The BE clone's source snapshot name should be more descriptive

[Zach Himsel]

Right now, the snapshot created to clone a BE dataset is named with just the date (e.g. ROOT/default@2020-03-23-14:46:49).

I'm thinking that the snapshot name should be one of the following:

• name of the new BE (so @new_be for zectl create new_be)
• @zectl_<be_name>
• or at least optionally overridden by the user (via commandline flag or dataset property)

This would make it easier to identify what a snapshot's purpose is when looking at snapshot lists.

[John]

zectl as a prefix might be worth having, and it could be made changeable by a property.

Perhaps:

<ds>@<prefix>-<be_name>-<date>

Not sure what I think about sticking the BE name there... If one does want to find the snapshot that their dataset has been cloned from they can use:

zfs get origin <ds>

IE:

zfs get origin vault/sys/enix/ROOT/linux-lts-5.4.19
NAME                                  PROPERTY  VALUE                                            SOURCE
vault/sys/enix/ROOT/linux-lts-5.4.19  origin    vault/sys/enix/ROOT/default@2020-02-18-21:05:41  -

[Zach Himsel]

If one does want to find the snapshot that their dataset has been cloned from they can use...

You could say the same for the date, too: zfs get creation <snap> :)

Personally, I don't find dates useful in "manual" snapshots (as opposed to automatically created ones, like via sanoid). However, I can see the merit to avoid namespace issues (say you have a BE named the same as a previous snapshot).

I like the @<prefix>-<be_name>-<date> format. I think zectl is a good default for <prefix>, which could, as you say, be made overrideable via a dataset property (or pool property?).

I just think the date-only snapshot name is way too vague if you're not looking at the clone itself. As far as I can tell, a snapshot has no property to indicate that it's an origin for a clone.

[John]

Yea, the date is mostly to make them unique.

I'll put this on my list of things to add 👍