The following description shows resolutions to problems relevant to system disks on which the file systems / (root), /usr, and /var operates.
In one of the following circumstances, take action as indicated for the relevant situation.
(2) System can be booted, but the system disk data is invalid.
(3) System cannot be booted (Failure of original boot disk).
(4) System cannot be booted (Failure in all boot disk devices).
(6) System cannot be booted (Incorrect boot disk device name).
(7) System cannot be booted in a ZFS Boot Environment (Abnormality of root class private slice).
(9) The system cannot be booted (Alias name error of the boot device).
(1) Create backup of root file system.
Explanation
Mirroring the system disk will protect data in case one of the physical disks crashes. However, to restore data damaged due to a critical fault as a result of multiple failures and so on or by an operational mistake, create backup data in advance.
Resolution
(2) System can be booted, but the system disk data is invalid.
Explanation
For some reason, the system disk data is invalid. You must restore data from the backup data created in advance.
Resolution
For ZFS boot environment, see "6.2.2 Restoring." For UFS boot environment, see "6.3.2 Restoring (When the System Can Be Booted)."
(3) System cannot be booted (Failure of original boot disk).
Explanation
Even when you have the system disk mirrored, you may experience unsuccessful booting. For example, you may encounter such a problem when the file accessed during the booting process in the boot disk is physically damaged, or when a disk is swapped improperly.
If the console message of the unsuccessful boot disk indicates such is the case, try booting from the other boot disk, i.e. the mirror disk.
Resolution
1) Confirm the name of the mirror disk. Use the OpenBoot ok prompt shown below.
ok printenv boot-device |
In this example, the original disk name is /pci@1f,4000/scsi@3/disk@0,0:a, which is indicated on the left. The mirror disk is /pci@1f,4000/scsi@3/disk@1,0:a. When mirroring is configured with three-way or more multiplexing, multiple mirror disks exist.
2) Boot from the mirror disk.
ok boot /pci@1f,4000/scsi@3/disk@1,0:a |
3) If booting is successful, find out what caused the original disk to fail, and recover by swap the disk.
See
For details on disk swap, see "D.8 sdxswap - Swap disk" and resolutions described in "5.3.4 Disk Swap."
When the procedures above cannot resolve the problem, causes other than data errors in the original boot disk unit may exist, or data errors may also exist in the mirror disk.
If it would appear that data errors also exist in the mirror disk, see "(5) System cannot be booted (Boot disk data damage).."
(4) System cannot be booted (Failure in all boot disk devices).
Note
This troubleshooting is intended for the UFS boot environment. If a system cannot be booted in the ZFS boot environment, see the following:
When swapping physical disks or restoring a data of a system disk is necessary, see "6.2.2 Restoring."
When there is no error of physical disks and data in a volume, see "(7) System cannot be booted in a ZFS Boot Environment (Abnormality of root class private slice)."
Explanation
Booting may be failed even a system disk has been mirrored. If you want to check the restoring procedure when all the boot disks are physically failed (or assumed to be failed), perform the following procedure for restoring as all the data saved in a disk will be deleted.
Although it is unlikely to consider the situation where the mirrored physical disks would physically fail simultaneously, this section gives an explanation by assuming the worst case where you want to check the restoring procedure before performing.
For the following procedure, swap all the physical disks registered in a root class and restore the data based on collected backups. Then, re-construct the mirroring configuration.
Also, this section explains how to perform the mirroring cancellation to boot the system without swapping a physical disk or restoring data of the system disk by assuming that booting cannot be performed by some abnormality other than a physical disk failure.
For SAN Boot environment, the example for this procedure, a device name, a device format and a path are different.
For normal hard disk: /dev/[r]dsk/cXtXdXsX
For MPLB disk: /dev/FJSVmplb/[r]dsk/mplbXsX
Resolution
[Before the operation]
Before the operation and at the time of configuring a system, perform the following tasks.
1) Check the partition configuration and the cylinder size of a system disk. Then make a note of them.
This task must be performed before the system disk mirroring configuration. Execute the following command and take note of the underlined information.
Example: When the system disk is c4t50000394882261A0d0.
# format c4t50000394882261A0d0 format> verify Primary label contents: Volume name = <00000000> ascii name = <TOSHIBA-MBF2600RC-3706 cyl 64986 alt 2 hd 27 sec 668> pcyl = 64988 |
2) Check if MPxIO is enabled for the system disk and record the result.
This step must be completed before setting up mirroring for the system disk.
MPxIO is enabled, if the /scsi_vhci directory is in the logical path of the displayed boot disk.
Example: If the system disk is c0t50000394882261A0d0
# format c0t50000394882261A0d0 |
3) Collect backups of root file system and so on.
It is necessary to perform this operation while system disks are in mirroring state.
For the method of collecting backups, see "6.3.1 Backing Up."
Note
To prepare for unexpected failures, make backups regularly.
For Solaris 10, it is necessary to choose the OS to be booted when restoring depending on the kernel patch number. Therefore, when collecting backups, make a note of the patch application status by using the following command.
# /usr/bin/showrev -p |
4) Check the name of a root class.
Names of root classes will be used in the subsequent procedures. Check names in beforehand because it is difficult to check them when the system cannot be booted due to a failure in all boot disk devices.
Execute sdxinfo -C. The NAME field in the line where "root" is output in the TYPE field is the root class name. In the following example, the root class name is "RootClass."
# sdxinfo -C |
5) Check the physical disk name of root class.
Physical disk names of root class will be used in the subsequent procedures. Check the physical disk names beforehand because it is difficult to check them when the system cannot be booted due to a failure in all boot disk devices.
Execute sdxinfo -D. Specify the name of the root class that was checked using step 4) with -c option. The DEVNAM field is the physical disk name. In the following example, the physical disk names of root class are "c4t50000394882261A0d0" and "c4t5000039488226254d0."
# sdxinfo -D -c RootClass
OBJ NAME TYPE CLASS GROUP DEVNAM DEVBLKS DEVCONNECT STATUS
------ ------- ------ ------- ------- ------- -------- ---------------- -------
disk rootDisk0001 mirror RootClass rootGroup c4t50000394882261A0d0 1172087496 node1 ENABLE
disk rootDisk0002 mirror RootClass rootGroup c4t5000039488226254d0 1172087496 node1 ENABLE |
[Restoring]
1) Swap all the physical disks registered in the root class.
Swapping operation is performed according to a defined procedure for each device.
Note
When restoring to the disk that was used in GDS before without swapping a physical disk, boot the system from the CD-ROM device (or network) and restore all the physical disks registered in a root class to the initial status. For the method of booting from the CD-ROM device (or network), see step 2).
# format c0t0d0 |
2) Boot the system from a CD-ROM device (or the network).
For the procedure, see the Solaris manual.
Note
To restore Solaris 10, to which the kernel patch 137111-08 or earlier has been applied, the version of an OS booted from a CD-ROM device (also network or emergency startup disk for restoration) must correspond to the version of the OS, to which the kernel patch 137111-08 or earlier has been applied, you are restoring. To boot from a CD-ROM device, use the medium of Solaris 10 5/08 or earlier.
To restore Solaris 10, to which the kernel patch 137137-09 or later has been applied, the version of an OS booted from a CD-ROM device (also network or emergency startup disk for restoration) must correspond to the version of the OS, to which the kernel patch 137137-09 or later has been applied, you are restoring. To boot from a CD-ROM device, use the medium of Solaris 10 10/08 or later.
To restore Solaris 11, the version of an OS booted from a CD-ROM device (also network or emergency startup disk for restoration) must correspond to the version of the OS you are restoring.
In a system where mirroring is conducted by GDS, booting from a CD-ROM device (also network) and performing an illegal operation may disrupt the mirroring environment. Be sure to follow the procedure described here and to avoid operations not provided in this procedure.
When mounting the file system on a physical slice at the time of recovery, mounting as read-only is recommended to write-lock.
See
For details on step 2) to 5), see the Solaris manual.
3) Restore the system disk to the partition configuration confirmed in step 1) in [Before the operation].
Note
For the disk of Expansion File Unit connected to a 6Gbps SAS card, or the internal disks of SPARC M12/M10 and SPARC T4-4/T4-2/T4-1/T3-4/T3-2/T3-1 servers, the device name that is displayed in the format command differs with the device name before swapping because the physical disk was swapped in step 1). Execute format command to set the partition of the device that was output.
# format |
4) Restore backup data collected to a tape medium and so on to each physical slice.
In the following example, restore the data of each file system based on the backup data collected by using the ufsdump(1M) command in step 3) in [Before the operation].
Restore the data from the tape medium of tape device /dev/rmt/0 by using the ufsrestore(1M) command.
The following is a restore example of s0 slice. If other slices exist other than s0, restore also those slices.
# newfs /dev/rdsk/c0t50000394980066DEd0s0 |
5) Install the boot block to the root partition of the system disk.
When executing installboot command, the slice which includes /usr file system needs to be mounted.
If /usr is included in / (root) file system:
# mount /dev/dsk/c0t50000394980066DEd0s0 /mnt |
If /usr is a different file system with / (root) file system:
# mount /dev/dsk/c0t50000394980066DEd0s3 /mnt |
6) Mount the new system disk. When there is /etc in the other partition, also mount the file system under the mount point of the root file system as /etc.
# mount -F ufs /dev/dsk/c0t50000394980066DEd0s0 /mnt |
7) Cancel the system disk mirroring.
Update /mnt/etc/system, /mnt/etc/vfstab, /mnt/kernel/drv/sfdsk.conf, and /mnt/etc/opt/FJSVsdx/dev/devname.db.
7-1) Check if MPxIO is enabled for the system disk and record the result.
Execute the current subcommand of the format(1M) command. MPxIO is enabled, if the /scsi_vhci directory is in the logical path of the displayed boot disk.
Example: If the system disk is c0t50000394882261A0d0
# format c0t50000394882261A0d0 format> current Current Disk = c0t50000394882261A0d0: 00000000 <TOSHIBA-MBF2600RC-3706 cyl 64986 alt 2 hd 27 sec 668> /scsi_vhci/disk@g50000394882261a0 /dev/chassis/SYS/HDD0/disk |
7-2) Edit /mnt/etc/system.
An example is shown below.
# vi /mnt/etc/system |
7-3) Edit the mount setting lines of the root class partitions in /mnt/etc/vfstab according to the partitions of the new system disk.
An example is shown below.
Examples before edit
/dev/sfdsk/RootClass/dsk/swapVolume - - swap - no - /dev/sfdsk/RootClass/dsk/rootVolume /dev/sfdsk/RootClass/rdsk/rootVolume / ufs 1 no - |
Examples after edit
/dev/dsk/c0t50000394980066DEd0s1 - - swap - no - /dev/dsk/c0t50000394980066DEd0s0 /dev/rdsk/c0t50000394980066DEd0s0 / ufs 1 no - |
Note
The current physical disk name (cXtXdX) is not the same as the one when starting from the system disk as in the step 7-1) of [Restoring] if MPxIO is enabled in the step 2) of [Before the operation] and MPxIO is disabled in the step 7-1) of [Restoring]. In this case, describe the current physical disk name in /mnt/etc/vfstab.
7-4) Delete all the lines where sysvol-db and SDX_DEV_TREE is stated from /mnt/kernel/drv/sfdsk.conf.
An example is shown as below.
# vi /mnt/kernel/drv/sfdsk.conf |
7-5) Delete the lines which include physical disk names of root class that was checked in step 5) of [Before the operation] from /mnt/etc/opt/FJSVsdx/dev/devname.db.
An example is show as below:
# vi /mnt/etc/opt/FJSVsdx/dev/devname.db |
8) Edit /mnt/kernel/drv/mpt_sas.conf so that MPxIO is not enabled.
This step is necessary only if MPxIO is enabled in the step 2) of [Before the operation] and MPxIO is disabled in the step 7-1) of [Restoring].
Examples before edit
mpxio-disable="no"; |
Examples after edit
mpxio-disable="yes"; |
9) Apply the device name of the disk.
This operation is necessary to the disk of Expansion File Unit connected to a 6Gbps SAS card, or the internal disks of SPARC M12/M10 and SPARC T4-4/T4-2/T4-1/T3-4/T3-2/T3-1 servers.
# cd /devices ; tar cf - . | ( cd /mnt/devices ; tar xf - ) |
10) Update the boot-archive file.
# cd / # /sbin/bootadm update-archive -R /mnt |
11) Unmount the file system.
# umount /mnt/etc |
12) Reboot the system from the new boot disk by the following procedure.
The procedure to reboot from a boot disk is different for the disk of Expansion File Unit connected to a 6Gbps SAS card or the internal disks of SPARC M12/M10 and SPARC T4-4/T4-2/T4-1/T3-4/T3-2/T3-1 servers, and the other cases.
For the procedure for SAN Boot environment, see SAN Boot manual.
12a-1) Check the root partition's slice names of the new boot disk and make a note of them.
# ls -l /dev/dsk |grep c0t50000394980066DEd0s0 |
12a-2) Check and record the boot disk information and make a note of them.
How to check information depends on whether or not MPxIO is enabled in step 7-1).
When MPxIO is disabled in step 7-1).
Check and record the obp-path parameter of the Hardware properties.
# prtconf -v /dev/rdsk/c0t50000394980066DEd0s2 |
When MPxIO is enabled in step 7-1).
Check and record the obp-path parameter of the Paths from multipath bus adapters.
# prtconf -v /dev/rdsk/c0t50000394980066DEd0s2 |
12a-3) Enter the OpenBoot environment.
# halt |
Note
If the system is started from an emergency startup disk for restoration, enter the OpenBoot environment using the shutdown(1M) command.
12a-4) Set the boot-device property in OpenBoot.
Combine the obp-path parameter output in step 12a-2) and the slice name (:a in this example) output in step 12a-1), and then set the boot-device property.
ok setenv boot-device /pci@8000/pci@4/pci@0/pci@0/scsi@0/disk@w50000394980066de,0:a |
12a-5) Set the boot-device-index property of OpenBoot to 0.
ok setenv boot-device-index 0 |
12a-6) Boot and re-configure the system.
ok boot -r |
Note
The following warning messages may be output during the booting process, but it does not affect the system. The current swap area is set as a dump device. If you want to change the setting of dump device, use dumpadm(1M).
savecore: open("/dev/sfdsk/System/dsk/swap"): No such file or directory
dumpadm: cannot use /dev/sfdsk/System/dsk/swap as dump device:
No such file or directory |
12b-1) Check the root partition's device path of swapped boot disks and make a note of them.
# ls -l /dev/dsk | grep c0t0d0s0 |
12b-2) Enter the OpenBoot environment.
# halt |
Note
If the system is booted from an emergency startup disk for restoration, enter the OpenBoot environment using the shutdown(1M) command.
12b-3) Execute the show-devs command in the OpenBoot environment and display the device name.
ok show-devs /pci@1f,4000/scsi@3 |
12b-4) Set the boot-device property in OpenBoot.
When the device name "disk" in the last element path is output in step 12b-3):
The path that ../../devices was excluded from the one output in step 12b-1) and the part before @ of the last element path (sd in this example) is replaced to disk, is set to the boot-device property.
ok setenv boot-device /pci@1f,4000/scsi@3/disk@0,0:a |
When the device name "disk" in the last element path is not output in step 12b-3):
The path that ../../devices was excluded from the one output in step 12b-1) is set to the boot-device property.
ok setenv boot-device /pci@1f,4000/scsi@3/sd@0,0:a |
12b-5) Check if there is the boot-device-index property in OpenBoot. If there is, set the boot-device-index property to 0.
ok printenv boot-device-index |
If there is boot-device-index property, "boot-device-index = number"is displayed. In this case, execute the following command.
If "boot-device-index = number"is not displayed, do not execute the following command.
ok setenv boot-device-index 0 |
12b-6) Boot the system.
ok boot |
Note
The following warning messages may be output during the booting process, but it does not affect the system. The current swap area is set as a dumpdevice. If you want to change the setting of dump device, use dumpadm(1M).
savecore: open("/dev/sfdsk/System/dsk/swap"): No such file or directory
dumpadm: cannot use /dev/sfdsk/System/dsk/swap as dump device:
No such file or directory |
13) Reset the content of /kernel/drv/mpt_sas.conf file that was edited in step 8) to its original content.
This step is necessary only after you do the step 8).
Example before edit
mpxio-disable="yes"; |
Example after edit
mpxio-disable="no"; |
14) Execute the following command to enable MPxIO and reboot system.
This step is necessary only after you do the step 13).
# stmsboot -u |
15) If it is the cluster system, delete the cluster resource of the root class.
The target is the only root class. Do not delete the cluster resource of local classes and shared classes.
The following examples show the case where the following classes exist:
Root Class: RootClass
Local Class: class0001
Shared Class: class0002
Check whether the class resource of the root class exists.
# /etc/opt/FJSVsdx/bin/sdxdcrsc -I |
If the class resource of the root class is not displayed, go to the next step 16).
If the class resource of the root class is displayed, delete the class resource of the root class.
# /etc/opt/FJSVsdx/bin/sdxdcrsc -R -c RootClass |
Confirm that the class resource of the root class has been deleted.
Execute the following command and check that the target class name is not displayed.
# /etc/opt/FJSVsdx/bin/sdxdcrsc -I |
16) Re-configure mirroring of the system disk.
Note
When this procedure is performed by assuming that a system disk is broken, that is a disk was not swapped in step 1), the following messages may be output at the time of re-setting the mirroring.
# sdxdisk -M -c System -a type=root -d c0t0d0=Root1:keep,c1t0d0=Root2:undef |
When the above errors are output, re-boot the system and set the mirroring again.
Information
If there are missing parts and errors during this procedure, the mirroring status may be unstable and the subsequent re-settings may be failed.
You can check the unstable parts by using the sdxrootck command after performing this procedure or failing the re-setting.
See
For details on the sdxrootck command, see "F.3.2 sdxrootck - Confirmation of System Disk Mirroring Cancellation."
(5) System cannot be booted (Boot disk data damage).
Explanation
For some reason, the system disk data is invalid. You must restore data from the backup data created in advance.
If restoring data from the mirror disk was unsuccessful, follow the procedures shown below.
Resolution
For the ZFS boot environment, see "6.2.2 Restoring." For the UFS boot environment, see "6.3.3 Restoring (When the System Cannot Be Booted)."
(6) System cannot be booted (Incorrect boot disk device name).
Explanation
By mirroring system disks, the system's boot disk device name (boot-device parameter) is automatically set and updated. However, in some systems GDS cannot recognize correct device names and thus sets incorrect device names.
In that event, the following console message is output and booting the system fails.
ok boot
Rebooting with command: boot
Boot device: /ssm@0,0/pci@18,700000/pci@1/scsi@2/sd@0,0:a File and args:
Evaluating: boot
Can't open boot device |
In this example the boot disk device name is "/ssm@0,0/pci@18,700000/pci@1/scsi@2/sd@0,0:a", but the correct device name is "/ssm@0,0/pci@18,700000/pci@1/scsi@2/disk@0,0:a" and thus booting the system fails.
Resolution
Create the bootdevtab file (boot disk device name file) for proper device name configuration.
1) Specify the correct boot device name and boot the system.
ok boot /ssm@0,0/pci@18,700000/pci@1/scsi@2/disk@0,0:a |
For the correct boot disk device name, perform the following.
1-1) View device names of disks connected to the system.
For the following procedures, make a copy of the displayed results.
ok show-disks |
1-2) Based on a) in the displayed results, "/ssm@0,0/pci@18,700000/pci@1/scsi@2/sd@0,0:a" should be replaced with "/ssm@0,0/pci@18,700000/pci@1/scsi@2/disk@0,0:a" as that is the correct boot disk device name.
2) Create the bootdevtab file.
2-1) Create a template.
# sdxinfo -x BootDev > /var/tmp/bootdevtab |
2-2) Edit the template.
# vi /var/tmp/bootdevtab |
Specify the physical disk name in the first field.
Specify the corresponding boot disk device name in the second field. Neither field requires partition assignment, such as s0 and :a.
To the template a boot disk device name assumed by GDS will be output. If the output boot disk name is incorrect, replace it with the correct device name. The correct device name can be checked as described in 1-1) through 1-2). Note that two boot disk device names may be output. In that event, delete one of the names and replace the remaining one with the correct device name.
Note
Separate one field from another in space or tab delimited format. Lines starting with the number sign (#) are assumed as comments.
All disks registered with a root class are output to the template.
Specify disks that may be used as the boot disk in the bootdevtab file. The relevant disks are:
Original boot disk
Mirror disk
Spare disk connected to a boot disk's group
Boot and spare disks in an alternative boot environment
(When using the system volume snapshot function of GDS Snapshot)
Do not specify the following disks in the bootdevtab file. Delete them from the bootdevtab file or comment out the relevant lines. Leaving them unedited will not affect the operation.
Disks not used as the boot disk
Disks for which GDS correctly assumed the boot disk device name and no assignment in the bootdevtab file is required
When FC-AL internal disks are swapped, the boot disk device name changes.
Declare as follows so that recreation of the bootdevtab file is not required every time disks are swapped. Put the asterisk (*) after the last at mark (@). GDS will then complement the asterisk portion automatically.
Before edit : c0t0d0 /ssm@0,0/pci@1d,700000/SUNW,qlc@1/fp@0,0/ssd@w50020f230000798c,0 After edit : c0t0d0 /ssm@0,0/pci@1d,700000/SUNW,qlc@1/fp@0,0/disk@* |
For MPLB disks, specify the boot disk unit names of their paths.
Before edit : # mplb0 c12t16d0 /pci@1f,4000/fibre-channel@4/disk@10,0 c13t16d0 /pci@1f,4000/fibre-channel@5/disk@10,0 # mplb1c12t17d0 /pci@1f,4000/fibre-channel@4/disk@11,0 c13t17d0 /pci@1f,4000/fibre-channel@5/dsik@11,0 After edit :
# mplb0c12t16d0 /pci@1f,4000/fibre-channel@4/sd@10,0 c13t16d0 /pci@1f,4000/fibre-channel@5/sd@10,0 # mplb1c12t17d0 /pci@1f,4000/fibre-channel@4/sd@11,0 c13t17d0 /pci@1f,4000/fibre-channel@5/sd@11,0 |
2-3) Place the template in the /etc/opt/FJSVsdx/lib directory.
# cd /etc/opt/FJSVsdx/lib |
2-4) Confirm that the bootdevtab file was created normally.
# sdxinfo -x BootDev |
Check whether the boot disk device name is output normally.
3) Check whether the system's boot disk device name (boot-device parameter) is set normally.
3-1) Stop the system and specify the correct boot disk device name.
# init 0 |
3-2) Check the boot-device parameter values.
# eeprom boot-device |
In the boot-device parameter, the device name of the boot device where synchronization copying is complete is set. If synchronization copying is incomplete, perform the check after it is complete.
To add a disk to the root class and specify the disk in the bootdevtab file, before connecting to the disk to a group, re-create the bootdevtab file according to the procedures described in 2).
When the disk is connected to a group and synchronization copying is complete, the boot disk device name is added to the boot-device parameter.
When the bootdevtab file is not re-created before synchronization copying is complete, a boot disk device name assumed by GDS is added to the boot-device parameter when the synchronization copying ends. In that event, re-create the bootdevtab file and set the correct boot disk device name as follows. Alternatively, reboot the system for proper boot disk device name configuration.
# eeprom boot-device # eeprom boot-device="... /ssm@0,0/pci@18,700000/pci@1/scsi@2/disk@1,0:a" Correct boot disk device name |
(7) System cannot be booted in a ZFS Boot Environment (Abnormality of root class private slice).
Explanation
Data of the private slice in a root class may become incorrect and the system cannot be booted by some trouble.
Resolution
See
For details of Solaris commands, see the Solaris manual.
1) Boot the system from a CD-ROM device (or the network).
For the procedure, see the Solaris manual.
Note
To restore Solaris 10, to which the kernel patch 137111-08 or earlier has been applied, the version of an OS booted from a CD-ROM device (also network or emergency startup disk for restoration) must correspond to the version of the OS, to which the kernel patch 137111-08 or earlier has been applied, you are restoring. To boot from a CD-ROM device, use the medium of Solaris 10 5/08 or earlier.
To restore Solaris 10, to which the kernel patch 137137-09 or later has been applied, the version of an OS booted from a CD-ROM device (also network or emergency startup disk for restoration) must correspond to the version of the OS, to which the kernel patch 137137-09 or later has been applied, you are restoring. To boot from a CD-ROM device, use the medium of Solaris 10 10/08 or later.
To restore Solaris 11, the version of an OS booted from a CD-ROM device (also network or emergency startup disk for restoration) must correspond to the version of the OS you are restoring.
In a system where mirroring is conducted by GDS, booting from a CD-ROM device (also network) and performing an illegal operation may disrupt the mirroring environment. Be sure to follow the procedure described here and to avoid operations not provided in this procedure.
When mounting the file system on a physical slice at the time of recovery, mounting as read-only is recommended to write-lock.
2) Check the disk to configure the ZFS root pool that can be imported.
# zpool import |
3) Create a symbolic link of the s0 slice for the disk where the ZFS root pool for import exists.
For the disk that there is the ZFS root pool checked in step 2), create the symbolic link of the device special file (/dev/dsk/c0t0d0s0) of s0 slice.
In the following example, use /tmp/work as a work directory.
# mkdir /tmp/work |
4) Import the ZFS root pool.
# zpool import -d /tmp/work rpool |
Note
The messages "cannot mount" and "failed to create mountpoint" may be output. If you can confirm that it has been imported normally in the next step 5), there is no problem.
5) Check the status of imported ZFS root pool.
Check the symbolic link path created in step 3) is displayed, ONLINE is displayed in the state field and no error messages are displayed.
# zpool status rpool |
Note
If ZFS mirror is configured, the following may be displayed.
# zpool status rpool |
In this case, after detaching devices other than devices selected in step 3), check the symbolic link path created in step 3) is displayed in the config field, ONLINE is displayed in the state field and, no error messages are displayed.
# zpool detach rpool 980382089098729109 |
6) Mount the root file system to restore to /mnt.
6-1) Change the mountpoint property of the root file system to restore to /mnt.
# zfs list |
6-2) Mount the root file system to restore.
# zfs mount rpool/ROOT/s10s_u6was_07b |
7) Delete all the lines where sysvo1-db is stated from /mnt/kernel/drv/sfdsk.conf.
An example is shown below.
# vi /mnt/kernel/drv/sfdsk.conf |
8) Check if there is the boot-archive file under /mnt/platform. If there is, execute the bootadm(1M) command and update the boot-archive file.
# ls /mnt/platform/`uname -m`/boot_archive |
If there is the boot-archive file, the file name is displayed as shown above.
In this case, execute the following command. In other cases, do not execute the following command.
# /sbin/bootadm update-archive -R /mnt |
9) Unmount the root file system mounted in step 6) and restore the mountpoint property to /, then export the ZFS root pool.
# zfs unmount /mnt |
10) Start the system from the disk, confirmed in step 2, where the imported ZFS root pool exists.
The procedure to reboot from a boot disk is different for the following three cases:
For the procedure for SAN Boot environment, see SAN Boot manual.
10a-1) Check the device path of a slice of the restore destination.
Execute the following command to check the device path of a slice. Note and save this path for later reference.
# ls -l /dev/dsk | grep c0t600000E00D0000000001060300040000d0s0 |
10a-2) Check the boot disk information.
Execute the following command to check the device path of a slice. Note and save this path for later reference.
# prtconf -v /dev/rdsk/c0t600000E00D0000000001060300040000d0s0 |
(*1) Disk node name (disk@w500000e0d0460306,0 in this example)
(*2) SAS Address (500000e0d0460306 in this example)
Note
If MPxIO is disabled, the line "Paths from multipath bus adapters:" is not displayed.
10a-3) Enter the OpenBoot environment.
# shutdown -y -g0 -i0 |
10a-4) Check the device path of SAS HBA that is connected to the disk of the restore destination.
Execute the following command to check the device path of SAS HBA of the restore destination.
ok probe-scsi-all |
(*1) Device path
(*2) SAS Address (SAS Address checked in *2 in step 10a-2)
10a-5) Set boot-device property.
Concatenate the device path checked in step 10a-4), the disk node name checked in step 10a-2), and the slice name checked in step 10a-1). Set this concatenated value to the boot-device property of OBP.
ok setenv boot-device /pci@8100/pci@4/pci@0/pci@0/LSI,sas@0/disk@w500000e0d0460306,0:a |
10a-6) Check if there is boot-device-index property in OpenBoot. If there is, set boot-device-index property to "0."
ok printenv boot-device-index |
If there is boot-device-index property, "boot-device-index = number" is displayed. In this case, execute the following command.
If "boot-device-index = number" is not displayed, do not execute the following command.
ok setenv boot-device-index 0 |
10a-7) Boot the system.
ok boot |
10b-1) Check the slice name of the root partition of the new boot disk. Note and save this slice name for later reference.
[Example for Solaris 11.3 or earlier]
# ls -l /dev/dsk | grep c0t5000CCA00AC113CCd0s0 |
[Example for Solaris 11.4 or later]
# ls -l /dev/dsk | grep c0t50000397683346A9d0s0 |
10b-2) Check the boot disk information. Note and save it for later reference.
How you check the information depends on the OS version and whether or not MPxIO is currently enabled.
You can check whether or not MPxIO is enabled by using the current subcommand of the format(1M) command. MPxIO is enabled if the logical path of the boot disk has the /scsi_vhci directory inside.
Example: when the boot disk is c0t50000394882261A0d0
# format c0t50000394882261A0d0 |
[Solaris 11.3 or earlier]
When MPxIO is currently disabled
Check and record the obp-path parameter of the Hardware properties.
# prtconf -v /dev/rdsk/c0t5000CCA0150FEA10d0s0 disk, instance #0 Driver properties: ... Hardware properties: ... name='obp-path' type=string items=1 value='/pci@400/pci@2/pci@0/pci@e/scsi@0/disk@w5000cca0150fea11,0' ... Device Minor Nodes: ... |
When MPxIO is currently enabled
Check and record the obp-path parameter of the Paths from multipath bus adapters.
# prtconf -v /dev/rdsk/c0t5000CCA0150FEA10d0s0 |
[Solaris 11.4 or later]
When MPxIO is currently disabled
Check and record the ddi-boot-path parameter of the Hardware properties.
# prtconf -v /dev/rdsk/c0t50000397683346A9d0s2 disk, instance ... Driver properties: ... Hardware properties: ... name='ddi-boot-path' type=string items=1 value='/pci@8000/pci@4/pci@0/pci@0/scsi@0/disk@w50000397683346aa,0' ... Device Minor Nodes: ... |
When MPxIO is currently enabled
Check and record the ddi-boot-path parameter of the Paths from multipath bus adapters.
# prtconf -v /dev/rdsk/c0t50000397683346A9d0s2 disk, instance ... Device Hold: ... Driver properties: ... Hardware properties: ... Paths from multipath bus adapters: ... name='ddi-boot-path' type=string items=1 value='/pci@8000/pci@4/pci@0/pci@0/scsi@0/disk@w50000397683346aa,0' ... Device Minor Nodes: ... |
10b-3) Enter the OpenBoot environment.
# halt |
Note
If the system is started from an emergency startup disk for restoration, enter the OpenBoot environment using the shutdown(1M) command.
10b-4) Set the boot-device property in OpenBoot.
[Solaris 11.3 or earlier]
Concatenate the obp-path parameter output in step 10b-2) and the slice name (:a in this example) output in step 10b-1). Set this concatenated value to the boot-device property.
ok setenv boot-device /pci@400/pci@2/pci@0/pci@e/scsi@0/disk@w5000cca0150fea11,0:a |
[Solaris 11.4 or later]
Concatenate the ddi-boot-path parameter output in step 10b-2) and the slice name (:a in this example) output in step 10b-1). Set this concatenated value to the boot-device property.
ok setenv boot-device /pci@8000/pci@4/pci@0/pci@0/scsi@0/disk@w50000397683346aa,0:a |
10b-5) Check if there is boot-device-index property in OpenBoot. If there is, set the boot-device-index property to 0.
ok printenv boot-device-index |
If there is the boot-device-index property, "boot-device-index = number" is displayed. In this case, execute the following command.
If "boot-device-index = number" is not displayed, do not execute the following command.
ok setenv boot-device-index 0 |
10b-6) Boot the system.
ok boot |
Note
The following warning messages may be output during the booting process, but there is no problem. The current swap area is set as a dump device. If you want to change the setting of a dump device, use dumpadm(1M).
savecore: open("/dev/sfdsk/System/dsk/swap"): No such file or directory
dumpadm: cannot use /dev/sfdsk/System/dsk/swap as dump device:
No such file or directory. |
10c-1) Check the device path of the root partition of the new boot disk. Note and save this device path for later reference.
# ls -l /dev/dsk | grep c0t0d0s0 |
10c-2) Enter the OpenBoot environment.
# halt |
Note
If the system is started from an emergency startup disk for restoration, enter the OpenBoot environment using the shutdown(1M) command.
10c-3) Execute the show-devs command in the OpenBoot environment and display the device name.
ok show-devs /pci@1f,4000/scsi@3 |
10c-4) Set the boot-device property in OpenBoot.
When the device name "disk" in the last element path is output in step 10c-3):
The path that ../../devices was excluded from the one output in step 10c-1) and the part before @ of the last element path (sd in this example) is replaced to disk, is set to the boot-device property.
ok setenv boot-device /pci@1f,4000/scsi@3/disk@0,0:a |
When the device name "disk" of the last element path is not output in step 10c-3):
Set the path that ../../devices was excluded from the path output in step 10c-1) to the boot-device property.
ok setenv boot-device /pci@1f,4000/scsi@3/sd@0,0:a |
10c-5) Check if there is the boot-device-index property in OpenBoot. If there is, set the boot-device-index property to 0.
ok printenv boot-device-index |
If there is boot-device-index property, "boot-device-index = number" is displayed. In this case, execute the following command.
If "boot-device-index = number" is not displayed, do not execute the following command.
ok setenv boot-device-index 0 |
10c-6) Boot the system.
ok boot |
Note
The following warning messages may be output during the booting process, but it does not affect the system. The current swap area is set as a dump device. If you want to change the setting of a dump device, use dumpadm(1M).
savecore: open("/dev/sfdsk/System/dsk/swap"): No such file or directory
dumpadm: cannot use /dev/sfdsk/System/dsk/swap as dump device:
No such file or directory. |
11) Delete the private slice of the system disk.
Check the partition configuration of the system disk using the prtvtoc(1M) command and if the private slice exists (the slice which the tag is 33), delete it using the sdxsystem Format command.
# prtvtoc /dev/rdsk/c0t0d0s2 # /etc/opt/FJSVsdx/bin/sdxsystem Format -d c0t0d0 |
12) Initialize the disk which was the mirror destination of the system disk.
# dd if=/dev/zero of=/dev/rdsk/c1t0d0s0 bs=1024k |
For the format -e command, select the disk (c1t0d0) which was the mirror destination, and input "y" to the inquiry "Disk not labeled. Label it now?"
13) If it is the cluster system, delete the cluster resource of the root class.
The target is the only root class. Do not delete the cluster resource of local classes and shared classes.
The following examples show the case where the following classes exist:
Root class: System
Local class: class0001
Shared class: class0002
Check whether the class resource of the root class exists.
# /etc/opt/FJSVsdx/bin/sdxdcrsc -I |
If the class resource of the root class is not displayed, go to the next step 14).
If it is displayed, delete the class resource of the root class.
# /etc/opt/FJSVsdx/bin/sdxdcrsc -R -c System |
Confirm that the class resource of the root class has been deleted.
Execute the following command and check that the target class name is not displayed.
# /etc/opt/FJSVsdx/bin/sdxdcrsc -I |
14) Re-configure mirroring of the system disk.
For details, see "7.1.1 System Disk Mirroring Settings in a ZFS Boot Environment (GDS Mirroring)" or "7.2.1 System Disk Mirroring Settings in a ZFS Boot Environment (ZFS Mirroring)."
Note
The following may be output at the time of re-setting the mirroring.
# sdxdisk -M -c System -a type=root -d c0t0d0=Root1:keep |
When the above errors are output, re-boot the system and set the mirroring again.
(8) The system cannot be booted (ufs logging errors).
Explanation
When backing up a ufs file system on a system disk using the procedure described in "a) When booting the system from a CD-ROM device (or the network) and creating backups" in "6.3.1 Backing Up," if step a1) is not performed for disk disconnection, the following errors may be output on OS startup and starting the OS may fail.
WARNING: Error processing ufs log data during scan WARNING: ufs log for /var changed state to Error WARNING: Please umount(1M) /var and run fsck(1M) WARNING: Error accessing ufs log for /var; Please run fsck(1M) mount: I/O error mount: /dev/sfdsk/RootClass/dsk/varVolume cannot be mounted. mount: mount-point /var/run does not exist. |
Resolution
Restore the backed up file system.
For the restore methods, see "6.3.2 Restoring (When the System Can Be Booted)" or "6.3.3 Restoring (When the System Cannot Be Booted)."
Even if operation is performed as described in [Explanation], an error may not occur immediately.
Additionally, if the fsck(1M) command is used according to the messages shown in [Explanation], it may be possible to start the OS and it appears that the file system was restored. However, in these situations, file system data and the synchronized status of mirrors are not ensured.
When performing operation as described in [Explanation], be sure to restore the file system.
(9) The system cannot be booted (Alias name error of the boot device).
Explanation
After mirroring the system disk, aliases of the boot devices are registered with the OpenBoot nvramrc variable. The alias names may not be correctly registered because the number of bytes in nvramrc is restricted. In this case, if you specify the alias name to boot the system, the following message is output and the system cannot be booted.
{0} ok boot sdx-boot-alt-device-32-2b
Boot device: /pci@8600/pci@4/pci@0/pci@9/LSI,sas@ File and args:
Boot load failed
ERROR: boot-read fail |
Resolution
Specify either another boot device or a path of the boot device to boot the system.
Example 1: To boot the system by specifying another boot device
{0} ok printenv nvramrc
nvramrc = devalias net /pci@8600/pci@4/pci@0/pci@11/network@0
devalias disk /pci@8600/pci@4/pci@0/pci@9/LSI,sas@0/disk
devalias sdx-root-slice-32-1
/devices/scsi_vhci/disk@g600000e00d1100000010003101800000:a,raw
devalias sdx-boot-device-32-1
/pci@8600/pci@4/pci@0/pci@9/LSI,sas@0/disk@w500000e0d4003185,1:a
...
{0} ok boot sdx-boot-device-32-1 |
Example 2: To boot the system by specifying a path of the boot device
{0} ok boot /pci@8600/pci@4/pci@0/pci@9/LSI,sas@0/disk@w500000e0d4003195,0:a |
See
For how to check a path of the boot device, see the following procedures:
SAN Boot environment
The other cases
