Bug 1473330

Created attachment 1301746 [details]
logs

Description of problem:
Trying to use the nodejs image as a base image for openshift-auth-proxy ends in segmentation fault within 10 seconds of running the container inside OpenShift.


Version-Release number of selected component (if applicable):
- base image: nodejs-{4,6}-{centos7,rhel7}
- OpenShift Origin v1.5.0
- openshift-ansible - current master [1]


How reproducible:
Always, as a docker image


Steps to Reproduce:
1. Deploy OpenShift logging using openshift-ansible
2. Apply a PR to change the base image for openshift-auth-proxy [2]
3. Build the image and deploy to own registry
4. In OpenShift logging project, point openshift/logging-auth-proxy image to the own registry
5. oc rollout latest logging-kibana


Actual results:
Starting up the proxy with auth mode "oauth2" and proxy transform "user_header,token_header".
/var/tmp/scl63Ncam: line 8:   395 Segmentation fault      (core dumped) node openshift-auth-proxy.js


Expected results:
No segmentation fault and running openshift-auth-proxy


Additional info:
Entering the container and installing nodejs from upstream [3] results in a successfully running proxy in the same image. I tried to verify in a standalone virtual machine using CentOS7 and KVM/QEMU but I was unable to reproduce the entire environment necessary for openshift-auth-image.



[1] https://github.com/openshift/openshift-ansible/tree/e29eab4e9813c46f4738a81e8698b001bd5981b9
[2] https://github.com/wozniakjan/openshift-auth-proxy/pull/1/files
[3] https://nodejs.org/en/download/package-manager/#enterprise-linux-and-fedora