Realtime task list/status and replica count

[mrmachine]

Is your feature request related to a problem? Please describe.
When a service is having trouble starting or staying up, e.g. stuck in a fail loop because of the retry policy, it can be difficult to know the current status of tasks for that service as by the time you refresh and scroll back to the list, you've lost a few seconds and context and have to try to compare the list from memory with what you see now.

For example, every time you refresh the page you might see the top most task is "running" and a few failed ones below it. Is it actually running? Or is it just the latest attempt which will run for 1 second then fail? Right now, I have to reload the page, scroll down, try to remember the task ID and see if it has changed.

Likewise, when I deploy an update to a stack I want to just sit and watch the services status as they are each restarted, without having to constantly refresh.

Describe the solution you'd like
Realtime updates to the task list without a page reload (e.g. via websocket), would solve this problem. You just look at the page for a few seconds, and if nothing changes in front of your eyes, it is stable.

If something does change, you can click straight into it to investigate, which is handy because I have noticed that often with the top most task is already gone by the time i scroll the page, find the top most task and click it. If I had my mouse positioned ready to click and it updated status without any scrolling or reloading, I could click in immediate and see the error message (even if the task is then immediately removed and a page reload shows "task not found").

Describe alternatives you've considered

• Not trusting the displayed status and having to reload the page to see if it is stable.
• Always clicking on the second most recent task (or the first failed/rejected task) in the list, as this is usually more stable and does not report "task not found".

Additional context
Docker Cloud (which we are migrating from) had great live updates for things like service/container status (terminating, pending, running, stopped), so I could kick off a redeploy and then sit back and watch as each service or each container redeployed in sequence.

[vecernik]

+1