Skip to content

ci: ISO builds fail intermittently due to ENOSPC during OSTree extraction #796

Closed

Description

@castrojo

What: The ISO build workflow (build-iso-bluefin.yml) intermittently fails with ENOSPC on GitHub-hosted runners when building non-composefs (OSTree) variants like bluefin-lts-hwe. These variants unpack all ~120 layers of their payload image onto the runner, which can exhaust the default disk space.\n\nFix: In projectbluefin/dakota-iso, we switched to aggressively freeing disk space at the start of the job using the org's standard ublue-os/remove-unwanted-software@v9 action, which reclaims ~30 GB. We should evaluate if this needs to be applied to other workflows producing OSTree artifacts.\n\nAutomatable: yes

Activity

  1. bketelsen commented on Aug 4, 2026

    @bketelsen
    Collaborator

    Audit complete: projectbluefin/dakota-iso already has the disk cleanup described here, and Common’s regular image build already receives the current v10 cleanup through projectbluefin/actions/bootc-build/setup-runner. The remaining gap was projectbluefin/iso’s shared Titanoboa workflow, which builds Stable and LTS-HWE OSTree ISOs without reclaiming runner space.

    I opened projectbluefin/iso#69 to run the org-standard v10 cleanup with extra-squeeze at the start of every ISO matrix job.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    blockedWaiting on something outside this repository. Use `/label blocked` or `/remove-label blocked`.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions