[WIP] Update to CustomResourceDefinition instead of ThirdPartyResource. #20
Conversation
* Delete WaitForTPR; CRD has new methods for handling the waiting. * Create the CRD resource.
pkg/spec/tf_job.go
Outdated
| TPRGroup = "mlkube.io" | ||
| TPRVersion = "v1beta1" | ||
| TPRDescription = "TensorFlow training job" | ||
| CRDKind = "tf-job" |
There was a problem hiding this comment.
Do you still use this anywhere? If this means Kind in the k8s API sense of Group-Version-Kind, then it should be camel case.
There was a problem hiding this comment.
Thanks; you're right we don't use it anymore. The only thing I specify is the PluralKind which I specify as "tfjobs". This ends up producing
spec:
group: mlkube.io
names:
kind: TfJob
listKind: TfJobList
plural: tfjobs
singular: tfjob
scope: Namespaced
version: v1beta1
status:
acceptedNames:
kind: TfJob
listKind: TfJobList
plural: tfjobs
singular: tfjob
This looks correct; i.e. the plural name are all lower case no hypens which follows the CronTab example in the docs
https://kubernetes.io/docs/tasks/access-kubernetes-api/extend-api-custom-resource-definitions/
* Fix bad import that was breaking the build.
* Looks like E2E test is having a problem creating the TFJob. * Tag the image with the git commit to ensure unique names. * Update the helm test to get the image from a value.
* We need to set Kind and ApiVersion when creating the request otherwise the ApiServer will reject the request with a 400 error. * Update tpr_util to use the helper methods to construct the URI as opposed to constructing the URI manually. * Add some logging to the E2E test to make it easier to debug.
…ce to update. * Update readme; replace references to TPR with CRD. * E2E test is still failing because the PS and Master are crashing because we try to run the operator as the program.
… the operator. * Attempt to fix travis by removing some of the vendor libraries.
Signed-off-by: Syulin7 <[email protected]>
No description provided.