Nested znapzend schedules issue?

[c0ldtech]

Hi,

I'm new to znapzend (as well as ZFS). I think I misunderstand something about ZFS or about znapzend (not sure which one). My idea was to make regular one day interval snapshots of a parent dataset, and use shorter intervals for some of the child datasets. So what I have is the following structure (obviously the children contain appdata for docker containers):

tank/docker
tank/docker/dockerA
tank/docker/dockerB
tank/docker/dockerC

So what I did was create a snapshot schedule for tank/docker, which creates a daily snapshot using --recursive. Then I created another schedule for tank/docker/dockerA, which creates snapshots every 6 hours.
That way I do not need to create an extra schedule for everything that gets created under tank/docker later, and do not need to create as many snapshots for tank/docker/B and tank/docker/C where a daily snapshot is good enough.

First question: is there any obvious issue with this?
I'm asking because the following happened: when stopping dockerA and then starting it again, its integrated SQLite db was suddenly missing data up until the last 6h snapshot. I.e. I restarted the DB at 10am, and suddenly all the data that was added after 6am (when the last snapshot was created) was gone. To be clear: I did NOT restore a snapshot at any point. Can anybody explain this? This makes me really uncomfortable not to know how it happened, since of course the whole point of using snapshots is to have more data security, not less.

Another question related to ZFS and datasets, not so much znapzend itself: what happens when I create a dataset where there already is a directory with the same name that already has data in it? It seems the data is then gone?

Example: the following dataset exists: tank/datasetA
Now I'm creating a directory "nameA" in tank/datasetA and putting data into it. Later I realise that I forgot to create a separate dataset for nameA and create the following dataset: tank/datasetA/nameA.
What happens with teh content of former directory nameA?

Also, can I mix datasets and directories (I do not want to do so intentionally, but it may happen), i.e.

tank/datasetA/dataset1
tank/datasetA/directory1
tank/datasetA/dataset2
tank/datasetA/dataset3
tank/datasetA/directory2

Is this an issue in any way? Or are datasets just regular directories with additional metadata, so nothing unexpected will happen with this configuration?

Any help or insight would be appreciated, thanks!

Best regards
Hendrik