"getting-started-gke" installer fails

[mmatiaschek]

I had these issues on google cloud shell so i tried with a fresh install on Bash on Ubuntu on Windows, Ubuntu 16.04, and got the same problems.

1. install and initialize gcloud
2. install ksonnet v0.11.0 from github
3. install kubeflow

First issue with the kubeflow documentation/installer was that the tmp folder was not found, my workaround was to mv /tmp/kubeflow-0.2.1/ /tmp/kubeflow-v0.2.1 and start again.

+ mv /tmp/kubeflow-v0.2.1 /mnt/c/Users/mmatiaschek/cmder/kubeflow_repo
mv: Aufruf von stat für '/tmp/kubeflow-v0.2.1' nicht möglich: Datei oder Verzeichnis nicht gefunden
$ mv /tmp/kubeflow-0.2.1/ /tmp/kubeflow-v0.2.1
$ curl https://raw.githubusercontent.com/kubeflow/kubeflow/v${KUBEFLOW_VERSION}/scripts/gke/deploy.sh | bash

The next error was not recoverable for me, any help would be greatly appreciated!!

kubeflow exists
+ gcloud deployment-manager --project=*********** deployments update kubeflow --config=cluster-kubeflow.yaml
The fingerprint of the deployment is ***************==
Waiting for update [operation-1531486959504-570e118330a81-c859d6b5-78eef300]...
....failed.
ERROR: (gcloud.deployment-manager.deployments.update) Error in Operation [operation-1531486959504-570e118330a81-c859d6b5-78eef300]: errors:
- code: CONDITION_NOT_MET
  location: /deployments/kubeflow/resources/kubeflow-gpu-pool-v1->$.properties
  message: '"": domain: validation; keyword: properties; message: required property(ies)
    not found; missing: ["nodePoolId"]; required: ["clusterId","nodePoolId","zone"]'

[mmatiaschek]

it seem like kubectl 1.11.0 is buggy and responsible for the failed deployment.
1.10.5 seems to work.
A fix for the tmp issue would be nice.

[ankushagarwal]

/assign me

[k8s-ci-robot]

@ankushagarwal: GitHub didn't allow me to assign the following users: me.

Note that only kubeflow members and repo collaborators can be assigned.

In response to this:

/assign me

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.

[ankushagarwal]

/assign ankushagarwal

[ankushagarwal]

@mmatiaschek : Do you know what the problem with ksonnet was? I've not had issues with 1.11.0

Regarding the gcloud deployment-manager error. Can you first delete the deployment and it's config directory and try again?

DEPLOYMENT_NAME=kubeflow
gcloud deployment-manager deployments delete --quiet \
 ${DEPLOYMENT_NAME} --project=${PROJECT}

KUBEFLOW_DM_DIR="${DEPLOYMENT_NAME}_deployment_manager_configs"
rm -rf ${KUBEFLOW_DM_DIR}

[jlewi]

@mmatiaschek The tmp issue should be fixed in 0.2.2-rc.0

[jlewi]

0.2.2-rc.0 has been promoted to 0.2.2.

[mmatiaschek]

Thank you, it works... i still have some problems though, i will open another issue