disk replacement howto

Any disk replacement should only be attempted by qualified service personnel, or
with the assistance of a qualified person. This document is only intended as an
introduction or reminder of the multiple software/hardware considerations that
can affect a safe and successful disk replacement. Full documentation should be
consulted when anything is unclear: http://docs.sun.com

When planning to replace a disk, determine the following:


1. Is the underlying hardware SCSI or fibre channel?
Is it a Storage Area Network(SAN) environment?

2. Is the disk under logical volume management?

2a. What software or hardware is being used for logical volume management?

2b. Is the volume a stripe(RAID 0), concatenation(RAID 0), mirror(RAID 1),
or RAID 5?

3. If the disk is not under volume management, are any partitions on the
disk mounted? Disks should not be removed while there is a potential for
I/O operations to occur to the disk.

4. Has the disk been allocated for RAW usage (i.e for a database)

5. Is the disk shared in a cluster environment? (Clusters have considerations
for disks which may contain quorum information.)
Cluster Note:
Details for cluster environments are outside of the scope of this document.
Reference Cluster documentation, or get assistance from Sun Support Personnel.

The primary types of logical volume management used with SUN storage:
Software Volume Management

JBOD (Just a Bunch of Disks)

Mostly used with: internal disks, Sun StorEdge[TM] A5000 family, Sun StorEdge
D1000, Sun StorEdge D240, Netra[TM] st D130, Sun StorEdge S1, Sun StorEdge
UniPack, Sun StorEdge MultiPack, and others.

Disks may be mounted individually without any logical volume management, and
would show up in mount or df -k output in the form:


Commands to check:

# mount
# df -k

NOTE: if a disk is being used for raw access by a data base, it will not appear in mount or df -k. Confirm this with a System administrator familiar with the system.

Solaris[TM] Volume Manager(SVM) a.k.a. Solstice DiskSuite[TM] (SDS) or Online: Disk Suite[TM] (ODS)

Mostly used with: any JBOD disks

Commands to check:

# df -k Shows volumes mounted as /dev/md/dsk/d#
# metastat Lists volume configuration, shows disk states.
# /usr/opt/SUNWmd/sbin/metastat SDS or ODS may require the full path to the command.

Veritas Volume Manager (VxVM) aka Sun Enterprise Volume Manager[TM] (SEVM)

Mostly used with: any JBOD disks, automatically licensed by Sun if an Sun StorEdge A5000 family array is attached to the system

Commands to check:

# df -k Shows volumes mounted as /dev/vx/dsk/ for rootdg,
or /dev/vx/dsk// for other disk groups.
# vxdisk list Lists disks, shows if they’re under VxVM control, shows their
current state. The "error" state indicates that the disk is
not under VxVM control.
# vxprint -ht Lists volume configuration.
# vxdiskadm Text-based menus for VxVM administration. Options 4 and 5 are
used to replace failed disks.

Hardware commands

Fibre channel(FC-AL or SAN) hardware may require additional commands to ensure
creation of proper device links to the disk world wide number(WWN). Fibre
channel disks require luxadm commands to remove the WWNs in the device path. In
Solaris 8 OS and later, adding / removing / replacing a SCSI disk drive,
requires the use of cfgadm(see cfgadm_scsi(1M) for more details). Consult the
documentation or man commands for further information.

Commands used:

# luxadm probe Lists arrays that use the SES
driver for array enclosure
# luxadm display Displays details of array or
# luxadm remove_device Removes device paths to a fibre channel disk
# cfgadm -x replace_device Starts the process of replacing
a SCSI disk drive in Solaris 8
OS and later.

Important documentation – more information on cfgadm:
Hardware Volume Management

Raid Manager 6 (RM6)

Used with: Sun StorEdge A1000, Sun RSM Array[TM] 2000, Sun StorEdge A3000, Sun StorEdge A3500FC

Features: Hardware RAID controller, managed from the host by GUI or command line.

Commands to check:

# /usr/lib/osa/bin/rm6 Launches management GUI for the
hardware raid controllers. The
Recovery Guru application in the GUI
is extremely useful for handling
failure events.
# /usr/sbin/osa/lad Lists c#t#d# of logical disks
available in hardware
# /usr/sbin/osa/drivutil -l List individual disks in logical disk
# /usr/sbin/osa/healthck -a Show array hardware status

Important documentation:
Version Matrix:

Sun StorEdge[TM] T3 and Sun StorEdge[TM] 6120 arrays

Used with : Sun StorEdge T3 and Sun StorEdge 6120 arrays

Features: Internal management by a hardware RAID controller which can be
accessed like a UNIX system through telnet or a serial service port.

Commands to check:

# telnet Username=root, password set by customer
# fru stat Field replaceable unit status
# fru list Field replaceable unit list
# vol stat Volume status
# vol list Volume list
# proc list Process list (Example: logical volume
construction or recovery)

Important documentation:

Sun StorEdge 3310, Sun StorEdge 3510 FC Array

Used with: Sun StorEdge 3310, Sun StorEdge 3510 FC Array

Features: Internal management by a RAID controller, which can be accessed
through an ethernet management port or serial port. Features a text terminal
menu interface. Can also be managed by a UNIX GUI or UNIX command line interface
from the host.

Commands to use:

For in-band management (through SCSI or FCAL interface)

# sccli Starts command line interface (cli) for array
sccli> show configuration Shows array components status
sccli> show disks display status of disks, note: faulted disks
do not appear in the listing.
sccli> show events display system log
sccli> show FRUs list fru’s; contains part numbers and serial numbers.
sccli> show logical-drives display LUN’s and LUN status; shows if spare
is configured and any failed disks.
sccli> show redundancy-mode displays controller status
sccli> help Displays cli commands available.

Manual: http://www.sun.com/products-n-solutions/hardware/docs/html/817-4951-10/

#ssconsole Launches the UNIX GUI for monitoring and

Leave a Reply

Your email address will not be published. Required fields are marked *