-
Hi, I've been trying to use the volumeClaimTemplate ability with RunnerSet, to give my ephemeral runners docker caching support. Currently, I'm seeing an issue where my next job automatically brings up containers from the previous job. I guess the reason is that when mounting the whole /var/lib/docker folder we also mount the What will be the right solution to take advantage of PVC but only for the docker layer caching and not to persist in running containers between jobs? any other methods you guys use to speed up docker build on ephemeral containers? Thanks! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 4 replies
-
@shaikatz Hey! There seems like no single solution to your problem, so you need to figure out how yourself mostly. But recently, awesome @TheDarkula has contributed an example config and his associated experience to set up a cache for rust builds. One of the essential learnings we have along the way was that you need to be sure which container within the runner pod you need to mount volumes, depending on whether you run dind within the runner or not, and whether you use containerized job steps or not. I think you can extend @TheDarkula's work to Also note that you'd need this patch #1340 to let ARC correctly clean and reuse PVs for RunnerSet-managed pods. |
Beta Was this translation helpful? Give feedback.
@shaikatz Hey! There seems like no single solution to your problem, so you need to figure out how yourself mostly.
But recently, awesome @TheDarkula has contributed an example config and his associated experience to set up a cache for rust builds. One of the essential learnings we have along the way was that you need to be sure which container within the runner pod you need to mount volumes, depending on whether you run dind within the runner or not, and whether you use containerized job steps or not.
I think you can extend @TheDarkula's work to
/var/lib/docker
for your goal.Also note that you'd need this patch #1340 to let ARC correctly clean and reuse PVs for RunnerSet-managed pods.