1243769 - [meta] [tracker] Separating infra from task issues - line in the sand

Status: RESOLVED INVALID

(Taskcluster :: General, defect)

Description

[Selena Deckelmann :selenamarie :selena]

This bug is for collecting issues related to infrastructure error/exception reporting and task exceptions/errors/failures. I'm hoping to come to some conclusion about where we should draw the line, along with an explanation, based on the kinds of problems that have been raised so far.

TaskCluster is trying to put the focus on tasks, and moving to make infra failures less or entirely irrelevant during task execution. The line can be blurry because of the dependancies tasks have on external services we don't control, and perhaps can't control. 

For many infra issues, I wonder what more we can do to make them fail less. For the things we'll never be able to control, I wonder if we should care less and retry/rerun more.

Comment 1

[Selena Deckelmann :selenamarie :selena]

Key questions that "infra" logging answers: 

* should I retrigger this task? 
* how should I categorize this failure? (before trying to find another human to help figure it out)

Example: taskcluster-proxy failures due to task configuration