CODEWITHSUNDEEP
google-app-enginegoogle-cloud-platformgoogle-cloud-tasks
Magoo
Magoo

Reputation: 2638

Cloud Tasks with an automatically scaling App Engine that last longer than 10 minutes

I have a backend for an iOS app that I've built on App Engine and I'm looking to do potentially long running background tasks to add records to my Cloud SQL database. Is this possible without Compute Engine? I've seen Cloud Tasks can do asynchronous work and you can set the dispatchDeadline to basically anything you want, but I've also read in the documentation

For App Engine tasks, 0 indicates that the request has the default deadline. The default deadline depends on the scaling type of the service: 10 minutes for standard apps with automatic scaling, 24 hours for standard apps with manual and basic scaling, and 60 minutes for flex apps. If the request deadline is set, it must be in the interval [15 seconds, 24 hours 15 seconds]. Regardless of the task's dispatchDeadline, the app handler will not run for longer than than the service's timeout. We recommend setting the dispatchDeadline to at most a few seconds more than the app handler's timeout. For more information see Timeouts.

I don't particularly need the App Engine instance to care if the task completes or not... so I'm not sure why the recommendation is at most a few seconds more than the app handler's timeout ... can anyone shed any light on this? What am I missing? Adding a Compute Engine for these relatively simple tasks that will take at most a few ours to complete seems like a lot of overhead and I don't want this to dictate which scaling options I choose.

Thanks for your time.

Upvotes: 0

Views: 761

Answers (1)

guillaume blaquiere
guillaume blaquiere

Reputation: 75880

The recommendation is only for logging purpose. If your task timeout is shorter that your app timeout, you never know if there is an error from your app, because you don't have the return.

If you have longer timeout on Cloud Task, you can catch and trace in Cloud Task logs the app return code and thus gracefully track the errors.

App Engine with a basic scaling mode is a great solution.

  1. You have 9H free per days (B instance type)
  2. App Engine scale to 0 automatically after a period of inactivity (that you can define in the basic scaling: idle_timeout parameter)
  3. You have a regional available service. Not a zonal like a compute engine, or you need to have 9 computes engine, to cover the regional High Availability (3 per zone, over 3 zones)
  4. You don't have server to manage: no update, no patching, no network/ip/firewall rule...

If you ask me about the overhead, I will anser Compute Engine and not App Engine (even if you need few configuration)

Upvotes: 1

Related Questions