Replicating datasets within datasets

[survive]

Hi guys,

I have a filer running 9.3 that zfs replicates a single dataset locally from my 8 drive raidz2 pool (zpool0) to a single 3TB drive (backup_pool). The other day the dataset on "zpool0" finally outgrew "backup pool."

What I would like is to define replication tasks for 3 of the 4 datasets nested within so I exclude my media dataset. When I go to the "Replication Tasks" tab in the GUI to set up a new task the only dataset I can select is zpool0/dataset0, which is the only dataset I have ever replicated from. The manual makes it sound like I should be able to set up zfs replication using any dataset as a source, but I could be totally misunderstanding things as well.

I don't see any sort of way to explore for a dataset so I assume the FreeNAS should auto-populate the "Volume\Dataset" dropdown box in the replication setup.

So can anybody fact-check me here? Can you replicate a nested dataset & if so how do you tell FreeNAS which one you want to use as the source?

-Will

 

[Ericloewe]

I believe you can only replicate datasets which are directly snapshotted. I don't think you can replicate just a child dataset if you have a snapshot of the parent.

 

[cyberjock]

Ericloewe kind of sums it up.

ZFS will replicate snapshots (and can only replicate snapshots). Datasets are file system boundaries and as such aren't included in snapshots of parent datasets.

So unless you snapshot the actual dataset you want to replicate you cannot have any replication actually going on for that dataset.

 

[Glorious1]

Presumably if you snapshot recursively the parent dataset, you could select individual child datasets to replicate?

 

[cyberjock]

Presumably if you snapshot recursively the parent dataset, you could select individual child datasets to replicate?

Absolutely!

 

[survive]

Hi guys,

So here is what I figured out.....

I had a single recursive snapshot setup on the zpool (zpool0) and I couldn't "browse" within it to separate out my nested "Media" dataset. It looks like Ericlowe was on the money with the solution. I set up individual snapshots for each of the nested datasets and I was able to set up replication jobs the way I wanted to.

In retrospect it makes sense!

-Will

 

[Glorious1]

I don't know, it doesn't really make sense to me. The manual says for the recursive snapshot checkbox "Select this box to take separate snapshots of the volume/dataset and each of its child datasets". If you had that recursive checkbox checked, you should have snapshot of each of the child datasets. That means you should be able to select each one when you set up a replication. One thing that makes me wonder is you mention "browsing" within the parent dataset to select a child dataset for replication. You don't "browse" to select a dataset for replication, you just select it from the list in the drop-down box.

 

[Ericloewe]

I don't know, it doesn't really make sense to me. The manual says for the recursive snapshot checkbox "Select this box to take separate snapshots of the volume/dataset and each of its child datasets". If you had that recursive checkbox checked, you should have snapshot of each of the child datasets. That means you should be able to select each one when you set up a replication. One thing that makes me wonder is you mention "browsing" within the parent dataset to select a child dataset for replication. You don't "browse" to select a dataset for replication, you just select it from the list in the drop-down box.

I can confirm that I have snapshots for child datasets (recursive option checked on the parent, nothing configured for the child) but that the child datasets cannot be selected for replication tasks.

 

[Glorious1]

Well that's it then. It seems like it's either a bug or that limitation should be clarified in the manual.

 

[Ericloewe]

Filed as Bug 8068.