Cloud Tasks suddenly failing to create tasks for HTTP Target tasks with AppEngine permissions error

[acoulton]

Hi,

We have a number of systems and projects using Google Cloud Tasks, all creating and dispatching HTTP target tasks.

As of approximately 2024-07-22 21:07 UTC all our task creation attempts have been failing with a PERMISSION_DENIED and the message "Failed to create cloud task: App Engine targets require "appengine.applications.get" IAM permission (https://cloud.google.com/appengine/docs/admin-api/access-control) and the API_CLOUD_PLATFORM scope (https://www.googleapis.com/auth/cloud-platform)."

Nothing has changed in our application code, deployment, or GCP configuration in this time window - the running apps were creating tasks fine up until 21:07 UTC and then suddenly started to hit the permissions error.

We don't use AppEngine at all - all of our tasks are HTTP target tasks, some to GCP-native endpoints (e.g. Cloud Run) and others to our own applications behind Google Loadbalancer.

As far as I can see this is a change / error at the GCP end, but there is not currently anything showing on the service health dashboard.

We don't have a paid technical support contract with GCP, so I can't find any way to report the issue.

Any suggestions welcome!

[acoulton]

I've also reported this on the Cloud Tasks issue tracker at https://issuetracker.google.com/issues/354741406 - although given it looks like Cloud Tasks issues there haven't had any responses since late 2023 I'm not sure if anyone will see that.

I have  also confirmed that adding the `appengine.applications.get` permission to the service accounts that create tasks (which currently only have the Cloud Tasks Enqueuer role) seems to workaround the issue for now.

[pierry01]

Same problem here !!!

[annieredso]

Restarting my app temporarily solved the issue. But the problem appeared again after a few days. Will try OP's way: granting the unnecessary permission.

[annieredso]

Tried OP's way on July 29, it worked for a couple of days and then the problem appeared again today.