Destroy Dataset = Delete Files?

[ppmax]

Hi--

Pretty basic question, but I couldn't find the answer in the documentation.

I have a dataset ("media") that has ~450G of files in it. If I destroy this dataset are the files destroyed as well?

Also, is there any advantage or disadvantage to having "more" datasets or "fewer" datasets? In other words, is it preferable to have a single large dataset or a bunch of smaller datasets?

Scenario one:
volume1 //volume
volume1 //dataset
--dataset1
--dataset2


Scenario two:
volume1 //volume
volume1 //dataset
--directory1 //not a dataset
--directory2 //not a dataset



Thanks much--

 

[danb35]

If I destroy this dataset are the files destroyed as well?

Yes. Destroying a dataset destroys all its contents--files, zvols, snapshots, etc.

 

[nojohnny101]

Also, is there any advantage or disadvantage to having "more" datasets or "fewer" datasets? In other words, is it preferable to have a single large dataset or a bunch of smaller datasets?

Not really, the stock answer here would be "it depends". Datasets offer more granular control over permissions, quotas, etc.

To my knowledge, FreeNAS doesn't care how many datasets you have, but obviously don't create more than are necessary.

 

[ppmax]

Thanks for the replies--much appreciated. Im doing a little restructuring and wanted to confirm a few things before rm'ing a bunch of stuff.

 

[Arwen]

The naming, amount and attributes, (sharing, compression, etc...) of your datasets should reflect usage. For example, I have;

tank/Media
tank/backups
tank/backups/HOST1
tank/backups/HOST2
tank/distfiles
tank/storage

Each with it's own use and sharing.

 

[ppmax]

awesome, thanks for the additional information. I've been using FreeNAS for years but have never learned best practices (or seen discussions about them).

As I mentioned earlier I'm doing a reorganization of my storage scheme. Instead of 1 dataset with an assortment of subdirs which are share points, I'm creating multiple datasets (which are shared) like your scheme above.

Do you use compression on any of your datasets?

 

[Arwen]

awesome, thanks for the additional information. I've been using FreeNAS for years but have never learned best practices (or seen discussions about them).

As I mentioned earlier I'm doing a reorganization of my storage scheme. Instead of 1 dataset with an assortment of subdirs which are share points, I'm creating multiple datasets (which are shared) like your scheme above.

Note that for me, I have separate child datasets for some of my client backups. This allows me to snapshot them individually, and then full backups run faster. (Rsync does not have to copy files that have not changed...) Basically I am using snapshots to perform de-duplication of my backups. Thus, no need to perform any incrementals, since the snapshots are a type of incremental. Kinda confusing until you plan it out and then use it.

Do you use compression on any of your datasets?

Yes, even though for the Media and distfiles it's not really relevant. Basically, ZFS's new default compression algorythm, LZ4, is fast enough and will bailout if it encounters less compressable data, that there is less reason to not use LZ4 compression as a default for the entire pool.

 

[ppmax]

Cool--I'll read up on snapshots...haven't really used those in the past since I had one large dataset with lots of churn.

So you do snapshots and then rsync those to another instance or offsite?

Thanks for all your comments/help

 

[Arwen]

...
So you do snapshots and then rsync those to another instance or offsite?
...

No, I simply use the snapshots on my FreeNAS for a history of the client backups. I then copy all
the current datasets, (not the snapshots), to an 8TB disk I use for backup of my FreeNAS. That does
get taken off-site occasionally.

 

[Robert Trevellyan]

Regarding organizing datasets, divide things up as much as makes sense for your needs, but remember most sharing methods don't span filesystems. In that context, a dataset is a filesystem, which means sharing a parent dataset doesn't share a child dataset.

 

[gpsguy]

Instead of rsync, you might want to look at replication - http://doc.freenas.org/9.10/storage.html#replication-tasks

So you do snapshots and then rsync those to another instance or offsite?