How to Break the ZFS mirror on Solaris ?

This article is going to explain how to break the ZFS mirror on Solaris.I am not sure how this will be useful in real production environment. Because splitting can be possible only on mirrored zpool. Unlike veritas volume manager diskgroup split, ZFS is not going to divide the zpool. Its just remove the mirror and keep the copy on mirrored disk. For an example, once you split the mirrored zpool , you will be having two copy of same zpool data.(Existing zpool will keep one disk and new Zpool will be created using the mirror disk). This operation should not done when the zpool used by application or database. In a order to split the zpool, you need to halt the database and perform this operation.

[box type=”error”]

Mirroed root zpool can’t be split and its not supported.In ZFS, you no need to break the mirror disks for OS patching.All the ZFS root systems should use liveupgrade for OS patching and rollback operation.Note:Live upgrade commands are differnet in solaris 11(BE commands) compare to solaris 10 (LU commands)

[/box]
Here we will how to split the existing mirrored zpool.

1. Login to Solaris system where you want to break the mirrored data zpool .Check the zpool status.

root@UAAIS:~# zpool status MUA1
  pool: MUA1
 state: ONLINE
  scan: none requested
config:

        NAME        STATE     READ WRITE CKSUM
        MUA1        ONLINE       0     0     0
          mirror-0  ONLINE       0     0     0
            c8t2d0  ONLINE       0     0     0
            c8t3d0  ONLINE       0     0     0

errors: No known data errors
root@UAAIS:~#

2.Break the mirror using zpool split command.The below command is going to break the mirror of MUA1 zpool and its going to create a new zpool called MUA2 using the detached disk of MUA1 zpool.

root@UAAIS:~# zpool split MUA1 MUA2
root@UAAIS:~#

3.Check the zpool status of MUA1.You can see c8t3d0 disk has been removed from MUA1 zpool.

root@UAAIS:~# zpool status MUA1
  pool: MUA1
 state: ONLINE
  scan: none requested
config:

        NAME      STATE     READ WRITE CKSUM
        MUA1      ONLINE       0     0     0
          c8t2d0  ONLINE       0     0     0

errors: No known data errors
root@UAAIS:~#

4.MUA2 zpool will be in exported state. You check the status of exported zpool using below command.

root@UAAIS:~# zpool import
  pool: MUA2
    id: 231980248923509617
 state: ONLINE
action: The pool can be imported using its name or numeric identifier.
config:

        MUA2      ONLINE
          c8t3d0  ONLINE
root@UAAIS:~#

5. MUA2 zpool can be imported to this host ,but make sure that there is no conflict on the zfs datasets mountpoints. Since its just clone of MUA1, there is high possibility of using same mounts for datasets. In this case you need to modify the zfs dataset mount points manually.

root@UAAIS:~# zpool status MUA2
  pool: MUA2
 state: ONLINE
  scan: none requested
config:

        NAME      STATE     READ WRITE CKSUM
        MUA2      ONLINE       0     0     0
          c8t3d0  ONLINE       0     0     0

errors: No known data errors
root@UAAIS:~#
root@UAAIS:~# df -h |grep MUA
MUA1                   3.9G    31K       3.9G     1%    /MUA1
MUA2                   3.9G    31K       3.9G     1%    /MUA2
root@UAAIS:~#

6.The conflict occurs on the following situations.Here is the MUA1 dataset and static mount point.

root@UAAIS:~# df -h |grep MUA1
MUA1                   3.9G    31K       3.9G     1%    /MUA1
MUA1/unixarena         3.9G    31K       3.9G     1%    /unixarena
root@UAAIS:~#

MUA2 also having same dataset with same point since its just cloned copy it.When we try to import the zpool,You will get below warning.The zpool will be imported but dataset can’t be mounted since its used by MUA1 dataset.

root@UAAIS:~# zpool import MUA2
cannot mount 'MUA2/unixarena' on '/unixarena': mountpoint or dataset is busy
root@UAAIS:~# zpool list
NAME       SIZE  ALLOC   FREE  CAP  DEDUP  HEALTH  ALTROOT
MUA1      3.97G   174K  3.97G   0%  1.00x  ONLINE  -
MUA2      3.97G   148K  3.97G   0%  1.00x  ONLINE  -
rpool     15.6G  11.6G  3.99G  74%  1.00x  ONLINE  -

7.This conflict issue can be overcome by using the below set of commands.By using -R options, system will assume the new zpool as other system zpool and conflict can be avoided .

root@UAAIS:~# zpool status MUA1
  pool: MUA1
 state: ONLINE
  scan: resilvered 138K in 0h0m with 0 errors on Tue Feb 11 15:15:46 2014
config:

        NAME        STATE     READ WRITE CKSUM
        MUA1        ONLINE       0     0     0
          mirror-0  ONLINE       0     0     0
            c8t2d0  ONLINE       0     0     0
            c8t3d0  ONLINE       0     0     0

errors: No known data errors
root@UAAIS:~# zpool split -R /MUA2 MUA1 MUA2
root@UAAIS:~# zpool list
NAME       SIZE  ALLOC   FREE  CAP  DEDUP  HEALTH  ALTROOT
MUA1      3.97G   215K  3.97G   0%  1.00x  ONLINE  -
MUA2      3.97G   180K  3.97G   0%  1.00x  ONLINE  /MUA2
rpool     15.6G  11.6G  3.99G  74%  1.00x  ONLINE  -
testrepo  15.9G  13.8G  2.03G  87%  1.00x  ONLINE  -
root@UAAIS:~# df -h |grep MUA
MUA1                   3.9G    31K       3.9G     1%    /MUA1
MUA1/unixarena         3.9G    31K       3.9G     1%    /unixarena
MUA2                   3.9G    31K       3.9G     1%    /MUA2/MUA2
MUA2/unixarena         3.9G    31K       3.9G     1%    /MUA2/unixarena
root@UAAIS:~#

[starlist]

This feature is available only for mirrored zpools.
Database and application operations should be halted before performing the zpool split operation.
zpool can not be splited while resilvering aka mirroring in progress
If you the zpool is configured with two way mirroring, then after the split operation, your zpool will be having two disks and new pool will be created using the third disk.

[/starlist]
Hope this article is informative for you . Please leave a comment if you have any doubt on this .

Thank you for visiting UnixArena.

Reader Interactions

Leave a Reply Cancel reply