What to do with a 2nd Truenas Scale server?

[otpi]

I happened to get my hands on a 2nd server. I read all sorts of fancy terms on the TNS page, Scaled-Out ZFS, Converged, Active-Active, Linux Containers, Easy to Manage. I have tinkered with proxmox and clusters before. In terms an entusiast hobbyist can understand: What can the 2nd server do if I install Truenas Scale?

 

[sretalla]

At the moment... GlusterFS to allow clustered storage and clustered SMB shares. (need to manage this with TrueCommand, not in the TrueNAS SCALE GUI)

Later with Bluefin... Apps via kubernetes cluster.

 

[joeschmuck]

You could make a backup server which many have done and Rsync them so you always have a second backup in a family/friends house. But it depends on your use case.

 

[Ericloewe]

At the moment... GlusterFS to allow clustered storage and clustered SMB shares. (need to manage this with TrueCommand, not in the TrueNAS SCALE GUI)

Later with Bluefin... Apps via kubernetes cluster.

You'd need a third server, though.

 

[otpi]

Offsite backup is on my list. Although I plan to repurpose my current "main" NAS as a local cold backup first. (I have a copy of all important files in OneDrive for now).

Is/will it be possible to use Scale/GlusterFS and have all my datasets replicate to an offsite server and seamlessly move between sites and everything behave as if shares are local, and both servers share loads for apps? (e.g. if I put one at my parents and set up nextcloud/plex/smb/nfs)

Why a third server? Is it for quorum (or equivalent)?

 

[Ericloewe]

Is/will it be possible to use Scale/GlusterFS and have all my datasets replicate to an offsite server and seamlessly move between sites and everything behave as if shares are local, and both servers share loads for apps? (e.g. if I put one at my parents and set up nextcloud/plex/smb/nfs)

Conceptually yes, but expect latency to kill this in practice.

Why a third server? Is it for quorum (or equivalent)?

Something along those lines. Fundamentally, Gluster has a very hard time dealing with determining which data is good if there is no clear majority agreeing on that. It's much closer to a traditional filesystem than it is to ZFS in that and other design criteria.
What does this imply? For data integrity, you need an absolute minimum of three servers. If you don't care about uptime, the third can be just an arbiter, which takes up less disk space. If you do care about uptime, the third one needs to be a full-blown mirror, so that writes can continue in the absence of a single server.