Integrate "pool checkpoints" into the GUI, middleware, and automation

Agreed.. for 100 a script to automate would be needed.

If we’re talking about accidental deletion then I don’t think this would help. The administrator would surely know how to remove the hold and then delete the dataset so we are in the same boat. The checkpoint is so powerful because it’s such a sledgehammer you hope to never use. I can’t see anyone accidentally rolling back a checkpoint but I can see someone deleting a dataset and hold and regretting it. I could also see this happening in a malicious attack. I’ve lost count of the times I’ve seen people on this forum that have lost data through mistakes often replicating over a live dataset among other things. A checkpoint would have helped out so many users over the years from losing data.

It would be useful to document a sane way of using them since there are restrictions and rolling back will lose any recent data.

Agreed. Personally I have a cron job that takes a checkpoint every week and releases it a minute before as you can only have one. For me this seems a reasonable timeframe to cover disasters but not to grow too large to annoy. As I mentioned above consideration would be needed around disk replacements and hot-spares as neither work when a checkpoint is in-place.

In the event of a disaster then you would need to decide if A. a permanent rollback is desirable or B. a read-only temporary rollback is needed.

This is a bit older but since feedback from enterprise was requested: I can say from experience that losing recent data is absolutely fine in a business environment if the alternative is to lose all the data.

Source: A restore after a malware attack in a business where backups were used which - of course - did not contain all the recent data.

I’d see this feature like a rolling backup but done within ZFS and I voted for it and will implement it as a scheduled task first thing tomorrow on my & our systems (thanks @winnielinnie for brining this to our collective attention!)

Not that I’m working for them - I don’t - but I’m sure MAERSK would agree (look up their attack story or listen to the darknet diary episode. Scary stuff)

Seems I posted to the wrong thread, so I’ll repeat myself a bit here.

I think the main things I want to add are:

  1. Another useful place for creating checkpoints is on a system upgrade. It’d be nice to have a list of pools with the option to set which should be checkpointed (defaulting to all?) as checkpoints make disaster recover a lot easier, especially if a new ZFS version was involved.
  2. Anywhere a checkpoint can be created should include an option to set a removal reminder, enabled by default (48 hours? One week?) after which TrueNAS will check if the checkpoint still exists*, and if so will alert the user and provide an option to remove it (with confirmation). This way users will be less likely to forget about checkpoints and then wonder why their capacity is disappearing.

*To determine whether the checkpoint for a reminder still exists, TrueNAS will need to save the checkpoint creation date from zpool status <pool> and compare against it when the reminder occurs — if the dates are the same, then it’s the same checkpoint so the reminder can be shown. It doesn’t look like there are currently any properties that could be used instead (unless they’ve been added in ZFS 2.4.x? I’m still on TrueNAS 25 for now)? All I could find was a property for the size (how much deleted data the checkpoint is holding onto).

Anyway, this has my vote! Checkpoints have saved my ass a few times, and my worst disaster recovery experiences have been when I failed to take any, so making it opt-out would provide a good safety net.