Closed hugy718 closed 2 years ago
It can be fixed by using a newer version of npm in the docker image of ow-utils. Configuring the helm charts in openwhisk-deploy-kube
to use the updated image, I have successfully deployed it as before.
Thank you for reporting this. Why don't you contribute the change to fix the issue as you already clearly analyzed the culprit?
Thank you for reporting this. Why don't you contribute the change to fix the issue as you already clearly analyzed the culprit?
Sure. I'd love to :). Opened a PR here #5261. As for the contribution guideline, I have submitted signed Apache CLA, waiting for the reply.
Update of image tags at openwhisk-deploy-kube helm charts has been merged.
@hugy718 Nice! Thank you for your contribution.
Environment details:
local deployment on self-built kubernetes cluster Kubernetes version: 1.23.3 ubuntu version: 20.04.4 LTS (kernel: Linux 5.13.0-51-generic)
Steps to reproduce the issue:
I had labeled my node as invoker.
Provide the expected results and outputs:
owdev-install-packages pod has status Completed
Provide the actual results and outputs:
owdev-install-packages pod has status Completed
Additional information you deem important:
This happened last week when re-deploy openwhisk. After doing some investigation, I found the problem is the npm package install for alarm package. I inspected the log of the failing pods and successfully reproduced issue by interactively launching the container. The
owdev-install-package
pod uses container image by openwhisk/tols/ow-utils/Dockerfile. The npm installed by apt has the version:Install fails for the openwhisk-deploy-kube/helm/openwhisk/configMapFiles/installPackages/myTask.sh at L104
Looking at the log of failed pod, problem is the npm install command at L95 in openwhisk-package-alarms/installCatalog.sh
The output:
That seems to be an issue of using old version of npm. A recent issue under npm/cli here also report their application suddenly broke last week.