queue:work or queue:listen for Docker?

I have the very same problem, has anyone figured this one out yet?

I've the same problem as well over docker

Hey guys,

I should have posted an update about this when I found a solution; but totally forgot.

We ended up going with k8s Deployments which doing a rolling update of containers. When using this implementation you have the PreStop Lifecycle hook you can leverage.

https://kubernetes.io/docs/concepts/workloads/pods/pod/#termination-of-pods

For this we wrote a script that runs the following:

#!/bin/bash

cd /app
php artisan queue:stop

while [ `ps aux | grep "queue:work" | wc -l` -gt 1 ];
do 
  echo "Waiting for the queue to shutdown...";
  sleep 1
done

exit 0

This sends the shutdown signal to the queue worker. Then checks every second to see if the worker has been shutdown. The -gt 1 check is there because ps aux | grep always returns 1 result. You could modify this to exclude the grep, but why? This works.

Then once the queue:work process is dead, the script exits with a zero status code to signal a success.

This works because the PreStop lifecycle hook is blocking (synchronous). So the container will not shut down before that script exits. This allows the current job to be processed, and finished, before shutting the container down and replacing it with the updated container running the new app code.

To configure this within your Deployment file, you would do the following:

apiVersion: apps/v1
kind: Deployment
spec:
  replicas: 3
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: app
        image: app:{version}
        ports:
        - containerPort: 80
        lifecycle:
          preStop:
            exec:
              command: ["/app/bin/stopQueue.sh"]