BYOC: Pathology in "evenly" group scheduling of jobs

Jan Vesely's Avatar

Jan Vesely

26 May, 2020 07:00 PM

Hi,

I've noticed that when a job gets canceled because of a newer commit, the new job gets scheduled before the old one gets canceled.
This leads to the following pathology with "evenly" with distribution strategy;
one job is running on "cloud A" one job is running on "cloud B"

a new commit creates a new job, it gets scheduled on "cloud A", however the commit forces cancellation the job on "clout B".
The result is 2 jobs on "cloud A" and 0 jobs on "cloud B".
There's 50:50 chance of this happening with 2 clouds.
However, if the cancelled jobs were removed before the new one is scheduled it should always schedule the job on the idle cloud.

Reply to this discussion

Internal reply

Formatting help / Preview (switch to plain text) No formatting (switch to Markdown)

Attaching KB article:

»

Attached Files

You can attach files up to 10MB

If you don't have an account yet, we need to confirm you're human and not a machine trying to post spam.

Keyboard shortcuts

Generic

? Show this help
ESC Blurs the current field

Comment Form

r Focus the comment reply box
^ + ↩ Submit the comment

You can use Command ⌘ instead of Control ^ on Mac