Fix: Root User Required for Operation
Summary
After experimentation, liveness must run as root for any CSI interface to properly attach volumes to hosts.
Steps to reproduce
Deploy any CSI implementation (EBS CSI driver or other, all of which require this image) and see that the pods do not correctly deploy.
What is the current bug behavior?
Pods running with the livenessprobe will fail to fully come up properly. They will not be able to correctly.
What is the expected correct behavior?
EBS CSI driver should be able to come up and allow for volume attachment.
Relevant logs and/or screenshots
Warning Unhealthy 2m18s (x3 over 2m38s) kubelet, ip-146-69-200-77.us-gov-east-1.compute.internal Readiness probe failed: Get "http://10.42.4.50:9808/healthz": dial tcp 10.42.4.50:9808: connect: connection refused
Warning Unhealthy 2m17s (x3 over 2m37s) kubelet, ip-146-69-200-77.us-gov-east-1.compute.internal Liveness probe failed: Get "http://10.42.4.50:9808/healthz": dial tcp 10.42.4.50:9808: connect: connection refused
Possible fixes
Run as root user was the only fix we found that allowed it to work.
Defintion of Done
-
Bug has been identified and corrected within the container
Edited by Joshua Carnes