Cache-aware placement: jobs go where their cache is issue #51 closed: completed andymccall opened this 22 days ago · 0 comments

Each runner keeps its own build cache (issue #28), and nothing steers a job to it: in oxygenhub run 226, build and clippy ran on the other runner from the run before and started cold (build 251 s cold on plex, 350 s on Earth). Runners report the caches they hold on each claim (repository and job), and the hub prefers the runner holding a job's cache as it prefers one holding its image (held up to 15 s, the reason in placed).