zfs discuss - Dec 2008 - zpool detach on non-mirrored drive

I''m using zfs not to have access to a fail-safe backed up system, but
to easily manage my file system. I would like to be able to, as I buy new
harddrives, just to be able to replace the old ones.  I''m very
environmentally concious, so I don''t want to leave old drives in there
to consume power as they''ve already been replaced by larger ones.
However, ZFS doesn''t currently let me detach a non-mirrored device.  Is
this planned for the future at all? I would imagine something like this:

zpool detach --non-mirrored dev ...
   detaching non-mirrored dev... wait for data to be copied

or even

zpool detach --non-mirrored dev ...
   detaching non-mirrored dev... there''s not sufficient space to be
able to remove dev

etc.

I hope my explanation is clear in that obviously the data would have to be
copied, possibly to the new drive I''ve added, as I want to remove the
old one.
-- 
This message posted from opensolaris.org

On Tue, Dec 16, 2008 at 7:10 PM, Daniel <nefar at hotmail.com> wrote:
> I''m using zfs not to have access to a fail-safe backed up system,
but to
> easily manage my file system. I would like to be able to, as I buy new
> harddrives, just to be able to replace the old ones.  I''m very
> environmentally concious, so I don''t want to leave old drives in
there to
> consume power as they''ve already been replaced by larger ones.
However, ZFS
> doesn''t currently let me detach a non-mirrored device.  Is this
planned for
> the future at all? I would imagine something like this:
>
> zpool detach --non-mirrored dev ...
>   detaching non-mirrored dev... wait for data to be copied
>
> or even
>
> zpool detach --non-mirrored dev ...
>   detaching non-mirrored dev... there''s not sufficient space to be
able to
> remove dev
>
> etc.
>
> I hope my explanation is clear in that obviously the data would have to be
> copied, possibly to the new drive I''ve added, as I want to remove
the old
> one.
>

vdev evacuation has been talked about, but there''s still no plans for
anyone
from Sun to implement it that I''m aware of as none of their large
enterprise
customers have requested it.

--Tim
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20081216/375e0d59/attachment.html>

tcook,

Thanks for your response. Well, I don''t imagine there would be a lot of
requests from enterprise customers with deep pockets. My impression has been
that OS is targetting the little guy though, and as such, this would really be a
welcome feature.
-- 
This message posted from opensolaris.org

Err... you can''t remove drives that are in use, but for what
you''re describing can''t you just use zfs replace and then
remove the old drive?
-- 
This message posted from opensolaris.org

>I''m using zfs not to have access to a fail-safe backed up system,
but to easily
>manage my file system. I would like to be able to, as I buy new harddrives,
just
>to be able to replace the old ones.  I''m very environmentally
concious, so I
>don''t want to leave old drives in there to consume power as
they''ve already been
>replaced by larger ones. However, ZFS doesn''t currently let me
detach a non-
>mirrored device.  Is this planned for the future at all? I would imagine
>something like this:
>
>zpool detach --non-mirrored dev ...
>   detaching non-mirrored dev... wait for data to be copied
>
>or even
>
>zpool detach --non-mirrored dev ...
>   detaching non-mirrored dev... there''s not sufficient space to be
able to remove dev
>

It is unfortunately that you ask this question after you''ve installed
the
new disks; now both the old and the new disks are part of the same zpool.

     zpool replace [-f] pool old_device [new_device]

         Replaces old_device with new_device. This is  equivalent
         to attaching new_device, waiting for it to resilver, and
         then detaching old_device.

         The size of new_device must be greater than or equal  to
         the minimum size of all the devices in a mirror or raidz
         configuration.

         new_device is required if the pool is not redundant.  If
         new_device  is not specified, it defaults to old_device.
         This form of replacement is  useful  after  an  existing
         disk  has  failed  and has been physically replaced.  In
         this case, the new disk may have the same /dev/dsk  path
         as  the  old  device,  even though it is actually a dif-
         ferent disk. ZFS recognizes this.


So in order to replace an old disk, you''d use:

	zpool replace pool olddisk newdisk

Casper

> It is unfortunately that you ask this question after
> you''ve installed the 
> new disks; now both the old and the new disks are
> part of the same zpool.
That''s awesome, I did not know that this would work. I''m glad
I made this post.

I actually have not yet replaced any drive, in fact, this very issue had made me
not span multiple drive. I installed zfs to deal with an outdated partition
issue, so I only have had the pool span partitions until now. But now that I
know I can use this as a solution, I will be less hesitant to add new drives :)
-- 
This message posted from opensolaris.org

Is it possible to do a replace on the root filesystem as well?
-- 
This message posted from opensolaris.org

Is it possible to do a replace on  / as well?
-- 
This message posted from opensolaris.org

Daniel,

You can replace the disks in both of the supported root pool
configurations:

- single disk (non-redundant) root pool
- mirrored (redundant) root pool

I''ve tried both recently and I prefer attaching the replacement disk to
the single-disk root pool and then detaching the old disk, using the
steps below. I like to provide confirmation between various steps. The
steps are included for replacing the root pool disk with zpool replace,
which works fine too, but doesn''t provide the safety nets that the
first
approach does.

Cindy

Before considering either approach, make sure you understand the boot
device pathnames of the current and new disk.

Replace root pool disk with zpool attach

1. Attach the new disk:

# zpool attach rpool old-disk new-disk

This step creates a mirrored root pool.

2. After the new disk is resilvered, apply the boot blocks to the new 
disk, like this:

# installboot -F zfs /usr/platform/`uname -i`/lib/fs/zfs/bootblk new-disk

3. Verify that you can boot from the new disk.

4. If the system boots from the new disk, detach the old disk:

# zpool detach rpool old-disk

5. Set up the system to boot automatically from the new disk,
either by using the eeprom command, the setenv from the SPARC
boot prom, or configure the PC BIOS.

Replace root pool disk with zpool replace

1. Replace the root pool disk:

# zpool replace rpool c1t10d0s0 c4t0d0s0

2. Check the status of the disk replacement:

# zpool replace rpool c1t10d0s0 c4t0d0s0
# zpool status rpool
   pool: rpool
  state: ONLINE
status: One or more devices is currently being resilvered.  The pool will
         continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
  scrub: resilver in progress for 0h0m, 4.73% done, 0h2m to go
config:

         NAME           STATE     READ WRITE CKSUM
         rpool          ONLINE       0     0     0
           replacing    ONLINE       0     0     0
             c1t10d0s0  ONLINE       0     0     0
             c4t0d0s0   ONLINE       0     0     0


3. When the replacement and resilvering are complete, add the boot
blocks to the  new disk:
# zpool status
   pool: rpool
  state: ONLINE
  scrub: resilver completed after 0h4m with 0 errors on Fri Dec 12 
10:32:40 2008
config:

         NAME        STATE     READ WRITE CKSUM
         rpool       ONLINE       0     0     0
           c4t0d0s0  ONLINE       0     0     0

# installboot -F zfs /usr/platform/`uname -i`/lib/fs/zfs/bootblk
c4t0d0s0

4. Confirm that you boot from the new disk.

5. Set up the system to boot automatically from the new disk,
either by using the eeprom command, the setenv from the SPARC
boot prom, or configure the x86 BIOS.

Daniel wrote:
> Is it possible to do a replace on  / as well?

Cindy,

This is helpful! Thank you very much :)
-- 
This message posted from opensolaris.org

I have only one question: why? ZFS is a great filesystem, but, why we
can''t shrink a zpool? This is a very important feature to ZFS adoption
in enterprises, or not?  Is it at least a planned feature?

Thanks,

Joel

On Thu, Dec 18, 2008 at 3:45 PM, Daniel <nefar at hotmail.com>
wrote:
> Cindy,
>
> This is helpful! Thank you very much :)
> --
> This message posted from opensolaris.org
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>


-- 
Joel Cesar Zamboni
joel.zamboni at gmail.com

Actually, I think it''s smaller companies and home users who want this
most.  Sun target enterprises with their storage kit which tends to be servers
that are fully populated with drives.  Look at the thumper for example - a 4U
server that already comes with 48 drives.  You''re not going to be
adding or removing drives from that, it''s pretty much a drop in storage
box.

There has been a lot of talk over the last year about making it possible to
remove drives, and I believe Sun are working on the code to do this.  It sounds
like it''s quite a complicated thing to do however, so it''s
likely to be a good while before it happens.

For now it just needs to be accepted as one of the limitations of ZFS. 
It''s easy enough to work around when you know about it.
-- 
This message posted from opensolaris.org