TrueNAS Core as virtual filer - possible with just 1 data vdisk? datasets later extendable?

callmr

Dabbler
Joined
Feb 9, 2020
Messages
13
I am thinking about using TNC as a virtual filer, i.e. installing TNC in a VM on ESXi.
I do not need and want to use redundancy features that cost storage space on vdisks (as the vdisks are actually stored via NFS on a physical TNC).

I do have a physical TNC with ZFS using HDDs and SSDs for caching, but my technical understading of details on ZFS is, alas, rather limited.
So before I proceed i would like to know

(a) is it possible to create an FS starting with just 1 vdisk for data? (The how-to that I found uses 3 vdisks)

(b) when a dataset later runs out of space, is it possible to manually extend the dataset with additional storage space by adding another vdisk?
 

sretalla

Powered by Neutrality
Moderator
Joined
Jan 1, 2016
Messages
9,700
(a) is it possible to create an FS starting with just 1 vdisk for data? (The how-to that I found uses 3 vdisks)
With a single disk (virtual or not) as stripe, yes. but... https://www.truenas.com/community/t...ide-to-not-completely-losing-your-data.12714/

(b) when a dataset later runs out of space, is it possible to manually extend the dataset with additional storage space by adding another vdisk?
Yes, possible... terrible practice in general and with no redundancy (as you said you're happy with), see link above also.


NOTE: what you're doing (together with what I confirmed) will technically work, but is in no way endorsed by me as a good way to run things.
 

callmr

Dabbler
Joined
Feb 9, 2020
Messages
13
Thank you!

I'm possibly not looking right, but I do not see where the linked description has a procedure to extend a dataset by adding a (v)disk to the array.
I also looked in the TNC documentation under Datasets, Manage Datasets, but also nothing there.
What would be the right chapter of the TNC documentation to read up on the topic of eexetending a dataset?
 

callmr

Dabbler
Joined
Feb 9, 2020
Messages
13
NOTE: what you're doing (together with what I confirmed) will technically work, but is in no way endorsed by me as a good way to run things.
what would you consider a good way to have several virtual filers, each located in a different VLAN? (due to security reasons I do not want to have a multi-homed filer that would in theory be able to route between different VLANs)
 

sretalla

Powered by Neutrality
Moderator
Joined
Jan 1, 2016
Messages
9,700
I do not see where the linked description has a procedure to extend a dataset by adding a (v)disk to the array.
You don't see it as it's not there... datasets are able to consume the full capacity of the pool they reside on.

Extending a pool with a disk (stripe) is trivial and can be found in the documentation.

Again, not a good general recommendation.

what would you consider a good way to have several virtual filers, each located in a different VLAN? (due to security reasons I do not want to have a multi-homed filer that would in theory be able to route between different VLANs)
You can assess for yourself what's good for you.

When I write something on these forums, I'm careful not to endorse things that are bad ideas for most people. (you can decide if you're most people or not)
 

callmr

Dabbler
Joined
Feb 9, 2020
Messages
13
Extending a pool with a disk (stripe) is trivial and can be found in the documentation.



You can assess for yourself what's good for you.

When I write something on these forums, I'm careful not to endorse things that are bad ideas for most people. (you can decide if you're most people or not)

Thanks, that was the pointer that I needed.
Sure I can assess that for myself, but my question was honestly meant, as you might have provided an idea that I liked better than my original plan.

Thanks again.
 

sretalla

Powered by Neutrality
Moderator
Joined
Jan 1, 2016
Messages
9,700
Perhaps one consideration would be ZFS x2... if you're passing block storage which sits on ZFS to make a ZFS pool on top of, you're checksumming twice on the same blocks.

Selecting another option that supports non-ZFS volume sharing might be wise if you want the maximum performance and don't care enough about the data to checksum it twice.

Something like OpenMediaVault might be suitable and also a bit lighter on RAM requirements.
 
Top