mirror of
https://github.com/onyx-dot-app/onyx.git
synced 2026-10-22 19:32:55 -05:00
roshan
92d5149cb5
## Description Idle cleanup preserves recoverable files after checkpoint or workspace discovery failures, including on unhealthy endpoints. Kubernetes requires complete workspace-list output. Docker rejects failed or incomplete snapshot probes. Migration `45662b7b63a2` adds nullable `sandbox.cleanup_checkpoint` JSONB storage. Cleanup commits proof after snapshots and final checks, before deletion. The proof update requires the current provisioning attempt and observed heartbeat. Retry requires a confirmed absent runtime and matching attempt, workspace storage identity, and heartbeat. Heartbeats and file mutations invalidate old proof before files change. Successful sleep and new provisioning clear the proof. ## How Has This Been Tested? 614 sandbox tests passed with two pytest workers. Tests cover partial Docker deletion across workers, changed storage, stale attempts, resumed activity, and failed checkpoint commits. Migration upgrade and downgrade SQL generation passed. Pre-commit checks passed. PostgreSQL persistence and heartbeat-race tests were added. Their local run was blocked because PostgreSQL was unavailable. Live Docker services were also unavailable locally. ## Known CI blocker Preview deployment failed with an unauthorized deploy-token error after one infrastructure retry. See the [failed preview run](https://github.com/onyx-dot-app/onyx/actions/runs/39660940780). Code, Craft, database, and type checks passed. Other CI jobs are still running. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> <!-- coderabbit-ai-generated:v1 --> ## Summary by CodeRabbit * **Bug Fixes** * Idle sandboxes remain running when workspace discovery or snapshot checks fail, preserving their files, runtime, and ports. * Interrupted cleanup can resume when the sandbox state is unchanged. Existing workspace storage is preserved when cleanup cannot be verified. * Uploads and file deletions are coordinated with session activity, preventing cleanup from interfering with active file changes. * Cleanup checkpoints are cleared when sandbox activity or lifecycle changes make them outdated. <!-- end of auto-generated comment: release notes by coderabbit.ai -->