Azure Custom Vision AI : From training to deploying the container export on the Azure Kuberenetes Service (AKS)

Introduction

Todays post will be the backend tour of “Frietjes-of-Niet” (translated from Dutch : “French-Frites-or-Not?”). A big part of the mission of Azure is about democratizing technology so it becomes accessible to organizations in order for them to achieve more. AI (Artificial Intelligence) is a key part of that vision.

 

End-to-End Flow

What will be the flow for today?

• We’ll train a model to recognize fries
• Next we’ll be exporting that model to be used as a container
• Afterwards we’ll build that container
• To end with deploying (and testing) it onto AKS

Sound cool? Let’s get to it..

Taking a glance at Rancher’s ability to manage the Azure Kubernetes Service (AKS)

Introduction

Pfew, it’s odd to admit that it has been a while since I’ve posted about Rancher. Though today is as good a day as any to pick up that thread… So today we’ll go through give or take the same objective as in the past, where we’ll notice that the integration has improved significantly with the arrival of AKS! Let’s get today’s post underway and deploy AKS from our Rancher control plane.

Preparation

Before the below started, I already had the following things ready ;

• A service principal with appropriate rights
• Installed Rancher 2.0

Replatforming Azure Functions into an Azure Functions Container

Introduction

A while ago I talked about  “Faas/Serverless” in relation to vendor lock-in. Today we’ll be continuing in that road, where we’ll be doing a small proof-of-concept (PoC). In this PoC, we’ll be replatforming existing Azure Functions code into an Azure Functions container!

 

Things to know

Since Azure Functions 2.0 (in preview at the time of writing this post), you are able to leverage containers. Though be aware that there are several known issues. Do check them out first before embarking on your journey!

 

Testdriving 2.0

So first, we’ll start off with testing the Azure Functions Core Tools!  If you’re looking to follow this guide, be sure to have the Azure Functions Core Tools installed, which also depends on .NET Core 2.0 and Nodejs. Once you have those installed, do a “func –help”, and you’ll see what capabilities are at hand…

From Github to ACI – A tale how to use Visual Studio Team Services & Azure Container Registry for Container CI/CD

Introduction

Today’s post is conceptually a rather simple one… Let’s see how we can go from this ;

To here ;

By using a CI/CD pipeline.

 

Flow of the day

What will we be doing today?

• Kick-off a VSTS build once a change has been made to our Github repo
  • Build a container via VSTS
  • Publish the container to an ACR (Azure Container Registry)
• Kick-off a VSTS release once the build succeeded
  • Use an ARM template to deploy an ACI (Azure Container Instance) with our docker container underneath

Sound cool? Let’s get to it!

Azure Service Fabric : Deploying your first container…

Introduction

In the past I’ve already done several posts about containers. This by using various orchestrators & workflow management tools. Today’s post will be about deploying a Linux container with Service Fabric… The main goals is to provide you with the look & feel of the initial steps. In a future posts, I’ll delve into the more advanced stuff (like data persistence & inter-container connectivity).

Issue : Failed to pull image : rpc error: code = 2 desc = unknown blob

Introduction

When you are deploying an image, which is hosted on a private registry, to a kubernetes (k8s) cluster with windows nodes… Then you might get the following error ;

Failed to pull image “kvaes.azurecr.io/kvaes2017:v1“: rpc error: code = 2 desc = unknown blob

Error syncing pod, skipping: failed to “StartContainer” for “private-reg-container” with ErrImagePull: “rpc error: code = 2 desc = unknown blob”

 

Context

So what did my setup look like?

• Orchestrator : Kubernetes for Windows (Azure Container Service)
• Registry : Private (Azure Container Registry)
• Image : Windows Nano Based

 

Technical Check

Let’s deploy two pods…

The first I’ll deploy via yaml, which is basically the example from the kubernetes docs on pulling an image from a private repo…

Now the second one is an adaptation of the example flow from the Azure Container Service documentation ;

Now let’s see how that one went…

The first one failed, and the second one passed! What was the difference?

 "nodeSelector": {
    "beta.kubernetes.io/os": "windows"
    }

Apparently this one forces the switch to “windows container mode” (or something like that…). As it seems very similar to the following thread…

 

TL;DR

When deploying windows containers to a kubernetes cluster. be sure to the set the “nodeSelector” or you might end up with errors on pulling the image.

How to try out the experimental windows 2016 support in the Rancher 1.3.0 release candidate?

Introduction

Yesterday Rancher commented on my github request for windows support ;

Tested with rancher-server version – v1.3.0-rc1 with catalog “library” set to vnext branch in https://github.com/rancher/rancher-catalog

Able to add “Windows Server 2016 Standard Evaluation” hosts successfully to rancher environment with orchestration set to “windows”.

Able to launch containers in “nat” network and “transparent” network.

@kvaes , Windows 2016 support will be available as experimental feature in rancher-server 1.3.0 release.

Great news! Let’s take it out for a spin… 😀

 

Rancher Host

Installing the host(s) is the same as any other time…  Though the host will still be a Linux machine off course ;

sudo docker run -d –restart=unless-stopped -p 8080:8080 rancher/server:v1.3.0-rc1

Though notice that I specified the v1.3.0-rc1 tag… And let the system do its magic!

(Update : For the stable, release you can omit the -rc1 part!)

Note ; Be aware that this is an early release candidate. Do not use this for your production! There is for instance a bug with the GUI, where the “Auto”-theme is malfunctioning. So switch to light or dark to get that one fixed. 😉