Checkpoint
Capture full process state with CRIU for forensic analysis.
This guide covers CLI commands for testing and one-off captures. For automated checkpoints, see Scheduling.
Complex workloads may encounter CRIU limitations. For backup/recovery, consider Layer operations or Memory dumps. See Known Limitations.
What is a Checkpoint?
A checkpoint is a complete snapshot of a running container's state—like pausing a video game and saving exactly where you are. Using CRIU (Checkpoint/Restore In Userspace), reel freezes the container and captures everything needed to understand or resume its execution:
- • Memory contents — variables, secrets in memory, cached data
- • Process state — what code was executing, call stacks
- • Open connections — active network sockets, open files
- • Pending operations — timers, signals, async work
This is different from layers, which only capture files on disk. Checkpoints let you analyze what a process was doing, not just what files it had.
Before You Start
Checkpoints require CRIU on your nodes. Most clusters don't have this by default—see Requirements for setup.
Check if your cluster supports checkpoints:
Find Your Workloads
List pods that reel can operate on:
Create Checkpoint
Output:
Export Checkpoint
Save a checkpoint to a local file or S3 for offline analysis:
Checkpoint Methods
By default, reel creates checkpoints through the Kubernetes API. If you prefer to use containerd or CRI-O directly, you can specify a different method—see Checkpoint Reference for options.
When to Use Checkpoint vs Layer
| Need | Use |
|---|---|
| Filesystem backup/restore | Layer |
| Memory analysis | Checkpoint |
| Quick rollback | Layer |
| Process state forensics | Checkpoint |
| Reliable backup | Layer |