Checkpoint/restore basics
Last updated
Was this helpful?
Last updated
Was this helpful?
The Cedana daemon is designed to checkpoint/restore processes as well as containers.
To checkpoint:
Where <type> can be process, containerd, runc, job, etc. See for all plugins that support checkpointing.
For example, to checkpoint a process:
A --dir flag can be used to specify the parent directory where the checkpoint will be stored. If not provided, the checkpoint will be stored in the default checkpoint directory as specified in the , or in /tmp if not set. You may also specify a --name flag to give a custom name to the checkpoint file.
See for all available options for process checkpoint.
To restore:
Where <type> can be process, containerd, runc, job, etc. See for all plugins that support restoring.
For example, to restore a process:
Notice that for restore the flag is called --path instead of --dir (as in dump), this is because it can be a path to a compressed file, or to a directory if not compressed.
The cedana dump/restore job subcommands have the same options as their non-managed counterparts, but with pretty good defaults. For e.g., the --path flag is not required for cedana restore job, as the checkpoint path is stored in the job metadata.
If you do cedana job list after checkpointing a job, you will see the latest checkpoint time and size:
To view all checkpoints for a job, use cedana job checkpoints <job_id>:
The cedana dump subcommand supports a --compression flag to specify the compression algorithm to use. For example:
This will create a compressed checkpoint file with the path /tmp/xyz.tar.gz. The --name flag is optional, and if not provided, the daemon will choose a unique name based on some metadata.
When restoring, the daemon will automatically detect the compression algorithm used and decompress the file. Simply provide the path to the compressed file:
Supported values for --compression are none, tar, gzip, lz4, zlib.
See for all available options for process restore.
As explained in , a job can be of any type, and thus can be checkpointed and restored using the cedana dump job and cedana restore job subcommands.
You may also specify the default compression algorithm in the .