successfully pulled image "icr.io/ext/logdna-agent:3.5.2" in 1.282495926s

due to

- name: logdnahostname
          hostPath:
            path: /etc/hostname

Not sure when the change happened in OCP, but workers in fact don't have the hostname file by default.

Replacing path: /etc/hostname with :
path: /proc/sys/kernel/hostname

seems to resolve the problem.

[dkhokhlov]

that hostname line has not been touched for very long time and that file should exist in Linux.

I see that file when I run the image:

~$ docker run -it icr.io/ext/logdna-agent:3.5.2 bash
Unable to find image 'icr.io/ext/logdna-agent:3.5.2' locally
3.5.2: Pulling from ext/logdna-agent
54e56e6f8572: Pull complete
4f8ddd7f5a75: Pull complete
769927f5652f: Pull complete
1e852efbf2ff: Pull complete
Digest: sha256:1bb082da732c6e2d6bda583832c4f4dd3f2891cd824682b61c80988d52dc0dcb
Status: Downloaded newer image for icr.io/ext/logdna-agent:3.5.2

[root@ec6a9b7c3744 work]# cat /etc/hostname
ec6a9b7c3744

[thanusiak]

If you run it via docker locally or on standard Linux box it will be present, however it will not be present when run off a CoreOs worker (Openshift).
See similar discussion (it does not seem to be specific to FCOS):
https://ask.fedoraproject.org/t/location-of-etc-hostname-on-coreos/13426

[dkhokhlov]

ic, you are using FCOS.

Probably something todo with FCOS vs RHCOS.

Red Hat Enterprise Linux CoreOS (RHCOS) is the only operating system supported on the OpenShift control plane nodes.
https://cloud.redhat.com/blog/red-hat-enterprise-linux-coreos-customization

@c-nixon ?

[thanusiak]

@dkhokhlov Im referring to RHCOS, just couldnt find any discussion specific to it.

[dkhokhlov]

ic. might be OCP 4 vs OCP 3.
some search on RH portal:
https://access.redhat.com/search/#/?q=%22%2Fetc%2Fhostname%22%20openshift

@thanusiak
are you happy with proc/sys/kernel/hostname solution?

[thanusiak]

It seems to works as expected, so Im good with this solution