Expand my zpool

[zeroARMY]

I was never a ZFS or Solaris expert, but with some help a couple years ago I created a 12 drive RAID-Z2 in OpenIndiana. The pool was shared out with SMB.

What I would like to do is add another 12 drive RAID-Z2 to the pool. This is how it currently looks:

With 12 new drives inserted, what would be the commands to expand the pool? Does it add to tank, or become a separate listing? Would the SMB share show the total increased amount?

Also after expanding the zpool, does ZFS rebalance the data between all 24 drives? What I worry about is infant mortality with the new drives though. It would be nice if the drives could just run for a month without rebalancing, so I could weed out early-life issues.

Thanks for any help

 

[bleomycin]

See here: http://docs.oracle.com/cd/E19253-01/819-5461/gazgw/index.html

The command in your case assuming you would like to add the new drives as raidz2 as well is:

zpool add tank raidz2 drive drive drive.....

If you are worried about infant mortality of the drives before adding them how about running a few passes of badblocks on them each individually first? Also, I believe that once added to the existing pool data will not be balanced over to the new disks, it will remain where it is.

 

[danswartz]

Correct, rebalancing does not happen automatically. If files are written to, modified data should be balanced. If you have large, static files, you can do something like:

cp foo bar
rm foo
mv bar foo

 

[Hameno]

See here: http://docs.oracle.com/cd/E19253-01/819-5461/gazgw/index.html

The command in your case assuming you would like to add the new drives as raidz2 as well is:

zpool add rpool raidz2 drive drive drive.....

If you are worried about infant mortality of the drives before adding them how about running a few passes of badblocks on them each individually first? Also, I believe that once added to the existing pool data will not be balanced over to the new disks, it will remain where it is.

It should be
zpool add tank raidz2 drive drive drive.....
right?

 

[brutalizer]

Correct, rebalancing does not happen automatically. If files are written to, modified data should be balanced. If you have large, static files, you can do something like:

cp foo bar
rm foo
mv bar foo

Or you could add your 12 new disks, and after that rebalance by sending(copying) each zfs snapshot to your pool.

 

[bleomycin]

It should be
zpool add tank raidz2 drive drive drive.....
right?

D'oh! I read misread it. Thanks, I corrected my post as well.

 

[zeroARMY]

Correct, rebalancing does not happen automatically. If files are written to, modified data should be balanced. If you have large, static files, you can do something like:

cp foo bar
rm foo
mv bar foo

So if I only have 500GB of space left on the original vdev in the zpool, then I add 1TB of data, 500 goes to the new and 500 to the old? What happens when the original runs out of space without rebalancing?

 

[Windreaper]

So if I only have 500GB of space left on the original vdev in the zpool, then I add 1TB of data, 500 goes to the new and 500 to the old? What happens when the original runs out of space without rebalancing?

To my best knowledge zfs wouldn't rebalance the data but would favor the larger vdev for new writes, correct me if I'm wrong.

 

[danswartz]

You are correct.

 

[SirMaster]

I recommend adding -n to your zpool add command to perform a simulation of the add operation.

It will then print out what the will array look like after the add.
Then just remove the -n to actually do it.

Yes re-balancing will NOT happen. I would only worry about balancing if you really cared about performance, otherwise I don't think it's a big problem.

Maybe just copy 1TB of your existing data to the new folder, delete the original, and move it back to free up some space on the old vdev.