dockerfiles | Deep Learning Dockerfiles | Machine Learning library

I have been working with dockerfiles for a while now but today I was working on a small project and somehow on my VScode I typed # then CTRL+SPACE on the first line I got this syntax=docker/dockerfile:experimental.

I don't understand what this does and can't find documentation about it can somebody explain what's with that weird comment.

PS I found some people using this so it's not just some random comment generated by vscode I guess.

I have a few Dockerfiles that are dependant on a "pat" (Personal Access Token) file to be able to access a private nuget feed. I have taken some inspiration from somakdas to get this working.

To run my single Dockerfile I first create a "pat" file containing my token and build with docker build -f Services/User.API/Dockerfile -t userapi:dev --secret id=pat,src=pat .

This works as intended, but my issue is getting this to work using a docker-compose.yml file.

First I took a look at using docker-compose secrets, but it came to my attention that docker-compose secrets are access at runtime, not build-time. https://github.com/docker/compose/issues/6358

So now I'm trying to create a volume containing my pat file but I get cat: /pat: No such file or directory when the command RUN --mount=type=secret... is running. This may not be secure but it will only be running locally.

My Dockerfile

So I've got a application that consists of lets say 4 APIs and a Frontend, all saved on a monorepo. Everything is setup with docker, every Service has its own Dockerfile. The file structure would look something like this:

I have created through a Dockerfile a Docker image on which Ubuntu 18.04 and ROS Melodic are properly installed. Please note that I have not used the official ROS-Docker image for creating my Docker image.

The Docker container derived from this Docker image is working fine. However, every time I want to work with the container, I need to execute the following commands:

1. Terminal window:

docker run -d -it --name container_name docker_image; docker exec -it container_name bash

Then, after I am within the Docker container: roscore

1. Terminal window:

docker exec -it container_name bash

Then, after I am within the Docker container: rosrun ROS_PackageName PythonScript.py

Please note that through the above-mentioned Docker commands, both terminals operate in the same Docker container.

I find my way to start the PythonScript.py inefficient. Therefore, I would like to ask for the best practice to start "roscore" followed by "rosrun ...." on Docker Container startup.

In some Dockerfiles I see at the end of the file the command ENTRYPOINT as well as CMD. However, I do not know if these commands can help me to make my Docker container execute "roscore" followed by "rosrun" on container startup.

I've been creating a micro-frontend project and the glue (nginx) isn't working as expected.

My projects are structured as such:

Hi. I'm trying to come up with a solution to the following problem. I have one host machine running docker which should serve multiple independent web apps. Each app has a frontend (in angular) and backend (in NodeJS). I have nginx running on the same host machine (it has access to the SSL certificates, etc.).

See the diagram below describing what I'm trying to achieve. When a user requests app1.example.com the frontend files should be served. When app1.example.com/api/* is called, nginx passes the requst to the NodeJS backend.

Most of the articles I found online use containerized nginx and a single compose file. I want to have separate compose files for easy and independent updates - I just bump version numbers of web and API image in the .yml file and restart both containers with docker-compose. It won't affect other running containers or the nginx itself.

I'm having issues on how to serve the static sites. One obvious solution would be to include another nginx in each of the web containers (Dockerfile.web in the diagram) and serve the static files from the container that way. I find that a bit wasteful to have more instances of nginx running when there is already one on the host.

Another way I thought about is to mount or copy the static files from the container to the host (for example to /var/www) when the container is started.

Ideal solution would be that if the container is running the static files are accessible to the host nginx. When the container is stopped, static files become inaccessible and nginx can return 404 or a "maintenance page".

Do you think I'm approaching this the right way? I don't want to end up devising some non-standard niche solution but I would still like to have host-running nginx and independent updates with docker-compose for reasons described above.

Dockerfile.api

The ingress-nginx pod I have helm-installed into my EKS cluster is perpetually failing, its logs indicating the application cannot bind to 0.0.0.0:8443 (INADDR_ANY:8443). I have confirmed that 0.0.0.0:8443 is indeed already bound in the container, but bc I don't yet have root access to the container I've been unable to glean the culprit process/user.

I have created this issue on the kubernetes ingress-nginx project that I'm using, but also wanted to reach out to a wider SO community that might lend insights, solutions and troubleshooting suggestions for how to get past this hurdle.

Being a newcomer to both AWS/EKS and Kubernetes, it is likely that there is some environment configuration error causing this issue. For example, is it possible that this could be caused by a misconfigured AWS-ism such as the VPC (its Subnets or Security Groups)? Thank you in advance for your help!

The linked GitHub issue provides copious details about the Terraform-provisioned EKS environment as well as the Helm-installed deployment of ingress-nginx. Here are some key details:

1. The EKS cluster is configured to only use Fargate workers, and has 3 public and 3 private subnets, all 6 of which are made available to the cluster and each of its Fargate profiles.
2. It should also be noted that the cluster is new, and the ingress-nginx pod is the first attempt to deploy anything to the cluster, aside from kube-system items like coredns, which has been configured to run in Fargate. (which required manually removing the default ec2 annotation as described here)
3. There are 6 fargate profiles, but only 2 that are currently in use: coredns and ingress. These are dedicated to kube-system/kube-dns and ingress-nginx, respectively. Other than the selectors' namespaces and labels, there is nothing "custom" about the profile specification. It has been confirmed that the selectors are working, both for coredns and ingress. I.e. the ingress pods are scheduled to run, but failing.
4. The reason why ingress-nginx is using port 8443 is that I first ran into this Privilege Escalation issue whose workaround requires one to disable allowPrivilegeEscalation and change ports from privileged to unprivileged ones. I'm invoking helm install with the following values:

I've been trying to dockerize an angular application with python backend following this. I did all but for the docker-compose file, it says that my app.py is not present in the folder.

my python app is in 'C:\Users\Fra\Desktop\Fra\uni\Tesi\Progetto\backend'

and my angular app is in 'C:\Users\Fra\Desktop\Fra\uni\Tesi\Progetto\frontend'

i'm leaving you the dockerfiles, that should help

('Dockerfile' placed in the backend folder)

I'm trying to get to grips with both vite and docker so I apologise if I've made stupid mistakes.

I'm running into an issue with esbuild inside docker. I'm trying to get a dev setup going, so I want to mount my code in my containers so that changes should be reflected in real time.

Previously I used Dockerfiles which copied /frontend and /backend into their respective containers and that worked, I had my web and api containers running and happily talking to each other. However, it meant it didn't pick up any code changes so it wasn't suitable for development.

So I've switched to volume mounts in the hope that I can get my dockerized apps to hot reload, but hit this error instead.

Here's my docker-compose.yml