iSCSI extent type

[Andres Gonzalez]

Hello,

I'm configuring an iSCSI target and I found that there are two options when configuring the extent:

- device
- file

Is there any difference regarding performance, use, etc. or it's the same ?

Also, It's better to create first a datastore then the zvol on the volume, or creating the zvol under the volume without a datastore it's ok ?

Thanks.

 

[cyberjock]

I have no clue what you are referring to when you say "datastore" as FreeNAS has no "datastore".

Right now file-based provie better performance, but once 9.3 comes out zvols will provide better performance. So if I were you I'd go with zvols. The reason why it changes is that 9.3 uses a kernel-based iSCSI service and that changes the mechanics of iSCSI.

 

[Andres Gonzalez]

I have no clue what you are referring to when you say "datastore" as FreeNAS has no "datastore".

Sorry, I wanted to say "dataset".

Right now file-based provie better performance, but once 9.3 comes out zvols will provide better performance. So if I were you I'd go with zvols. The reason why it changes is that 9.3 uses a kernel-based iSCSI service and that changes the mechanics of iSCSI.

OK! Thanks.
It's better to create a dataset and zvol under the dataset, or just create de zvol "under" de volume without dataset ?

 

[cyberjock]

Doesn't matter where the zvol is.

 

[toadman]

True, it doesn't matter, unless you are trying to manage the zvol I suppose. For the OP, remember, you can set dataset specific attributes, so putting the zvol in a separate dataset will allow you to manage it differently than other datasets presumably used for different purposes.

For example, I put my zvols into a separate dataset so I can manage the specifics of compression, dedup, snapshots, and replication independently of my other datasets. I (currently) only use these zvols for vmware storage (datastores).

 

[cyberjock]

True, it doesn't matter, unless you are trying to manage the zvol I suppose. For the OP, remember, you can set dataset specific attributes, so putting the zvol in a separate dataset will allow you to manage it differently than other datasets presumably used for different purposes.

For example, I put my zvols into a separate dataset so I can manage the specifics of compression, dedup, snapshots, and replication independently of my other datasets. I (currently) only use these zvols for vmware storage (datastores).

Not true. zvols get their own properties just like a dataset. You can set the dedup, compression, etc for a zvol directly.

Just do "zfs get all <pool>/<zvol name>" and you can see for yourself.

 

[toadman]

Very nice, yes, you are correct (and I stand corrected). ZVOLs do have independent attributes. I do manage my zvols in a single dataset as my needs have it such that all the attributes are the same. And I replicate the entire dataset.

However, this OP may be able to simply assign attributes to a single zvol.