- Category
- Developer Tools
- Rank
- No. 1847Tools index
Previous survey · No. 1766 ·
- Type
- TOOL
- Use case
- Deployment & Operations
- GitHub
- 10 stars
- Latest release
- v0.1.0
- Date
About
Corral runs a command with a wall-clock limit and aims to leave nothing it started still running when it finishes. The command gets its own session so killing it cannot kill the caller. It uses a cgroup when available to kill the whole process tree, falls back to tracking the tree through /proc, and exits with code 120 if it cannot confirm everything is dead. This addresses agent runners that only kill the process they started, leaving background or double-forked children alive.
At a glance
Corral runs a command with a deadline and checks whether its descendants have stopped. It is worth evaluating when timed-out jobs leave background processes holding ports or files.
Good fit
Linux developers who need predictable cleanup after an agent or test command exits.
Consider something else when
Use a security sandbox when you need to restrict filesystem, network or privilege access.
Start here
Try a harmless command that spawns a background child in a disposable environment. After it exits, confirm the child is gone and check how your caller handles unverified cleanup.
Cost and requirements
Requires Linux 5.11 or later. The project documents Linux 5.14 or later and cgroup v2 for enforced mode.
Limits to know
The fallback uses subreaping and process inspection. Corral reports a separate failure when it cannot establish cleanup. It does not restrict what commands can do while running.
Sources checked 2026-11-05. Source-based assessment.
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.