dockerfiles | Docker files | Continuous Deployment library

Docker is still supported after 1811. It's all described on the SAP Help

You can either create your own recipe. Or make use of the ant command createPlatformImageStructure

It's been three weeks with no responses, but I just wanted to update with what I've found. In all cases where I've left "docker-compose up" running it eventually started.

At times it took 30 minutes, but it started every time.

Found the answer, just need to use the right tagged build that was built on the OS I need. In the dockerfile below, it is the tag 5.0-windowsservercore-ltsc2019. That was built with server core ltsc2019. Works in Container Service now.

https://hub.docker.com/_/microsoft-dotnet-sdk

Your first form uses a Docker multi-stage build. This has the specific advantage of not depending on any particular tools being installed on the host system. I'd consider this path:

• If you have a clustered CI system, it can be much easier to run the build in a container than to try to get the required tools manually installed on every worker system.
• If you're working in a compiled language (C, Go, Rust, ...), you can consistently use Linux containers even if developers use a different host OS.
• If you need to do extra work to set up the build system, like installing extra C development packages.
• If you need a really exact version of the language runtime for whatever reason.

The second path depends on having the toolchain available outside the container environment. That's not necessarily a problem; most front-end developers will probably have Node installed anyways. I'd consider this path:

• If the result of the build is extremely portable across environments (compiled HTML and Javascript; Java .jar files; interpreted text-format Python or Ruby scripting code).
• If there aren't big differences in different versions of the language runtime itself to produce the build. (Does your Webpack build do anything different on Node 8, 10, 12, or 14?)
• If the build system is something that's easy to install, or preinstalled in most host environments. (Most Linux and MacOS systems have Python, for example.)
• If a host build system can do incremental builds or otherwise run much faster than a clean-slate container build.

For what you show, with a simple front-end application that's compiled to static files, your second form is just fine. If you look at SO Java questions they almost universally COPY a prebuilt .jar file into an image, without including a build chain.

I dont know why but separating the install scripts into separate layers makes an image that works.

You can try to add buildContext parameter in the Docker task.

buildContext: Path to the build context

Here is a case you can refer to.

I found a working solution using nginx in https://github.com/adhavpavan/ContainerizingApps/tree/master/vue .

Dockerfile:

When you go to your browser and type http://localhost:3000 in, the browser makes a call to the client container, does an HTTP GET to retrieve the Javascript code, and the browser actually runs the code. This is a critical difference: any fetch calls or similar from your React application are running in the end user's browser, not inside Docker.

If a call is coming from the browser; it is part of your front-end application; or it is otherwise coming from outside Docker, then you need to use the host system's DNS name and the first published ports: number. If the browser and the containers are both running on the same system you can use localhost; if Docker is running in a VM (perhaps on the older Docker Toolbox setup) you need the VM's IP address.

I think you've identified your problem.

You're building your images on Apple's M1 chip, which is an ARM architecture. Fargate is probably running on the more common Intel x86-64 architecture. Images you build locally on your Mac aren't going to be able to run there.

The easiest solution is probably to have your images build automatically in Docker Hub (or use a Github Action to build them in Github).

I don't have all the details about how you're building and deploying your images, so it's possible I'm missing some details.