test

Platform Notes: Sun Enterprise 250 Server

Replacing a Faulty Hot-Pluggable Disk Drive

This section contains information on configuring your system to replace a disk drive while the power is on and the operating system is running.

The way in which you replace a faulty disk drive depends on the application you are using. Each application is different, but requires that you:

  1. Determine which disk drive is failing or has failed

  2. Remove the disk

  3. Add the replacement drive

  4. Reconfigure the operating environment.

In all cases you must stop any activity or applications on the disk; unmount it; physically remove the old drive and install the new one; and configure the Solaris environment to recognize the drive. Then you must configure your application to accept the new disk drive.

Prepare Spare Drives

If possible, prepare replacement disk drives in advance. Each replacement disk drive should be formatted, labeled, and partitioned the same as the disk it will replace. See the documentation for your application for instructions on how to format and partition the disk, and add that disk to your application.

Identifying the Faulty Disk Drive

Disk errors may be reported in a number of different ways. Often you can find messages about failing or failed disks in your system console. This information is also logged in the /usr/adm/messages file(s). These error messages typically refer to a failed disk drive by its physical device name (such as /devices/pci@1f,4000/scsi@3/sd@b,0) and its UNIX device instance name (such as sd11). In some cases, a faulty disk may be identified by its logical device name (such as c0t11d0). In addition, some applications may report a disk slot number (0 through 5) or activate an LED located next to the disk drive itself (see following figure).

Figure 2-2 Disk Slot Numbers and LED Locations

Graphic

In order to perform a disk hot-plug procedure, you need to know the slot number of the faulty disk (0 through 5) and its logical device name (for example, c0t11d0). If you know the disk slot number, it is possible to determine the logical device name, and vice versa. It is also possible to determine both the disk slot number and the logical device name from a physical device name (such as /devices/pci@1f,4000/scsi@3/sd@b,0).

To make the necessary translation from one form of disk identifier to another, see the Chapter Chapter 3, Mapping Between Logical and Physical Device Names . Once you have determined both the disk slot number and logical device name, you are ready to continue with this procedure.

Replacing the Disk Drive Within Your Application

Continue the disk replacement by following the instructions for your specific application.

UNIX File System (UFS)

The following procedure describes how to deconfigure a disk being used by one or more UFS file systems.

Caution - Caution -

These procedures should be performed only by a qualified system administrator. Performing hot-plug operations on an active disk drive can result in data loss if performed incorrectly.

  1. Type su and your superuser password.

  2. Identify activities or applications attached to the device you plan to remove.

    Commands to use are mount, showmount -a, and ps -ef. See the mount(1M), showmount(1M), and ps(1) man pages for more details.

    For example, where the controller number is 0 and the target ID is 11: 

    # mount | grep c0t11
/export/home1 on /dev/dsk/c0t11d0s2 setuid/read/write on
# showmount -a | grep /export/home1
cinnamon:/export/home1/archive
austin:/export/home1
swlab1:/export/home1/doc
# ps -f | grep c0t11
root  1225   450   4 13:09:58  pts/2   0:00 grep c0t11

    In this example, the file system /export/home1 on the faulty disk is being remotely mounted by three different systems--cinnamon, austin, and swlab1. The only process running is grep, which has finished.

  3. Stop any activity or application processes on the file systems to be deconfigured.

  4. Back up your system.

  5. Determine what file system(s) are on the disk: 

    # mount | grep cwtx

    For example, if the device to be removed is c0t11d0, enter the following: 

    # mount | grep c0t11
/export/home   (/dev/dsk/c0t11d0s7 ):   98892 blocks   142713 files
/export/home1  (/dev/dsk/c0t11d0s5 ):  153424 blocks   112107 files

  6. Determine and save the partition table for the disk.

    If the replacement disk is the same type as the faulty disk, you can use the format command to save the partition table of the disk. Use the save command in format to save a copy of the partition table to the /etc/format.dat file. This will allow you to configure the replacement disk so that its layout matches the current disk.

    Refer to the format(1M) man page for more information.

  7. Unmount any file systems on the disk.

    For each file system returned, type: 

     # umount file_system

    where file_system is the first field for each line returned in Step 5.

    For example: 

    # umount /export/home
# umount /export/home1

    Note -

    If the file system(s) are on a disk that is failing or has failed, the umount operation may not complete. A large number of error messages may be displayed in the system console and in the /var directory during the umount operation. If the umount operation does not complete, you may have to restart the system.

  8. Use the ssaadm replace_device command to take the device offline: 

    # ssaadm replace_device logical_device_name
ssaadm: warning: can't quiesce "/dev/rdsk/c0t11d0s2": I/O error
Bus is ready for the replacement of device
Replace device and reconfigure bus as needed
Press RETURN when ready to continue

    Here, logical_device_name is the full logical device name of the drive to be removed (/dev/rdsk/c0t11d0s2). You must specify slice 2, which represents the entire disk. Note that this command also accepts a physical device name as an alternative.

    You can safely ignore the warning message since the Enterprise 250 SCSI bus does not require quiescing.

  9. Remove the faulty disk drive and insert the replacement drive in its place.

    Refer to the Sun Enterprise 250 Server Owner's Guide for drive removal and replacement instructions.

  10. Press Return to complete the hot-plug operation.

    The ssaadm command brings the replacement drive back online.

  11. Verify that the device's partition table satisfies the requirements of the file system(s) you intend to re-create.

    You can use the prtvtoc command to inspect the label for your device. If you need to modify the label, use the format command. Refer to the prtvtoc(1M) and format(1M) man pages for more information. For example: 

    # prtvtoc /dev/rdsk/cwtxdysz

    If you have saved a disk partition table using the format utility and the replacement disk type matches the old disk type, then you can use the format utility's partition section to configure the partition table of the replacement disk. See the select and label commands in the partition section.

    If the replacement disk is of a different type than the disk it replaced, you can use the partition size information from the previous disk to set the partition table for the replacement disk. Refer to the prtvtoc(1M) and format(1M) man pages for more information.

    You have defined your disk's partition table and have identified the disk slice on which to build your UFS file system.

  12. Once you have selected a disk slice for your UFS file system, check and/or create a file system on the slice: 

    # fsck /dev/rdsk/cwtxdysz
# newfs /dev/rdsk/cwtxdysz

  13. Mount the new file system using the mount command: 

    # mount mount_point

    where mount_point is the directory on which the faulty disk was mounted.

The new disk is ready to be used. You can now restore data from your backups.

Solstice DiskSuite

The following procedure describes how to replace a disk in use by Solstice DiskSuite. Refer to the Solstice DiskSuite documentation for more information.

Caution - Caution -

These procedures should be performed only by a qualified system administrator. Performing hot-plug operations on an active disk drive can result in data loss if performed incorrectly.

  1. Back up your system.

  2. Type su and your superuser password.

  3. If possible, save the partition table for the disk you intend to replace.

    If the disk label can still be read, save the disk partitioning at this time.

    Note -

    Save all the disk partitioning information immediately after configuring metadevices or file systems for use when recovering from device failure later.

    Use the prtvtoc command to save the partition information. 

    # prtvtoc /dev/rdsk/cwtxdys0
 > saved_partition_table_file

    For example: 

    # prtvtoc /dev/rdsk/c0t11d0s0 > /etc/c0t11d0s0.vtoc

  4. Identify metadevices or applications using the device you plan to remove.

    For example: 

    # metadb | grep c0t11d0
# metastat | grep c0t11d0
# mount | grep c0t11d0

    Save the output of the commands to reconstruct the metadevice configuration after you replace the disk.

  5. Delete database replicas.

    If there are database replicas on the disk, these must be deleted. First record the size and number of database replicas on each slice; then delete them. 

    # metadb -d cwtxdysz

    For example: 

    # metadb -d c0t11d0s0

  6. Detach submirrors.

    If any slices of the disk are used by submirrors, the submirrors should be detached. For example: 

    # metadetach d20 d21

  7. Delete hotspares.

    If any slices are used by hotspare pools, remove them. Record the hotspare pools containing the slices; then delete them. For example: 

    # metahs -d all c0t11d0s1

  8. Terminate all other metadevice activity on the disk.

    Check metastat output for other slices of the disk used by metadevices that cannot be detached (stripes not in mirrors, etc.). These metadevices must be unmounted if they contain file systems, or they must otherwise be taken off line.

    Refer to the prtvtoc(1M) man page for more information.

  9. Unmount any file systems on the disk.

    Note -

    If the file system(s) are on a disk that is failing or has failed, the umount operation may not complete. A large number of error messages may be displayed in the system console and in the /var directory during the umount operation. If the umount operation does not complete, you may have to restart the system.

    For each file system returned, type: 

     # umount file_system

    where file_system is the first field for each line returned in Step 4.

    For example: 

    # umount /export/home
# umount /export/home1

  10. Use the ssaadm replace_device command to take the device offline: 

    # ssaadm replace_device logical_device_name
ssaadm: warning: can't quiesce "/dev/rdsk/c0t11d0s2": I/O error
Bus is ready for the replacement of device
Replace device and reconfigure bus as needed
Press RETURN when ready to continue

    Here, logical_device_name is the full logical device name of the drive to be removed (/dev/rdsk/c0t11d0s2). You must specify slice 2, which represents the entire disk. Note that this command also accepts a physical device name as an alternative.

    You can safely ignore the warning message since the Enterprise 250 SCSI bus does not require quiescing.

  11. Remove the faulty disk drive and insert the replacement drive in its place.

    Refer to the Sun Enterprise 250 Server Owner's Guide for drive removal and replacement instructions.

  12. Press Return to complete the hot-plug operation.

    The ssaadm command brings the replacement drive back online.

  13. Restore the disk partitioning.

    If you have saved the disk partitioning to a file, you may restore it with fmthard. For example: 

    # fmthard -s /etc/c0t11d0s0.vtoc  /dev/rdsk/c0t11d0s0

    If you have not saved the disk partitioning, use the format (1M) or fmthard(1M) command to repartition the disk.

  14. Replace any database replicas.

    For example: 

    # metadb -a -l 2000 -c 2 c0t11d0s0

  15. Reattach any submirrors.

    For example: 

    # metattach d20 d21

  16. Re-create hot spares for each hot spare pool that contained a slice on the new disk.

    For example: 

    # metahs -a hsp001 c0t11d0s1

  17. Fix any broken metadevices, using slices from the new disk.

    If the disk to be replaced had caused any metadevices to go into the maintenance state, these metadevices can be repaired by re-enabling the slices. 

    # metareplace -e mirror_or_RAID5_metadeice cwtxdysz

  18. Remount any file systems and restart any applications that were using metadevices that could not be taken off line. 

    # mount file_system

Refer to the Solstice DiskSuite documentation for more information.