-
Notifications
You must be signed in to change notification settings - Fork 835
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
configurable prepack images #1118
configurable prepack images #1118
Conversation
I think we need access to the configmap entries in the webhook, since there's a prepackaged model server scenario where some parts of the container spec might be set (e.g. resources) without an image. Then we have to fill in the image. But we don't currently have access to the configmap in the webhook |
I tried going the environment variable route by setting this in the manager.yaml:
And then loading the json from that env var string. But then the docker-build blows up, presumably with kubebuilder:
Might be best to try to get access to the client. |
Actually that error seems to have been caused by commenting out a test. If I delete the test then I don't get that. Would be better to make the test actually work but I guess for that it will be necessary to mock the configmap lookup (whether the env var or client way). |
Turns out I don't need to pass through the client. So what I do need to do is work through the other uses for the prepack image code and put the test back in. |
Seems I didn't need to mock the client call either. It just requires that the configmap be installed in kind before running the tests. /ok-to-test |
Tensorflow now seems to work but for SKLEARN the initContainer seems to have gone missing. |
/test integration |
Hmm, I managed to run tests locally by installing to the cluster before running the tests. In the cluster it's not doing it that way though. |
An example of using a configmap in a go test - https://github.com/kubeflow/kfserving/blob/master/pkg/controller/inferenceservice/controller_test.go |
Got the unit tests passing again, now just need to remove the unnecessary logging and double-check the e2e tests. |
/test integration |
@ryandawsonuk: you cannot LGTM your own PR. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: ryandawsonuk The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
/test integration |
@ryandawsonuk: The following test failed, say
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here. |
Not sure why the integration tests are failing here as they pass locally. |
So within a local kind cluster everything works. I've run e2e and tested manually. I've also built an image from the branch and run it in a GKE cluster and tested that manually and all works. I'm stumped as to why the e2e tests in the jx cluster are failing. They're actually all failing, even the ones not related to prepackaged model servers. (The only ones passing are the ones where resource creation is meant to fail.) |
Hmm, now it seems to have been automatically merged without the e2e tests passing. I have a theory now about what the problem was though. I think it's because we were on |
for #959