man pages need to clearly state that checkpoints are discarded upon rewind #12646
Description
System information
| Type | Version/Name |
|---|---|
| Distribution Name | Any |
| Distribution Version | Any |
| Kernel Version | Any |
| Architecture | Any |
| OpenZFS Version | Any |
Describe the problem you're observing
Checkpoints are silently discarded upon rewind, but this is not documented anywhere.
The man page does state:
A pool checkpoint can be thought of as a pool-wide snapshot
and
To discard the checkpoint from a pool...
In neither case, nor in discussing rewinding does it mention that the checkpoint will be discarded upon rewind.
The man page seems to softly imply that a checkpoint has similar semantics to a snapshot. It does make it clear that there can be only one checkpoint. But a user who wants to take a second try at a destructive operation before performing it for real, may incur severe data loss and will be extremely disappointed to discover that rewinding is not possible a second time.
The man page should make it clear that rewinding will discard the checkpoint.
Describe how to reproduce the problem
- Make a checkpoint
- Rewind to it
- Try to rewind a second time