You need to sign in or sign up before continuing.
Create process to troubleshoot Knapsack scheduling
Context
See gitlab-org&11260 (closed).
- We have noticed that RSpec parallelized jobs are sometimes timing out (investigated in #271 (closed))
- We have noticed that some jobs in the same
RSpec job classes
have sometimes very different runtimes (discussed in #272 (closed))
So far, we have used ad-hoc approaches to troubleshoot those issues (1,2).
Goal
Create a process to have a standard way to diagnose the following issues:
- Understand how Knapsack scheduled work in a given RSpec job class, and where the differences might come from (example)
- Understand why a given spec job timed out (example)
This process could be added to the pipelines RUNBOOK's learning section.
Technical notes
I just saw that we could change the Knapsack log level. Might be worth a try!
Edited by David Dieulivol