So this might be a newbie question but I just want to get my terms right.
so if I have a dataset with no snapshots, is what is exposed called the file sysytem?
If there are changes from after a snapshot was made then there is a layer of deltas which combined with the snapshot below presents as the file system.
Have I got the names and basic concept right here.? I just wonders if there was a name other than Files system to describe what is exposed from the dataset irrelevant of whether there are any snapshots.
In general, it is somewhat irrelevant if their are snapshots, clones or such on a Dataset / file system.
There is an odd case of the pool top level dataset. For NAS purposes, (and specifically TrueNAS), it is not advisable to share directly this top level dataset. Basically the pool name, like tank, is also a dataset called tank. But, it does not follow child dataset rules, like tank/tank, just simply tank.
You may know some or all of the below. In which case, this can be for future readers…
ZFS uses COW, Copy On Write. This means the “layer of deltas” is not like other systems that support snapshots. For example, both VMWare and Linux LVM snapshots are not meant to be kept more than a few hours, or days at most. They act more like write transaction log storage locations. And upon removal, it takes time to apply all the changes back to the “real” file system. Sometimes a long time, slowing down the server.
ZFS on the other hand, uses block pointers. So both the original dataset and snapshots can exist just fine, even for years. Each has pointers to their own version of the data. Similar behavior for cloned file systems. Just pointers in the pool. Any update to a file, causes the updated location to be allocated from empty space, and the block pointers for that specific dataset or clone is updated. (Snapshots are immutable… they always point to pre-existing data at the time of the snapshot.)
Thanks for the info, but how you refer to the exposed layer in a dataset, is it just called the file system, layer 1 etc.. it must have a specific name and I cannot figure out what it is called.
White stickers, taken as a whole, represent the live filesystem.
A complete set of stickers of a certain color (red, blue, or green), taken as a whole, represent a snapshot and the read-only filesystem that it is.[1]
A faded box is not seen by the live filesystem, but it remains and takes up space because it still has at least one colored sticker on it. (Referenced and protected by a snapshot.)
You can get an idea of what a snapshot’s read-only filesystem looks like by only considering the boxes that have a certain shared color, such as red, blue, or green.
All snapshots of a dataset are their own read-only filesystems, but they share blocks with the dataset’s live filesystem and other snapshots. This is why they don’t “add” used space equal to the amount of referenced data. More accurately, snapshots never “add” to the used space. They can only hold data that would otherwise be freed. ↩︎