1228585 - Running `npm test` fails: A task is created, but never get started

Status: RESOLVED WORKSFORME

(Taskcluster :: Workers, defect)

Description

[Johan Lorenzo [:jlorenzo]]

STR
1. Configure your environment, like described here[1]
2. Run npm test

Results
You get the following output:
>   live logging
> taskcluster/docker-worker-test:latest exists in the cache.
> Created Task ID: t0pvfunlSZ6i0_a2nFiKXQ
>     1) live log expiration date
> 
> 
>   0 passing (4m)
>   1 failing
> 
>   1) live logging live log expiration date:
>      Error: timeout of 240000ms exceeded. Ensure the done() callback is being called in this test.

The task is indeed created, but there is no worker that starts it up. At the beginning, I though that was due to the changes I made in bug 1132214, but even when I run the code present in master, I get the same results.

:wcosta, proposed to change the container[3] to "aws-provisioner-v1", but this requires the scope ["queue:create-task:aws-provisioner-v1/dummy-type-*"]. 


I tested it under Mac OS X (with the vagrant image) and under Linux (with no VM, docker being directly installed), both give me identical behavior.

:garndt, is it possible to have the right to create a task, but not the right to start it?

[1] https://github.com/taskcluster/docker-worker/pull/187/files
[2] https://tools.taskcluster.net/task-inspector/#t0pvfunlSZ6i0_a2nFiKXQ/0
[3] https://github.com/taskcluster/docker-worker/blob/544ba6f4c2ef41001f841d01bf3b218f2eaa0c94/test/testworker.js#L29

Comment 1

[Johan Lorenzo [:jlorenzo]]

After turning on the debug traces, I noticed I get a lot of authentication errors, once the task is created. I'm missing another scope:
> [["queue:poll-task-urls","assume:worker-type:no-provisioning-nope/dummy-type-*"]

Comment 2

[Johan Lorenzo [:jlorenzo]]

Thanks to :pmoore, I can now poll tasks. For the record, this other one is needed:
>  ["queue:claim-task","assume:worker-type:no-provisioning-nope/dummy-type-*"]

Comment 3

[Johan Lorenzo [:jlorenzo]]

Another missing:
> assume:worker-id:random-local-worker/dummy-worker-*

I'll list all the scopes I need in this bug. In the first place, back in bug 1228284 comment 0, I needed:
> ["queue:create-task:no-provisioning-nope/dummy-type-*"]

Comment 4

[Johan Lorenzo [:jlorenzo]]

> queue:create-artifact:public/*

Comment 5

[Johan Lorenzo [:jlorenzo]]

> ["queue:get-artifact:private/docker-worker-tests/*"]
> ["queue:cancel-task","assume:scheduler-id:docker-worker-tests/*"]