'Upgrade' replication tasks in 11.3?

[Basil Hendroff]

After upgrading a couple of FreeNAS servers to 11.3 (without any problems whatsoever I might add), I noticed that replication tasks that have come across in the upgrade are using the LEGACY transport. I understand from reviewing the 11.3 documentation, that other transports enable additional options for replication tasks. Several questions spring to mind:

1. Are there any immediate benefits in 'upgrading' replication tasks that use the LEGACY transport to say use the SSH transport instead if those tasks are already working as intended?
2. If there are additional immediate or longer-term benefits, is it as simple as editing the existing replication task and selecting SSH in the drop-down menu for the Transport field? If so, does it matter the SSH connection is using a Legacy Replication Key Pair?
3. Or should I be creating a new SSH connection, new key pair and new replication tasks to replace the legacy equivalents?
4. Am I correct in saying that the SSH+NETCAT transport is only useful in the instance that a high-speed data path exists between servers participating in the replication?

 

[Basil Hendroff]

*bump*

 

[Jedi940]

I'll give this a bump too. I am researching this new version before I upgrade and one thing I cannot afford is to have to start replication all over. I don't have the bandwidth to move 53TB of data all over again. A proper migration guide and how it works would be very helpful.

 

[JaimieV]

I'll give this a bump too. I am researching this new version before I upgrade and one thing I cannot afford is to have to start replication all over. I don't have the bandwidth to move 53TB of data all over again. A proper migration guide and how it works would be very helpful.

You won't need to restart the replication from scratch. Replication is copying of snapshots from pool A to pool B. If the replication task is changed/deleted/replaced/whatever, on first use ZFS just checks which snapshots are on each end for the defined datasets and copies any not on B from A.

 

[Jedi940]

In theory, you are correct, however, in practice, when I disabled the old replication task and created a new one to use the new system, selecting the old snapshot task, it deleted all my snapshots and I still have to start from scratch.

 

[JaimieV]

Huh. Worked fine for me in practice, my three pools picked up where they left off after I switched replication from old-style and push to new ssh+netcat pull on the backup NAS. No loss of snapshots.

Are you sure you got the periodic snapshot tasks and/or snapshot name scheme correct so it could find the existing ones? That took me a good half hour to get right, because I was typoing rather than copy-pasting and also didn't know what exactly was needed.

The documentation for the new replication and needs for snapshot name formatting is lacking right now. Badly needs a worked example in the docs, and fuller helptext in the UI on the (?) buttons.

 

[GusGus67]

I'd like to know about this too. Not urgent, but I'd like to take advantage of the higher speed offered by SSH+NETCAT. My Storage NAS and my backup NAS are connected by 10Gbe, so I think it would be an advantage

 

[Basil Hendroff]

I'd like to know about this too. Not urgent, but I'd like to take advantage of the higher speed offered by SSH+NETCAT. My Storage NAS and my backup NAS are connected by 10Gbe, so I think it would be an advantage

I've got the opposite issue. I've got a 20Mb/s DSL service to my backup NAS so I'm wondering if there's any advantage in using SSH+NETCAT. A more general question would be 'Is there a threshold speed that makes SSH+NETCAT the preferred option?'.

 

[geoff.jukes]

To "upgrade" my replication and Snapshot tasks, I did this:

1. Edit the Replication task
2. Changed "Transport" from "Legacy" to "SSH"
3. Selected the appropriate "Periodic Snapshot Tasks"
4. Click Save
5. Edit the Periodic Snapshot Tasks (State will show "Pending")
6. Disable "Allow taking Empty snapshots"
7. Click save

Because I disabled #6, I changed the snapshot cycle to 24-hour. I used to limit the snapshot hours to "when people were probably making changes" but there's no need to do that now.

YMMV