PRIMECLUSTER Global Disk Services Configuration and Administration Guide 4.1 (Solaris(TM) Operating System)
Contents PreviousNext
Appendix A General Notes> A.2 Important Points

A.2.31 Swapping FC-AL Internal Disks

If FC-AL internal disks are swapped, the device numbers of the replacement disks may change. For this reason, FC-AL internal disks cannot be swapped through the procedures for physical disk swapping with [Swap Physical Disk] and [Restore Physical Disk] in the GDS Management View or with the sdxswap command. For FC-AL internal disk swapping, remove the relevant disk from the class once and register the replacement with the class again.

The following example shows the procedures for swapping an FC-AL internal disk connected to a mirror group for another. In this example, the swapped disk is a mirrored disk and removal of volumes from the disk prior to swapping is unnecessary. If the swapped disk is not a mirrored disk, such as a single disk and a disk whose highest level group is not a mirror group, volumes must be removed from the disk in advance. If volume removal is required, back up volume data according to need, remove the volumes, remove the disk from the class, register the replacement with the class, then re-create volumes and restore the data.

  1. Check GDS I/O error messages output in the /var/adm/messages file for the device name (devname) of the failed disk.


    To search the GDS I/O error messages, use the string "sfdsk" as a key. In the following example, the device name is ssd@w210000203716ee10.

    WARNING: sfdsk: write error on mirror slice:
    volume info: devno(maj,min)=0x03f4002e(253,46)
    device info: devno(maj,min)=0x01d8001e(118,30)
    devname=/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0 (ssd3)
    error info: blk in disk=17308981, blk in slice=5232448
    length=8192, resid=8192, errno=5


  2. Check the physical disk name of the swapped disk (failed disk).

    For the argument of the grep command, specify the device name confirmed in step 1.

    In the following example, the physical disk name is c2t4d0.

    # ls -l /dev/rdsk | grep ssd@w210000203716ee10
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s0 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:a,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s1 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:b,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s2 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:c,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s3 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:d,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s4 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:e,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s5 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:f,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s6 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:g,raw
lrwxrwxrwx   1 root   root      78 Feb  3 18:49 c2t4d0s7 -> ../../devices/sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w210000203716ee10,0:h,raw


  3. Check the disk name, class name, and group name of the swapped disk.

    For these names, execute the sdxinfo -D command and check the NAME, CLASS, GROUP fields of the row that shows the physical disk name confirmed in step 2 in the DEVNAM field.

    In the following example, for physical disk c2t4d0, the disk name is Disk2, the class name is Class1, and the group name is Group1.

    # sdxinfo -D
    OBJ     NAME    TYPE   CLASS   GROUP  DEVNAM  DEVBLKS  DEVCONNECT      STATUS
------ ------- ------ ------- ------- ------- -------- -------------- --------
disk    Disk1   mirror Class1  Group1 c0t0d0  17682084  Node1          ENABLE
disk    Disk2   mirror Class1  Group1 c2t4d0  17682084  Node1          ENABLE


  4. Disconnect the swapped disk from the group.


    With the -c, -g, and -d options of the sdxdisk -D command, specify the class name, the group name, and the disk name confirmed in step 3.

    # sdxdisk -D -c Class1 -g Group1 -d Disk2


  5. Check if the swapped disk was disconnected from the group.

    For checking, execute the sdxinfo -D command and check if the asterisk (*) is displayed in the GROUP field of the row that shows the physical disk name of the disconnected disk in the DEVNAM field.

    # sdxinfo -D
OBJ    NAME   TYPE    CLASS   GROUP   DEVNAM  DEVBLKS   DEVCONNECT     STATUS
----- ------- ------- ------- ------- ------- --------- -------------- -------
disk   Disk1  mirror  Class1  Group1  c0t0d0  17682084  Node1          ENABLE
disk   Disk2  mirror  Class1  *       c2t4d0  17682084  Node1          ENABLE


  6. Remove the swapped disk from the class.


    With the -c and -d options of the sdxdisk -R command, specify the class name and the disk name confirmed in step 3.

    # sdxdisk -R -c Class1 -d Disk2


  7. Check if the swapped disk was removed from the class.

    For checking, execute the sdxinfo -D command and check if information of the removed disk is not displayed.

    # sdxinfo -D
OBJ     NAME    TYPE    CLASS   GROUP  DEVNAM  DEVBLKS  DEVCONNECT    STATUS
------ ------- ------- ------- ------- ------- -------- ------------- -------
disk    Disk1   mirror  Class1  Group1  c0t0d0 17682084  Node1        ENABLE


  8. Stop the system.

    # shutdown -y -i0 -g0


  9. Swap disks.

    Ask your local customer support (CE) for disk unit swapping.

    The CE will swap disk units according to the following procedures.

    1. Take a copy of the WWN (World Wide Name) of the swapped disk.

    2. Check the auto-boot? variable setting of OpenBoot, and if the auto-boot? variable value is "true", change the value to "false."

    3. Turn off power of the main unit.

    4. Swap disks.

    5. Turn on power of the main unit.

    6. When the auto-boot? value was changed to "false" in step 2, set it back to "true."


  10. Restart the system.


    An FC-AL disk has been assigned a unique WWN (World Wide Name) and the Solaris(TM) operating System uses this WWN as management information. For the Solaris(TM) operating System to recognize the WWN of the replacement disk, the system must be rebooted with the boot -r command.

    ok boot -r


  11. Together with the CE, check if the replacement is recognized by the Solaris(TM) operating System and check its physical disk name.

    In the following example, the WWN is 0020370e35e5 and the physical disk name is c2t4d0.

    # format
Searching for disks...done


AVAILABLE DISK SELECTIONS:
       0. c2t4d0 <sfdsk type1.0; SUN9.0G cyl 4924 alt 2 hd 27 sec 133>  00100004
          /sbus@2,0/SUNW,socal@d,10000/sf@1,0/ssd@w21000020370e35e5,0
    Specify disk (enter its number):


    Disks managed by GDS are not displayed with the format command.


  12. Check if the device special files of the replacement exist.

    For the argument of the ls command, specify the physical disk name confirmed in step 11 attaching the asterisk (*) to the end.

    # cd /dev/rdsk
# ls -lL c2t4d0*
crw-r-----   1 root     sys      118, 32 Feb  3 17:09 c2t4d0s0
crw-r-----   1 root     sys      118, 33 Feb  3 13:21 c2t4d0s1
crw-r-----   1 root     sys      118, 34 Feb  3 17:09 c2t4d0s2
crw-r-----   1 root     sys      118, 35 Feb  3 13:21 c2t4d0s3
crw-r-----   1 root     sys      118, 36 Feb  3 13:21 c2t4d0s4
crw-r-----   1 root     sys      118, 37 Feb  3 13:21 c2t4d0s5
crw-r-----   1 root     sys      118, 38 Feb  3 13:21 c2t4d0s6
crw-r-----   1 root     sys      118, 39 Feb  3 15:35 c2t4d0s7


  13. If the relevant system is a cluster system, re-create the resource of the replacement according to the following procedures.

    If not, go on to step 14).

    1. Check the resource of the original disk.

      Check if the resource of the original disk c2t4d0 connected to node Node1, confirmed in step 2, is displayed with the clgettree command.

      # /etc/opt/FJSVcluster/bin/clgettree
Cluster 1 cluster
                 Domain 2 Domain0
~
                   Node 3 Node1 ON
~
                           DISK 33 c2t4d0 UNKNOWN 
~


    2. Delete the original disk resource.

      # /etc/opt/FJSVcluster/bin/cldeldevice -n c2t4d0


    3. Check if the original disk resource was deleted.

      For checking, check if the resource confirmed in step 1 is not displayed with the clgettree command.

      # /etc/opt/FJSVcluster/bin/clgettree
Cluster 1 cluster
            Domain 2 Domain0
~
                    Node 3 Node1 ON
~


    4. Create the resource of the replacement.

      Execute the following command on any node that composes the cluster system.

      # /etc/opt/FJSVcluster/bin/clautoconfig -r


    5. Confirm that the resource of the replacement was created.

      For checking, check if the resource of replacement c2t4d0 confirmed in step 11 is displayed with the clgettree command.

      # /etc/opt/FJSVcluster/bin/clgettree
Cluster 1 cluster
           Domain 2 Domain0
~
                   Node 3 Node1 ON
~
                           DISK 44 c2t4d0 UNKNOWN
~


  14. Register the replacement with the class.

    With the -c option of the sdxdisk -M command, specify the class name confirmed in step 3. With the -d option, specify the physical disk name confirmed in step 11 and the disk name confirmed in step 3.

    # sdxdisk -M -c Class1 -d c2t4d0=Disk2


  15. Confirm that the replacement was registered with the class.

    # sdxinfo -D
OBJ     NAME   TYPE   CLASS   GROUP   DEVNAM  DEVBLKS   DEVCONNECT  STATUS
------ ------- ------ ------- ------- ------- -------- ------------ -------
disk    Disk1  mirror Class1  Group1  c0t0d0  17682084  Node1       ENABLE
disk    Disk2  mirror Class1  *       c2t4d0  17682084  Node1       ENABLE


  16. Connect the replacement to the group.

    With the -c, -g, and -d options of the sdxdisk -D command, specify the class name, the group name, and the disk name confirmed in step 3.

    # sdxdisk -C -c Class1 -g Group1 -d Disk2


  17. Confirm that the replacement was connected to the group.

    # sdxinfo -D
OBJ     NAME   TYPE   CLASS   GROUP   DEVNAM  DEVBLKS  DEVCONNECT    STATUS
------ ------ ------ ------- ------- -------- -------- ------------- -------
disk    Disk1  mirror Class1  Group1  c0t0d0  17682084  Node1         ENABLE
disk    Disk2  mirror Class1  Group1  c2t4d0  17682084  Node1         ENABLE


  18. When the replacement is connected to the group, resynchronization copying for the disk is performed.

    If the resynchronization copying is in process, with the sdxinfo -S command, COPY is displayed in the STATUS field of the row that shows the disk name of the replacement. If the resynchronization copying is complete, ACTIVE is displayed in the STATUS field.

    # sdxinfo -S
    OBJ    CLASS   GROUP   DISK   VOLUME  STATUS
----- ------- ------- ------- ------- --------
slice  Class1  Group1  Disk1  Volume1  ACTIVE
slice  Class1  Group1  Disk1  Volume2  ACTIVE
slice  Class1  Group1  Disk2  Volume1  COPY
slice  Class1  Group1  Disk2  Volume2  COPY


Contents PreviousNext
All Rights Reserved, Copyright(C) FUJITSU LIMITED 2005