Changing config from RAIDz2 to Mirrors

[BigDave]

My FreeNAS box is my backup and I'm going to make some changes to the pool.
Currently my two pools look like this.

Code:

[root@freenas] ~# zpool status
   pool: plumber
 state: ONLINE
  scan: scrub repaired 0 in 2h14m with 0 errors on Fri Jun  2 03:15:53 2017
config:

		NAME											STATE	 READ WRITE CKSUM
		plumber										 ONLINE	   0	 0	 0
		  raidz2-0									  ONLINE	   0	 0	 0
			gptid/9562a750-c341-11e6-a20a-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/c0a640f6-c3df-11e6-9356-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/18ed8eb6-f573-11e6-96fa-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/91cb5f7c-32a1-11e6-a5d4-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/1dc335b1-f551-11e6-acda-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/f7acbdd0-f5ab-11e6-9ee6-90e2ba0045fc  ONLINE	   0	 0	 0

errors: No known data errors

  pool: tanktest
 state: ONLINE
  scan: scrub repaired 0 in 0h0m with 0 errors on Fri Jun  2 16:15:35 2017
config:

		NAME											STATE	 READ WRITE CKSUM
		tanktest										ONLINE	   0	 0	 0
		  mirror-0									  ONLINE	   0	 0	 0
			gptid/d290bfd2-47d7-11e7-b000-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/d362dd3e-47d7-11e7-b000-90e2ba0045fc  ONLINE	   0	 0	 0
		  mirror-1									  ONLINE	   0	 0	 0
			gptid/f5671fee-47d7-11e7-b000-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/f69e4473-47d7-11e7-b000-90e2ba0045fc  ONLINE	   0	 0	 0
		  mirror-2									  ONLINE	   0	 0	 0
			gptid/110bd9f8-47d8-11e7-b000-90e2ba0045fc  ONLINE	   0	 0	 0
			gptid/12354a67-47d8-11e7-b000-90e2ba0045fc  ONLINE	   0	 0	 0

errors: No known data errors

My plan is to copy the data from "plumber" to "tanktest", verify the copy
of all the files is complete and use the original drives (as mirrors) to
expand the new pool and end up with six mirrored pairs (6-3TB and 6-4TB)
Please tell me if this move is stupid or my head is in that dark place...
Eventually I will replace the 3TB drives (one at a time) with 4TB and that
will be the end of it.

Having never used snapshots and replication before now, I'm needing some
verification that there is a proper amount of space to get this done. Anyone?

 

[Ericloewe]

If you want to transfer all the snapshots, you'll need pretty much what it's currently using (plus minus some details),

 

[Robert Trevellyan]

I'm needing some
verification that there is a proper amount of space to get this done.

It appears that you have roughly 8TB available on the destination and roughly 4TB in use on the source, so you should be all set.

 

[BigDave]

If you want to transfer all the snapshots, you'll need pretty much what it's currently using (plus minus some details),

Well the snapshots (after the first one) will take up no space because there will be no changes
made to the "plumber" pool. The system dataset resides in the boot pool and no one else has
anything other than read only access to the share. My thinking is correct, Yes?

 

[Robert Trevellyan]

Yes.

 

[Ericloewe]

Well the snapshots (after the first one) will take up no space because there will be no changes
made to the "plumber" pool.

I meant the current ones. If you can afford to get rid of some, you might save some space, too.

Well the snapshots (after the first one) will take up no space because there will be no changes
made to the "plumber" pool.

In that case why bother with a task? Just make the one you need manually.

 

[BigDave]

I meant the current ones. If you can afford to get rid of some, you might save some space, too.

I misunderstood but I get it now.
There are no current snapshots. The pool "plumber" has never been backed up in any way.
I created the new tanktest pool yesterday, for the specific purpose of backing up this data.

 

[Spearfoot]

@BigDave : I recently moved data between pools using replication, and found that I needed to create the datasets on the target pool before replicating the source to the target. If I didn't do this, the target pool ended up being 'Read Only'. YMMV...

 

[BigDave]

@BigDave : I recently moved data between pools using replication, and found that I needed to create the datasets on the target pool before replicating the source to the target. If I didn't do this, the target pool ended up being 'Read Only'. YMMV...

I actually just completed that very thing, but wrongly named the main dataset Share2 instead of Share

Of course now I'll have to delete Share2, cause AFAIK, you can't rename a dataset...

 

[JayG30]

You most certainly can rename a dataset.

As for read-only in replicated dataset you can see THIS thread. Not really a big issue though, it can be changed after.

Also, I don't think ZFS will redistribute the existing data optimally when you add the previous disks back to the pool. I think it only happens currently in initial write. Might not be anything you can do about it though.

 

[BigDave]

You most certainly can rename a dataset.

Yes, I did find out that renaming can be accomplished through the CLI
zfs rename tanktest/Share2 tanktest/Share