-
Type: Bug
-
Status: Resolved
-
Priority: Minor
-
Resolution: Fixed
-
Affects Version/s: None
-
Component/s: CI/CD
-
Epic Link:
-
Team:PLATFORM
-
Sprint:nxplatform 11.1.18
-
Story Points:3
After having fixed NXP-27988, see https://github.com/nuxeo/nuxeo/pull/3334, we've added a smoke test to run the image once it's built with the nuxeoctl start command and ensure that the exit code is 0, making the pipeline stage fail otherwise.
This test is now failing with:
22:34:25 Server started with process ID 118. 22:34:25 Could not start process: /var/run/nuxeo/nuxeo.pid (No such file or directory)
Possible lead: https://github.com/GoogleContainerTools/kaniko/issues/506
Yet, when performing a docker describe on a running container, there seem to be no volumes in its Mounts.
Maybe the problem is that
- The 'maven' container of the pod used to build the image with Kaniko has
/var/run/secrets/kubernetes.io/serviceaccount from jenkins-token-wrq8b (ro)
in its Mounts
- Kaniko magically performs the same mount on the built image.
- Since it's readonly, the mkdir -p /var/run/nuxeo instruction in the Dockerfile fails silently...
------------------------------------------------------------------------------------------------------------------------------------------------
In fact, most certainly a CentOS-related issue.
Fixed by using another directory to store the pid: /var/pid/nuxeo.
- is related to
-
NXP-27988 Fix ownership of /opt/nuxeo/server in the nuxeo Docker image
- Resolved