add `@primate/deploy`

[terrablue]

Different providers

• Repopack (default)
• Cloudflare workers
• Cloudflare pages
• Amazon
• Google
• Netlify
• Vercel
• Deno deploy
• ?

[ralyodio]

just a standalone server would be good too. no need to only support paid cloud offerings.

[terrablue]

@ralyodio how would such a deployment work?

[ralyodio]

Sync with a ssh

[ralyodio]

look at ./bin/deploy.sh and ./bin/post-deploy.sh in fastest-engineer-web -- that's how I deploy to a server.

There is some setup on the server that needs to happen first (nginx, systemd and directory paths)

[ralyodio]

// deploy.sh

#!/bin/bash

. $HOME/.bashrc
. .env
. .env.local

args=(-azvP --delete --exclude=node_modules --exclude=.idea --exclude=data --exclude=static/_posts)
hosts=($HOST_DOMAIN) # tornado lightning thunder tundra jefferson
dry=() #add --dry-run to enable testing
user=$HOST_USER
name=$HOST_PATH
project=$HOST_PROJECT

if [ -z "$name" ] || [ -z "$project" ]; then
    echo "One or both directories do not exist"
    exit 1
fi

result=$(ssh $user@$HOST_DOMAIN "[ -d ~/www/$name ] && [ -d ~/www/$name/$project ] && echo 'exists' || echo 'does not exist'")

if [ "$result" == "exists" ]; then
    # If both directories exist, run your command here
    echo "Both directories exist"

    for host in "${hosts[@]}"
    do
      echo ""
      date
      echo "---------------------"
      echo "syncing ${host}"
      echo "---------------------"
      rsync ${dry[@]} ${args[@]} ./ ${user}@${host}:www/${name}/${project}
      ssh -t ${user}@${host} \$HOME/www/${name}/${project}/bin/post-deploy.sh
    done

else
    echo "One or both directories do not exist"
fi

version=$(jq -r .version package.json)
say "$HOST_PROJECT is live!"
exit

//post-deploy.sh

#!/bin/bash

cd "$(dirname "$0")/.."
. $HOME/.bashrc
. .env
. .env.local

host=$HOST_DOMAIN
name=$HOST_PATH
project=$HOST_PROJECT

echo "current name: $name"

if [ -d "$HOME/www/${name}/${project}" ]; then
    # If both directories exist, run your command here
    echo "Both directories exist"
    cd $HOME/www/${name}/${project}
    nvm install v20
    node -v
    pnpm -v
    rm -f package-lock.json
    rm -f pnpm-lock.yaml
    rm -rf ./node_modules
        npm i -g pnpm
    #pnpm cache clean --force
    #pnpm cache verify
    pnpm i
        # curl --proto '=https' --tlsv1.2 -sSf https://install.surrealdb.com | sh -s -- --nightl
    sudo systemctl stop ${META_SERVICE}
    sudo systemctl stop surrealdb
    # surreal upgrade --nightly
    surreal upgrade
    sudo /etc/init.d/nginx reload
    sudo systemctl daemon-reload
    sudo systemctl start ${META_SERVICE}
    sudo systemctl start surrealdb
    # run migrations
    ./migrations/users.sh
    # ./migrations/comments.sh
    ./migrations/links.sh
    ./migrations/apikeys.sh
    ./migrations/products.sh
    # ./migrations/send_email.sh
    ./migrations/nostrusers.sh
    ./migrations/affiliates.sh
    ./migrations/appointments.sh
    ./migrations/payouts.sh
    ./migrations/waitlist.sh

  else
    echo "One or both directories do not exist"
fi

[ralyodio]

in my opinion if you gave us the ability to write our own deploy script and post deploy scripts (which runs on the server) that would be good enough for custom deploys.

[terrablue]

@ralyodio noted, I think that's the best way forward

[ralyodio]

maybe a ./bin/setup.sh too which would be ran once. I manually setup certs, symlink conf files and shit. I should script that.

[ralyodio]

Here's an example setup.sh script:

#!/bin/bash

. $HOME/.bashrc
. .env
. .env.local

# Check if the user provided the domain name and project name
if [ -z "$1" ] || [ -z "$2" ]; then
  echo "Usage: $0 <domain_name> <project_name>"
  exit 1
fi

# Variables
DOMAIN_NAME="$1"
PROJECT_NAME="$2"
TARGET_PATH="~/www/$DOMAIN_NAME/$PROJECT_NAME"
NGINX_CONF="$TARGET_PATH/etc/$DOMAIN_NAME.conf"
SYSTEMD_SERVICE="$TARGET_PATH/etc/$PROJECT_NAME.service"
NGINX_SYMLINK="/etc/nginx/sites-available/$DOMAIN_NAME.com.conf"
SYSTEMD_SYMLINK="/etc/systemd/system/$PROJECT_NAME.service"
user=$HOST_USER

# Create the directory if it does not exist
ssh $user@$HOST_DOMAIN "mkdir -p $TARGET_PATH"

# Deploy the code
rsync -azvP --delete --exclude=node_modules --exclude=.idea --exclude=data --exclude=static/_posts ./ ${user}@${HOST_DOMAIN}:${TARGET_PATH}

# Symlink the Nginx configuration
ssh $user@$HOST_DOMAIN "if [ -f $NGINX_CONF ]; then sudo ln -sf $NGINX_CONF $NGINX_SYMLINK && sudo ln -sf $NGINX_SYMLINK /etc/nginx/sites-enabled/; else echo 'Nginx configuration file not found at $NGINX_CONF'; exit 1; fi"

# Symlink the systemd service
ssh $user@$HOST_DOMAIN "if [ -f $SYSTEMD_SERVICE ]; then sudo ln -sf $SYSTEMD_SERVICE $SYSTEMD_SYMLINK; else echo 'Systemd service file not found at $SYSTEMD_SERVICE'; exit 1; fi"

# Generate Let's Encrypt certificates for the domain
ssh $user@$HOST_DOMAIN "sudo certbot certonly --nginx -d $DOMAIN_NAME"
ssh $user@$HOST_DOMAIN "sudo certbot certonly --nginx -d www.${DOMAIN_NAME}"

# Enable and start the systemd service
ssh $user@$HOST_DOMAIN "sudo systemctl enable $(basename $SYSTEMD_SERVICE) --now"

ssh -t ${user}@${HOST_DOMAIN} ${TARGET_PATH}/bin/post-deploy.sh

echo "Setup completed successfully!"

version=$(jq -r .version package.json)
say "$HOST_PROJECT is setup!"
exit

[ralyodio]

it could be smarter and write out .env.local based on questions you answer.

[ralyodio]

I also decided to support multiple environments. basically you just symlink .env.<envname> to .env.local and add those custom files to .gitignore

this makes it easiser for staging.example.com vs. example.com which a lot of clients want