984935 - Add task timeout after period of inactivity

Status: RESOLVED FIXED

(Taskcluster :: General, defect)

Description

[James Lal [:lightsofapollo]]

Usually this is defined as "no output in stdout or stderr in N seconds"

Comment 1

[James Lal [:lightsofapollo]]

This is a worker level option not to be confused with how timeouts work at the queue level (which exists for entirely different reasons).

Comment 2

[Jonas Finnemann Jensen (:jonasfj)]

Sounds good. This should be added to the task payload, at least for now.

We should also consider adding `maxRunTime` in the same go. As maximum number of seconds the worker is allowed to work on a task.

When either of these conditions is violated, we should abort the run and ask the queue to reset `takenUntil`.

Comment 3

[Jonas Finnemann Jensen (:jonasfj)]

`maxRunTime` have been introduced with 4a0fea8400c75fab823442569e528abe392c390e.
Inactivility is a little harder to define, specifically I suspect that node won't flush every single byte individually. For example python is known to only print full lines.

Anyways, `maxRunTime` should always be specified and makes this less critical.

Comment 4

[Pete Moore [:pmoore][:pete]]

Resetting Version and Target Milestone that accidentally got changed...