Migration of datasets with rsync not working as expected

[Todd Nine]

Hey all,
I recently migrated from an old volume of USB drives to a new volume I created of internal storage drives. My goal is to keep my system exactly the same, just move it to a new volume. I performed the following.

1. Set up the new drives as "internal" mount point
2. Stop all sharing services and jails
3. Rsync with this command
   Code:

   /usr/local/bin/rsync -avzh --delete /mnt/origdisk/ /mnt/internal/

4. Go to System-> System Dataset and set the the dataset pool to use the new "internal" point from "origdisk"
5. Go to sharing, and change all paths from /mnt/origdisk to /mnt/internal
6. Go to jails, and change all mount points from /mnt/origdisk to /mnt/internal
7. Detach origdisk volume.
8. Reboot to verify the system functions as expected

So far, all of my data is present, and working properly. However, the datasets I created in /mnt/origdisk didn't migrate to /mnt/internal. The files are exactly the same. I assumed since I rsynced everything from /mnt/origdisk to /mnt/internal, these settings would be copied. However they have not. What files do I need to copy to get the datasets? See the attached screenshot.

 

[danb35]

rsync will not move datasets. If you want an exact duplicate of the old pool's contents, you need to use zfs replication instead.

 

[ian351c]

Or create your datasets, then rsync.

Either way, rsync works at a file level, not a dataset or pool level.

 

[Todd Nine]

Thanks for the reply guys. I've done the following.

1. In my /mnt/internal point, create a "jails" dataset
2. Set up 10 minute snapshots and replication. I use the localhost ssh, and point to internal/jails in my replication
3. Let the replication run

However, at the end, I'm left with /mnt/internal/jails/jails, not /mnt/internal/jails as expected. See the attached screenshots. What am I setting incorrectly? The documentation in the help icon on the destination says that is should the destination dataset I want the data copied into. Should the destination just be "internal"?

 

[ian351c]

Just make your remote destination /internal. That should work. I'm more familiar with doing this via CLI rather than the GUI.

 

[Todd Nine]

Hey Ian,
Do you have any pointers to the documentation for the CLI? I'm a software engineer, and I write a distributed database, so I'm very comfortable with the terminal. I'd prefer to use it over the gui :)

 

[ian351c]

I'd start here: http://www.freebsd.org/cgi/man.cgi?...D+Ports+10.1-RELEASE&arch=default&format=html

 

[Todd Nine]

Sorry, to be clear I was talking about the zfs sync, so that I can retain my snapshots. I currently have 4 minute snapshots on all my data sets, and replication via localhost ssh. I wouldn't find a way to do local zfs replication without the need to go through an ssh loopback. I googled for some documentation, but it doesn't seem to be as straightforward as running rsync between the 2 mount points.

 

[ian351c]

Ah. In that case this is probably more what you're looking for: http://docs.oracle.com/cd/E18752_01/html/819-5461/gbchx.html

 

[Todd Nine]

So good news, I've managed to test replication successfull using this command

Code:

zfs send -R origdisk/jails@manual-2015-01-15 | zfs recv internal/jails

Bad news, it copies every snapshot that was taken on the 5 minute interval when I was trying to do near realtime replication. I have over 1,700 of these snapshots across the drive. I attempted to delete all these automatic snapshots with this command.

Code:

zfs list -t snapshot -o name | grep \\-2w | xargs -n 1 zfs destroy -d

Which does the following.

1. list all snapshots by name
2. Grep for all snapshots that have a "-2w" in them, since they are auto 2 week expiration snapshots
3. Delete them

This worked for around 1000 of the snapshots. Now when I run it, I'm stuck at 641 remaining snapshots. I can't remove them, there isn't an error. It just appears to succeed, and never disappears from the list of snapshots. Any ideas? I want to remove all the extra snapshots before I do the manual sync, so that I'm not retaining them.

Thanks for all the help!
Todd

 

[Todd Nine]

Found the issue, some of the snapshots were being held. I ran the following command to release the holds.

Code:

 nohup zfs list -t snapshot -o name | grep \\-2w | sort -nr | xargs -n 1 zfs holds -r | grep \\-2w | awk '{print $2 " " $1}' | xargs -n 2 zfs release -r

Which does the following

1. List all snapshots
2. Grep for them with the -2w in them, again from the 2 week
3. Reverse sort them newest to highest
4. invoke zfs holds -r to find all holds on the snapshot
5. Grep that holding snapshot name line only
6. Print the tag first, with a space, then the snapshot name
7. Pass those 2 arguments to zfs release to release the holds

After that I ran this command again to remove the snapshots

Code:

zfs list -t snapshot -o name | grep \\-2w | xargs -n 1 zfs destroy -d

This cleaned up all my snapshots.