Updating mnist example to fix minio compatibility (#108)

* Updating mnist example to fix minio compatibility

* Changing default sa user for ksonnet entrypoint

* Updating mnist example based on pr feedback.

mnist/Dockerfile.ksonnet

@@ -2,7 +2,7 @@
 FROM ubuntu:16.04
 
 ENV KUBECTL_VERSION v1.9.2
-ENV KSONNET_VERSION 0.8.0
+ENV KSONNET_VERSION 0.10.1
 
 RUN apt-get update
 RUN apt-get -y install curl 

mnist/Dockerfile.model

@@ -1,5 +1,5 @@
 #This container contains your model and any helper scripts specific to your model.
-FROM tensorflow/tensorflow:1.5.1
+FROM tensorflow/tensorflow:1.7.0
 
 ADD model.py /opt/model.py
 RUN chmod +x /opt/model.py

mnist/README.md

@@ -71,9 +71,9 @@ With our code ready, we will now build/push the docker image.
 
 ```
 DOCKER_BASE_URL=docker.io/elsonrodriguez # Put your docker registry here
-docker build . --no-cache  -f Dockerfile.model -t ${DOCKER_BASE_URL}/mytfmodel:1.0
+docker build . --no-cache  -f Dockerfile.model -t ${DOCKER_BASE_URL}/mytfmodel:1.7
 
-docker push ${DOCKER_BASE_URL}/mytfmodel:1.0
+docker push ${DOCKER_BASE_URL}/mytfmodel:1.7
 ```
 
 ## Preparing your Kubernetes Cluster
@@ -94,7 +94,7 @@ cd ${APP_NAME}
 
 ks registry add kubeflow github.com/kubeflow/kubeflow/tree/master/kubeflow
 
-ks pkg install kubeflow/core@1a6fc9d0e19e456b784ba1c23c03ec47648819d0
+ks pkg install kubeflow/core@v0.1.2
 ks pkg install kubeflow/argo@8d617d68b707d52a5906d38b235e04e540f2fcf7
 
 # Deploy TF Operator and Argo
@@ -133,15 +133,19 @@ $ argo list
 NAME   STATUS   AGE   DURATION
 ```
 
-### Creating secrets for our workflow
-For fetching and uploading data, our workflow requires S3 credentials. These credentials will be provided as kubernetes secrets:
+### Creating secrets for our workflow and setting S3 variables.
+
+For fetching and uploading data, our workflow requires S3 credentials and variables. These credentials will be provided as kubernetes secrets, and the variables will be passed into the workflow. Modify the below values to suit your environment.
 
 ```
-export S3_ENDPOINT=s3.us-west-2.amazonaws.com
-export AWS_ENDPOINT_URL=https://${S3_ENDPOINT}
+export S3_ENDPOINT=s3.us-west-2.amazonaws.com  #replace with your s3 endpoint in a host:port format, e.g. minio:9000
+export AWS_ENDPOINT_URL=https://${S3_ENDPOINT} #use http instead of https for default minio installs
 export AWS_ACCESS_KEY_ID=xxxxx
 export AWS_SECRET_ACCESS_KEY=xxxxx
+export AWS_REGION=us-west-2
 export BUCKET_NAME=mybucket
+export S3_USE_HTTPS=1 #set to 0 for default minio installs
+export S3_VERIFY_SSL=1 #set to 0 for defaul minio installs
 
 kubectl create secret generic aws-creds --from-literal=awsAccessKeyID=${AWS_ACCESS_KEY_ID} \
  --from-literal=awsSecretAccessKey=${AWS_SECRET_ACCESS_KEY}
@@ -174,9 +178,8 @@ First we need to set a few variables in our workflow. Make sure to set your dock
 DOCKER_BASE_URL=docker.io/elsonrodriguez # Put your docker registry here
 export S3_DATA_URL=s3://${BUCKET_NAME}/data/mnist/
 export S3_TRAIN_BASE_URL=s3://${BUCKET_NAME}/models
-export AWS_REGION=us-west-2
 export JOB_NAME=myjob-$(uuidgen  | cut -c -5 | tr '[:upper:]' '[:lower:]')
-export TF_MODEL_IMAGE=${DOCKER_BASE_URL}/mytfmodel:1.0
+export TF_MODEL_IMAGE=${DOCKER_BASE_URL}/mytfmodel:1.7
 export TF_WORKER=3
 export MODEL_TRAIN_STEPS=200
 ```
@@ -194,6 +197,8 @@ argo submit model-train.yaml -n ${NAMESPACE} --serviceaccount tf-user \
     -p job-name=${JOB_NAME} \
     -p tf-worker=${TF_WORKER} \
     -p model-train-steps=${MODEL_TRAIN_STEPS} \
+    -p s3-use-https=${S3_USE_HTTPS} \
+    -p s3-verify-ssl=${S3_VERIFY_SSL} \
     -p namespace=${NAMESPACE}
 ```
 

mnist/ksonnet-entrypoint.sh

@@ -1,6 +1,7 @@
 #!/bin/bash
 
-SERVICE_ACCOUNT=${SERVICE_ACCOUNT:-default}
+#FIXME this is set to tf-user by default in order to work around lack of Serviceaccount setting in argo.
+SERVICE_ACCOUNT=${SERVICE_ACCOUNT:-tf-user}
 
 create-kubeconfig ${SERVICE_ACCOUNT} > kubeconfig.tmp
 cp kubeconfig.tmp ~/.kube/config

mnist/model-deploy.yaml

@@ -115,7 +115,7 @@ spec:
       - name: model-serving-ks-tag
       - name: model-name
     script:
-      image: elsonrodriguez/ksonnet:0.8.0-test6
+      image: elsonrodriguez/ksonnet:0.10.1
       command: ["/ksonnet-entrypoint.sh"]
       source: |
         ks init my-model-server
@@ -135,8 +135,8 @@ spec:
         ks param set {{inputs.parameters.model-name}} s3_secret_secretaccesskey_key_name awsSecretAccessKey
         ks param set {{inputs.parameters.model-name}} s3_aws_region {{inputs.parameters.aws-region}}
         ks param set {{inputs.parameters.model-name}} s3_endpoint {{inputs.parameters.s3-endpoint}}
-        ks param set {{inputs.parameters.model-name}} s3_use_https \'{{inputs.parameters.s3-use-https}}\'
-        ks param set {{inputs.parameters.model-name}} s3_verify_ssl \'{{inputs.parameters.s3-verify-ssl}}\'
+        ks param set {{inputs.parameters.model-name}} s3_use_https {{inputs.parameters.s3-use-https}} --as-string
+        ks param set {{inputs.parameters.model-name}} s3_verify_ssl {{inputs.parameters.s3-verify-ssl}} --as-string
         ks apply default -c {{inputs.parameters.model-name}}
       #FIXME This doesn't actually work in the current version of argo. We're using a default of `tf-user` in the container entrypoint for now.
       env:

mnist/model-train.yaml

@@ -15,13 +15,13 @@ spec:
     - name: tf-ps # number of tf parameter servers
       value: 2
     - name: tf-model-image
-      value: elsonrodriguez/mytfmodel:1.0
-    - name: tf-serving-image #this image is a mirror of a private kubeflow-ci image
-      value: elsonrodriguez/model-server:1.0
+      value: elsonrodriguez/mytfmodel:1.7
+    - name: tf-serving-image #FIXME this image is a mirror of a private kubeflow-ci image, once we're building images swap this out. https://github.com/kubeflow/kubeflow/blob/dcf4adfe2dd1cec243647f3dd05d7c26246fddb1/components/k8s-model-server/images/Dockerfile.cpu
+      value: elsonrodriguez/model-server:1.6
     - name: tf-tensorboard-image
-      value: tensorflow/tensorflow:1.5.1
+      value: tensorflow/tensorflow:1.7.0
     - name: ks-image
-      value: elsonrodriguez/ksonnet:0.8.0-test7
+      value: elsonrodriguez/ksonnet:0.10.1
     - name: model-name
       value: mnist
     - name: model-hidden-units
@@ -354,8 +354,8 @@ spec:
         ks param set {{workflow.parameters.model-name}} s3_secret_secretaccesskey_key_name awsSecretAccessKey
         ks param set {{workflow.parameters.model-name}} s3_aws_region {{workflow.parameters.aws-region}}
         ks param set {{workflow.parameters.model-name}} s3_endpoint {{workflow.parameters.s3-endpoint}}
-        ks param set {{workflow.parameters.model-name}} s3_use_https \'{{workflow.parameters.s3-use-https}}\'
-        ks param set {{workflow.parameters.model-name}} s3_verify_ssl \'{{workflow.parameters.s3-verify-ssl}}\'
+        ks param set {{workflow.parameters.model-name}} s3_use_https {{workflow.parameters.s3-use-https}} --as-string
+        ks param set {{workflow.parameters.model-name}} s3_verify_ssl {{workflow.parameters.s3-verify-ssl}} --as-string
         ks apply default -c {{workflow.parameters.model-name}}
       #FIXME This doesn't actually work in the current version of argo. We're using a default of `tf-user` in the container entrypoint for now.
       env: