Use a GitHub Actions runner with a larger disk space

From community channel:

I have a really big Dockerfile:

ERROR: failed to build: failed to solve: ResourceExhausted: failed to compute cache key: failed to > copy: write /var/lib/buildkit/runc-> overlayfs/content/ingest/d352575557ef1bb14551b2d1ddc4776a3b6aca8f9a7f9797cedb14d6cf44e368/data: no > space left on device

Solution:

• if RunsOn version <2.9, simply use disk=large and ensure that your RunnerLargeDiskSize stack parameter ↗ is set to a value greater than 80.
• if RunsOn version >= 2.9, use the new volume label, and just set e.g. volume=120g in your runs-on: labels.
• for all RunsOn versions, you can also use EC2 instances with a locally attached SSD instance storage (usually instances with a d in their name, e.g. m6id.large). If local disks are detected, /var/lib/docker will automatically be mounted on it. Local disks range from tens of gigabytes to terabytes.
• finally, on larger memory instances you can use extras=tmpfs so that /var/lib/docker is mounted on a tmpfs volume. E.g. if you use instances with 128GB memory, you’ll get at least 100GB free for your builds. Also it should be faster as it skips the disk entirely.

Last updated: October 22, 2025

View original discussion ↗