any advice on how to run the migrations on the prod database when it's inside a VPC in GCP? I'd really really really like to avoid connecting to the VPC from GH actions to run that, there must be a better way, but upon google searching not a lot of 'meaty answers' have popped up so thought I'd ask here

I am/was facing the same issue, but i'm in AWS with ECS Fargate. I'm using nodejs as backend, so what i'm doing at the moment is using a process manager (pm2), that has this config:

app-cluster.config

module.exports = {
    apps: [
        {
            name: 'migrator',
            script: './migrationManager/index.js',
            instances: 1,
        },
        {
            name: 'app_1',
            script: './index.js',
            exec_mode: 'cluster',
            instances: 'max', // leave this so that it can take advantage from all computing power available.
        },
    ],
};

then my container simply does:

Dockerfile

CMD ['yarn', 'start:prod']

package.json

"start:prod": "pm2-runtime app-cluster.config.js"

That will start 2 processes: • 1 for an API • 1 for running migrations The migrations are executed in another child_process running in the background using

spawnSync

and after everything finished, using the pm2 API i can connect to the running instance of pm2 and stop the process, as I only need it to run migrations and nothing else... A better way to do this tho, would be to instead of running migrations from the script, just tell the script to put the migrations in a queue as a job (could be redis or whatever), and then having a second version of your API or script, to constantly check the queue for jobs, and execute/consume them. This second approach is better for scaling. The first one tho helps you to validate that everything is working up to that point 😛 But if there's a better way, I'm all ears! 🙂