zfs discuss - Sep 2008 - What is the correct procedure to replace a non failed disk for another?

Hi,

can anybody describe the correct procedure to replace a disk (in a working OK
state) with a another disk without degrading my pool?

For a mirror I thought off adding the spare, you''ll get a three device
mirror. Let it resilver. Finally remove the disk I want.
But what would be the correct commands?
And what if I''ve got a pool consisting of multiple mirror
vdev''s?

And what about a raid-z or raid-z2 vdev? I can pull a disk and let the hotspare
take it''s place. But that degrades the pool. I want to mirror the two
disks and when done remove the source disk. This way I''ll never have a
degraded pool. Or am I asking for a new zpool feature?

Thanks,

Frederik
--
This message posted from opensolaris.org

On 3 Sep 2008, at 05:20, "F. Wessels" <wessels147 at yahoo.com>
wrote:
> Hi,
>
> can anybody describe the correct procedure to replace a disk (in a  
> working OK state) with a another disk without degrading my pool?
This command ought to do the trick:

zfs replace <pool> <old-disk> <new-disk>

The type of pool doesn''t matter.


Regards,
markm

Mark J. Musante wrote:
> 
> On 3 Sep 2008, at 05:20, "F. Wessels" <wessels147 at
yahoo.com> wrote:
> 
>> Hi,
>>
>> can anybody describe the correct procedure to replace a disk (in a  
>> working OK state) with a another disk without degrading my pool?
> 
> This command ought to do the trick:
> 
> zfs replace <pool> <old-disk> <new-disk>
Slight typo above, it''s zpool replace is the command

By the way what is the pool config, I assume you have a pool that 
supports this :-)

Once the disk is added, a resilver will occur, so do not take snapshots 
till it has finished, as the resilver will be restarted, this is fixed 
in snv_94 though.

Enda
> 
> The type of pool doesn''t matter.
> 
> 
> Regards,
> markm
> 
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

I''m pretty sure you just need the zpool replace command:
# zpool replace <poolname> <olddisk> <newdisk>

Run that for the disk you want to replace and let it resilver.  Once
it''s done, you can unconfigure the old disk with cfgadm and remove it.

If you have multiple mirror vdev''s, you''ll need to run the
command a few times.  I expect you can replace several drives at once but
I''ve not tried that personally.
--
This message posted from opensolaris.org

Gaah, my command got nerfed by the forum, sorry, should have previewed.  What
you want is:
# zpool replace poolname olddisk newdisk
--
This message posted from opensolaris.org

How would this work for servers that support only (2) drives, or systems 
  that are configured to have pools of (2) drives, i.e. mirrors, and 
there is no additional space to have a new disk, as shown in the sample 
below.

I still support lots of V490''s, which hold only (2) drives.

Thanks,

Jerry



Ross wrote:
> Gaah, my command got nerfed by the forum, sorry, should have previewed. 
What you want is:
> # zpool replace poolname olddisk newdisk
> --

On Wed, 3 Sep 2008, Jerry K wrote:
> How would this work for servers that support only (2) drives, or systems
>  that are configured to have pools of (2) drives, i.e. mirrors, and
> there is no additional space to have a new disk, as shown in the sample
> below.
You may be able to accomplish what you want by using an intermediate 
temporary disk and doubling the work (two replacements).  Perhaps the 
server supports USB so it can use an external USB drive as the initial 
replacement.  There is also the possibility of replacing the disk with 
a suitably sized disk file which is stored on some other server or an 
independent local filesystem with enough space.  You could access 
temporary storage on another server using iSCSI.  Server performance 
may suck while the inferior temporary device is in place.

Whatever you do, make sure that the intermediate storage is never any 
larger than the final device will be.

Bob
=====================================Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

Hello Bob,

Thank you for your reply.  Your final sentence is a gem I will keep.

As far as the rest, I have a lot of production server that are (2) drive 
systems, and I really hope that there is a mechanism to quickly R&R dead 
drives, resilvering aside.  I guess I need to do some more RTFMing into 
this.

Jerry K.


Bob Friesenhahn wrote:
> On Wed, 3 Sep 2008, Jerry K wrote:
> 
>> How would this work for servers that support only (2) drives, or
systems
>>  that are configured to have pools of (2) drives, i.e. mirrors, and
>> there is no additional space to have a new disk, as shown in the sample
>> below.
> 
> You may be able to accomplish what you want by using an intermediate 
> temporary disk and doubling the work (two replacements).  Perhaps the 
> server supports USB so it can use an external USB drive as the initial 
> replacement.  There is also the possibility of replacing the disk with a 
> suitably sized disk file which is stored on some other server or an 
> independent local filesystem with enough space.  You could access 
> temporary storage on another server using iSCSI.  Server performance may 
> suck while the inferior temporary device is in place.
> 
> Whatever you do, make sure that the intermediate storage is never any 
> larger than the final device will be.
> 
> Bob
> =====================================> Bob Friesenhahn
> bfriesen at simple.dallas.tx.us,
http://www.simplesystems.org/users/bfriesen/
> GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

2008/9/3 Jerry K <sun.mail.list at oryx.cc>:
> Hello Bob,
>
> Thank you for your reply.  Your final sentence is a gem I will keep.
>
> As far as the rest, I have a lot of production server that are (2) drive
> systems, and I really hope that there is a mechanism to quickly R&R
dead
> drives, resilvering aside.  I guess I need to do some more RTFMing into
> this.
If the drive is dead the pool is already in degraded mode. You simply
replace the failed
drive and tell zfs that it was replaced:

zpool replace pool device

Thanks for the replies.

I guess I misunderstood the manual:
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 different disk. ZFS recognizes this.

In the last paragraph it stated failed and physically removed etc. Also in the
first paragraph it stated resilver.

To summarize, "zpool replace" can replace a disk in any vdev type
without compromising redundancy. If the new disk fails (during resilvering) the
old pool remains in it''s original state. After resilvering the new disk
the old one gets detached. All this time the pool remains in it''s
original state. (off course no other factor kicks in)

PS. Why can''t I see the comments from Bob and Jerry and others made
after the last comment from Ross? I can see the comments in the text based site
at http://mail.opensolaris.org/pipermail/zfs-discuss/2008-September but not at
http://www.opensolaris.org/ at which I''m currently posting this
message.
--
This message posted from opensolaris.org