svelte-manga-api

it has docker-compose for development and deploying. it also configured with .devcontainer så you can go into the container and work on it from there. another way is it has python environment using the commands in run.sh or run.bat.

a dev environment for python

Environment Variables

To control GPU usage, you can set environment variables before running docker-compose.

Without GPU

export GPU_COUNT=0
docker-compose --profile dev up
# or for production
docker-compose --profile prod up

With GPU

export GPU_COUNT=1
docker-compose --profile dev up
# or for production
docker-compose --profile prod up

Explanation

Environment Variables: The GPU_COUNT environment variable determines the number of GPUs to reserve. When set to 0, no GPU is reserved, effectively disabling GPU support. Conditional Logic: Docker Compose does not support direct conditional logic, so using environment variables allows you to control resource allocation. Deployment Configuration: The deploy.resources.reservations.devices section is included in both development and production services. It uses the GPU_COUNT environment variable to specify the number of GPUs. If GPU_COUNT is 0, it will effectively skip GPU reservation. This setup allows you to control GPU usage by simply setting an environment variable, providing flexibility without modifying the Docker Compose file each time.

(Optional) everything after this is optional (this has changed to include ddns)

deployment

there are two ways. one is simpler using azure container. it just deploys the app to your azure account. the 2nd option is more heavy duty but comes with more. while the first is just a container. the second one is a full on server. the server is a free tier ec2(if you only have one server). to choose, its the one in the workflows folder thats active. for azure have the deploy-azure.yml and svelte-manga-api.yml in the folder and the other two in the extra_workflows folder. for terraform have the deploy-terraform.yml, detroy.yml and svelte-manga-api.yml in the workflow folder and deploy-azure.yml in the extra_workflows folder for both you need docker hub account

docker hub

go to https://hub.docker.com/settings/security and create and copy the token. together with your username go to (your repository)https://github.com/valiantlynx/svelte-manga-api/settings/secrets/actions and make these repository secrets

DOCKER_HUB_USERNAME = your username
DOCKER_HUB_PASSWORD = your docker access token

azure

its just a github action that uses your dockerfile. got to azure and search for web app for containers choose this:

and then these details: basic container

in the docker section choose what your docker details will be in docker hub. (username/container-name):

leave the rest as they are(yes even deployment can be done later) and review and create. after the resource is built go to it then deployment center and choose this config(try to relly on azures auto fill to fill this form if possible): config stuff to remove that error so we can save. we need to set the variables in azure got to configurations and edit the two docker stuff that are empty: env in azure go to general settings tab and turn on SCM basic auth. save it and go back to deployment center. your might have to redo some of the former step. then save

we need some env variables for authentication in git hub action. go back to overview and on the top is the option Download publish profile. copy the whole file and lets go to github

first the login stuff. go to (your repository)https://github.com/valiantlynx/svelte-manga-api/settings/secrets/actions and make these repository secrets(you might see two env variables azure made delete them) if your planing to use azure

AZUREAPPSERVICE_PUBLISHPROFILE = the thing you downloaded and copied

now the next time you commit to there is a succesfull pull or push to the main branch and the docker image is on docker hub then azure will redeploy

to see the app azure provides a free url something like https://svelte-manga-api.azurewebsites.net/ go there and you might have to wait like 10 min the first time around. or make a commit to main again a couple times

terrafrom

this does alot for and is simpler to set up.

what you get

simply put its IaC with terraform, config with ansible, monitoring with prometheus-grafana and service-check with uptime-kuma and container management with portainer. it all automated as well

setup

you need an aws account. in it make an s3 bucket called ´svelte-manga-api´ it can be anything it just has to match with the one in terraform/provider.tf aws. and lastly access keys, you can get thenm in IAM. just create access key cause we need both the key and the secret accesskey

the login stuff. go to (your repository)https://github.com/valiantlynx/svelte-manga-api/settings/secrets/actions and make these repository secrets if your planing to use azure

AWS_ACCESS_KEY = that you copied
AWS_SECRET_ACCESS_KEY = the secret you copied

this devops sets up everything even ssl. so edit the somain ports and images to your needs

all the files you need to edit are in andible/roles/docker_deploy/files unfortunately for this you need some knowledge(basic is enough) in docker, nginx and certbot. specifically, andible/roles/docker_deploy/files/docker/docker-compose.yml - for ports and which domain you want ssl for and containers. andible/roles/docker_deploy/files/docker/prometheus.yml - together with the one above for your monitoring needs andible/roles/docker_deploy/files/nginx/http.conf - both this and below need to be edited on each edit. for http edits. and where the containers are accessed. reverse proxy andible/roles/docker_deploy/files/nginx/https.conf - both this and above need to be edited on each edit. for https edits. and where the containers are accessed. reverse proxy

you can look at my git is some of my repos logs to get an idea of how i do or contact me for help.

now the next time you commit to there is a succesfull pull or push to the main branch and the docker image is on docker hub then terraform with deploy the infra and ansible will configure everything all the way to ssl. this means the first time its building it might fail. that cause the domains that certbot is trying to get ssl for might not be pointing to the newly created ec2. you need to go to your dns and point it to the correct ip.