Chapter 9 Backing Up and Restoring a Cluster
This is a list of step-by-step instructions in this chapter.
Backing Up a Cluster
Table 9–1 Task Map: Backing Up Cluster Files
How to Find File System Names to Back Up
Use this procedure to determine the names of the file
systems you want to back up.
-
Display the contents of the /etc/vfstab file.
You do not need to be superuser to run this command.
-
Look in the mount point column for the name of the file system you want
to back up.
Use this name when you back up the file system.
Example—Finding File System Names to Back Up
In the following example, the names of available file systems listed
in the /etc/vfstab file are displayed.
% more /etc/vfstab
#device device mount FS fsck mount mount
#to mount to fsck point type pass at boot options
#
#/dev/dsk/c1d0s2 /dev/rdsk/c1d0s2 /usr ufs 1 yes -
f - /dev/fd fd - no -
/proc - /proc proc - no -
/dev/dsk/c1t6d0s1 - - swap - no -
/dev/dsk/c1t6d0s0 /dev/rdsk/c1t6d0s0 / ufs 1 no -
/dev/dsk/c1t6d0s3 /dev/rdsk/c1t6d0s3 /cache ufs 2 yes -
swap - /tmp tmpfs - yes -
|
How to Determine the Number of Tapes Needed for a Full Backup
Use this procedure to calculate the number of tapes
you will need to back up a file system.
-
Become superuser on the cluster node you want to back up.
-
Estimate the size of the backup in bytes.
-
S
-
Displays the
estimated number of bytes needed to perform the backup.
-
filesystem
-
Specifies the name of the file system you want to back up.
-
Divide the estimated size by the capacity of the tape to see how many
tapes you need.
Example—Determining the Number of Tapes Needed
In the following example, the file system size of 905,881,620 bytes
will easily fit on a 4 GB tape (905,881,620 ÷ 4,000,000,000).
# ufsdump S /global/phys-schost-1
905881620
|
How to Back Up the root (/) File System
Use this procedure to back up the root (/) file system of a cluster node. Be sure the cluster is running
problem-free before performing the backup procedure.
-
Become superuser on the cluster node you want to back up.
-
Switch each running data service from the node to be backed up to another
node in the cluster.
# scswitch -z -D disk-device-group[,...] -h node[,...]
|
-
-z
-
Performs the
switch.
-
-D disk-device-group
-
Name of the disk device group to be switched.
-
-h node
-
Name of the cluster node to switch the disk device group to.
This node becomes the new primary.
-
Shut down the node.
-
Reboot the node in non-cluster mode.
-
SPARC:
-
x86:
<<< Current Boot Parameters >>>
Boot path: /pci@0,0/pci8086,2545@3/pci8086,1460@1d/pci8086,341a@7,1/
sd@0,0:a
Boot args:
Type b [file-name] [boot-flags] <ENTER> to boot with options
or i <ENTER> to enter boot interpreter
or <ENTER> to boot with defaults
<<< timeout in 5 seconds >>>
Select (b)oot or (i)nterpreter: b -x
|
-
Back up the root (/) file system.
-
If the root disk is not encapsulated, use the following command.
# ufsdump 0ucf dump-device /
|
-
If the root disk is encapsulated, use the following command.
# ufsdump 0ucf dump-device /dev/vx/rdsk/rootvol
|
Refer to the ufsdump(1M)
man page for more information.
-
Reboot the node in cluster mode.
Example—Backing Up the root (/) File System
In the following example, the root (/) file system
is backed up onto tape device /dev/rmt/0.
# ufsdump 0ucf /dev/rmt/0 /
DUMP: Writing 63 Kilobyte records
DUMP: Date of this level 0 dump: Tue Apr 18 18:06:15 2000
DUMP: Date of last level 0 dump: the epoch
DUMP: Dumping /dev/rdsk/c0t0d0s0 (phys-schost-1:/) to /dev/rmt/0
DUMP: Mapping (Pass I) [regular files]
DUMP: Mapping (Pass II) [directories]
DUMP: Estimated 859086 blocks (419.48MB).
DUMP: Dumping (Pass III) [directories]
DUMP: Dumping (Pass IV) [regular files]
DUMP: 859066 blocks (419.47MB) on 1 volume at 2495 KB/sec
DUMP: DUMP IS DONE
DUMP: Level 0 dump on Tue Apr 18 18:06:15 2000
|
How to Perform Online Backups for Mirrors (Solstice DiskSuite/Solaris Volume Manager)
A mirrored metadevice can be backed up without
unmounting it or taking the entire mirror offline. One of the submirrors must
be taken offline temporarily, thus losing mirroring, but it can be placed
online and resynced as soon as the backup is complete, without halting the
system or denying user access to the data. Using mirrors to perform online
backups creates a backup that is a “snapshot” of an active file
system.
A problem might occur if a program writes data onto the volume immediately
before the lockfs command is run. To prevent this problem,
temporarily stop all the services running on this node. Also, be sure the
cluster is running problem-free before performing the backup procedure.
-
Become superuser on the cluster node you want to back up.
-
Use the metaset(1M)
command to determine which node has the ownership on the backed up volume.
-
-s setname
-
Specifies the diskset name.
-
Use the lockfs(1M)
command with the -w option to lock the file system from
writes.
Note –
You must lock the file system only if a UFS file system resides
on the mirror. For example, if the metadevice is set up as a raw device for
database management software or some other specific application, it would
not be necessary to use the lockfs command. You may, however,
want to run the appropriate vender-dependent utility to flush any buffers
and lock access.
-
Use the metastat(1M)
command to determine the names of the submirrors.
-
-p
-
Displays the
status in a format similar to the md.tab file.
-
Use the metadetach(1M)
command to take one submirror offline from the mirror.
# metadetach -s setname mirror submirror
|
Note –
Reads will continue to be made from the other submirrors. However,
the offline submirror will be out of sync as soon as the first write is made
to the mirror. This inconsistency is corrected when the offline submirror
is brought back online. You don't need to run fsck.
-
Unlock the file systems and allow writes to continue, using the lockfs command with the -u option.
-
Perform a file system check.
# fsck /dev/md/diskset/rdsk/submirror
|
-
Back up the offline submirror to tape or another medium.
Use the ufsdump(1M)
command or the backup utility that you usually use.
# ufsdump 0ucf dump-device submirror
|
Note –
Use the raw device (/rdsk) name for the submirror,
rather than the block device (/dsk) name.
-
Use the metattach(1M)
command to place the metadevice back online.
# metattach -s setname mirror submirror
|
When the metadevice is placed online, it is automatically resynced with
the mirror.
-
Use the metastat command to verify that the submirror
is resyncing.
# metastat -s setname mirror
|
Example—Performing Online Backups for Mirrors (Solstice DiskSuite/Solaris Volume Manager)
In the following example, the cluster node phys-schost-1
is the owner of the metaset schost-1, therefore the backup
procedure is performed from phys-schost-1. The mirror /dev/md/schost-1/dsk/d0 consists of the submirrors d10, d20, and d30.
[Determine the owner of the metaset:]
# metaset -s schost-1
Set name = schost-1, Set number = 1
Host Owner
phys-schost-1 Yes
...
[Lock the file system from writes:]
# lockfs -w /global/schost-1
[List the submirrors:]
# metastat -s schost-1 -p
schost-1/d0 -m schost-1/d10 schost-1/d20 schost-1/d30 1
schost-1/d10 1 1 d4s0
schost-1/d20 1 1 d6s0
schost-1/d30 1 1 d8s0
[Take a submirror offline:]
# metadetach -s schost-1 d0 d30
[Unlock the file system:]
# lockfs -u /
[Check the file system:]
# fsck /dev/md/schost-1/rdsk/d30
[Copy the submirror to the backup device:]
# ufsdump 0ucf /dev/rmt/0 /dev/md/schost-1/rdsk/d30
DUMP: Writing 63 Kilobyte records
DUMP: Date of this level 0 dump: Tue Apr 25 16:15:51 2000
DUMP: Date of last level 0 dump: the epoch
DUMP: Dumping /dev/md/schost-1/rdsk/d30 to /dev/rdsk/c1t9d0s0.
...
DUMP: DUMP IS DONE
[Bring the submirror back online:]
# metattach -s schost-1 d0 d30
schost-1/d0: submirror schost-1/d30 is attached
[Resync the submirror:]
# metastat -s schost-1 d0
schost-1/d0: Mirror
Submirror 0: schost-0/d10
State: Okay
Submirror 1: schost-0/d20
State: Okay
Submirror 2: schost-0/d30
State: Resyncing
Resync in progress: 42% done
Pass: 1
Read option: roundrobin (default)
...
|
SPARC: How to Perform Online Backups for Volumes (VERITAS Volume Manager)
VERITAS Volume Manager identifies a mirrored volume
as a plex. A plex can be backed up without unmounting it or taking the entire
volume offline. This is done by creating a snapshot copy of the volume and
backing up this temporary volume without halting the system or denying user
access to the data.
Be sure the cluster is running problem-free before performing the backup
procedure.
-
Log on to any node in the cluster, and become superuser on the current
primary node for the disk group on the cluster.
-
List the disk group information.
-
Run the scstat(1M)
command to see which node has the disk group currently imported, indicating
it is the primary node for the disk group.
-
-D
-
Shows the
status for all disk device groups.
-
Create a snapshot of the volume using the vxassist
command.
# vxassist -g diskgroup snapstart volume
|
Note –
Creating a snapshot can take a long time depending on the size
of your volume.
-
Verify the new volume was created.
When the snapshot is complete, a status of Snapdone
displays in the State field for the selected disk group.
-
Stop any data services that are accessing the file system.
# scswitch -z -g resource-group[,...] -h ““
|
Note –
Stop all data services to ensure that the data file system is
properly backed up. If no data services are running, you do not need to perform Step 6 and Step 8.
-
Create a backup volume named bkup-vol and attach
the snapshot volume to it using the vxassist command.
# vxassist -g diskgroup snapshot volume bkup-vol
|
-
Restart any data services that were stopped in Step 6,
using the scswitch(1M)
command.
# scswitch -z -g resource-group[,...] -h node[,...]
|
-
Verify the volume is now attached to the new volume bkup-vol using the vxprint command.
-
Register the disk group configuration change.
# scconf -c -D name=diskgroup,sync
|
-
Check the backup volume using the fsck command.
# fsck -y /dev/vx/rdsk/diskgroup/bkup-vol
|
-
Perform a backup to copy the volume bkup-vol to tape
or another medium.
Use the ufsdump(1M)
command or the backup utility you normally use.
# ufsdump 0ucf dump-device /dev/vx/dsk/diskgroup/bkup-vol
|
-
Remove the temporary volume using vxedit.
-
Register the disk group configuration changes using the scconf(1M) command.
# scconf -c -D name=diskgroup,sync
|
SPARC: Example—Performing Online Backups for Volumes (VERITAS Volume Manager)
In the following example, the cluster node phys-schost-2
is the primary owner of the metaset disk group schost-1,
therefore the backup procedure is performed from phys-schost-2.
The volume /vo101 is copied and then associated with
a new volume, bkup-vol.
[Become superuser on the primary node.]
[Identify the current primary node for the disk group:]
# scstat -D
-- Device Group Servers --
Device Group Primary Secondary
------------ ------- ---------
Device group servers: rmt/1 - -
Device group servers: schost-1 phys-schost-2 phys-schost-1
-- Device Group Status --
Device Group Status
------------ ------
Device group status: rmt/1 Offline
Device group status: schost-1 Online
[List the disk group information:]
# vxprint -g schost-1
TY NAME ASSOC KSTATE LENGTH PLOFFS STATE TUTIL0 PUTIL0
dg schost-1 schost-1 - - - - - -
dm schost-101 c1t1d0s2 - 17678493 - - - -
dm schost-102 c1t2d0s2 - 17678493 - - - -
dm schost-103 c2t1d0s2 - 8378640 - - - -
dm schost-104 c2t2d0s2 - 17678493 - - - -
dm schost-105 c1t3d0s2 - 17678493 - - - -
dm schost-106 c2t3d0s2 - 17678493 - - - -
v vol01 gen ENABLED 204800 - ACTIVE - -
pl vol01-01 vol01 ENABLED 208331 - ACTIVE - -
sd schost-101-01 vol01-01 ENABLED 104139 0 - - -
sd schost-102-01 vol01-01 ENABLED 104139 0 - - -
pl vol01-02 vol01 ENABLED 208331 - ACTIVE - -
sd schost-103-01 vol01-02 ENABLED 103680 0 - - -
sd schost-104-01 vol01-02 ENABLED 104139 0 - - -
pl vol01-03 vol01 ENABLED LOGONLY - ACTIVE - -
sd schost-103-02 vol01-03 ENABLED 5 LOG - - -
[Start the snapshot operation:]
# vxassist -g schost-1 snapstart vol01
[Verify the new volume was created:]
# vxprint -g schost-1
TY NAME ASSOC KSTATE LENGTH PLOFFS STATE TUTIL0 PUTIL0
dg schost-1 schost-1 - - - - - -
dm schost-101 c1t1d0s2 - 17678493 - - - -
dm schost-102 c1t2d0s2 - 17678493 - - - -
dm schost-103 c2t1d0s2 - 8378640 - - - -
dm schost-104 c2t2d0s2 - 17678493 - - - -
dm schost-105 c1t3d0s2 - 17678493 - - - -
dm schost-106 c2t3d0s2 - 17678493 - - - -
v vol01 gen ENABLED 204800 - ACTIVE - -
pl vol01-01 vol01 ENABLED 208331 - ACTIVE - -
sd schost-101-01 vol01-01 ENABLED 104139 0 - - -
sd schost-102-01 vol01-01 ENABLED 104139 0 - - -
pl vol01-02 vol01 ENABLED 208331 - ACTIVE - -
sd schost-103-01 vol01-02 ENABLED 103680 0 - - -
sd schost-104-01 vol01-02 ENABLED 104139 0 - - -
pl vol01-03 vol01 ENABLED LOGONLY - ACTIVE - -
sd schost-103-02 vol01-03 ENABLED 5 LOG - - -
pl vol01-04 vol01 ENABLED 208331 - SNAPDONE - -
sd schost-105-01 vol01-04 ENABLED 104139 0 - - -
sd schost-106-01 vol01-04 ENABLED 104139 0 - - -
[Stop data services, if necessary:]
# scswitch -z -g nfs-rg -h ““
[Create a copy of the volume:]
# vxassist -g schost-1 snapshot vol01 bkup-vol
[Restart data services, if necessary:]
# scswitch -z -g nfs-rg -h phys-schost-1
[Verify bkup-vol was created:]
# vxprint -g schost-1
TY NAME ASSOC KSTATE LENGTH PLOFFS STATE TUTIL0 PUTIL0
dg schost-1 schost-1 - - - - - -
dm schost-101 c1t1d0s2 - 17678493 - - - -
...
v bkup-vol gen ENABLED 204800 - ACTIVE - -
pl bkup-vol-01 bkup-vol ENABLED 208331 - ACTIVE - -
sd schost-105-01 bkup-vol-01 ENABLED 104139 0 - - -
sd schost-106-01 bkup-vol-01 ENABLED 104139 0 - - -
v vol01 gen ENABLED 204800 - ACTIVE - -
pl vol01-01 vol01 ENABLED 208331 - ACTIVE - -
sd schost-101-01 vol01-01 ENABLED 104139 0 - - -
sd schost-102-01 vol01-01 ENABLED 104139 0 - - -
pl vol01-02 vol01 ENABLED 208331 - ACTIVE - -
sd schost-103-01 vol01-02 ENABLED 103680 0 - - -
sd schost-104-01 vol01-02 ENABLED 104139 0 - - -
pl vol01-03 vol01 ENABLED LOGONLY - ACTIVE - -
sd schost-103-02 vol01-03 ENABLED 5 LOG - - -
[Synchronize the disk group with cluster framework:]
# scconf -c -D name=schost-1,sync
[Check the file systems:]
# fsck -y /dev/vx/rdsk/schost-1/bkup-vol
[Copy bkup-vol to the backup device:]
# ufsdump 0ucf /dev/rmt/0 /dev/vx/rdsk/schost-1/bkup-vol
DUMP: Writing 63 Kilobyte records
DUMP: Date of this level 0 dump: Tue Apr 25 16:15:51 2000
DUMP: Date of last level 0 dump: the epoch
DUMP: Dumping /dev/vx/dsk/schost-2/bkup-vol to /dev/rmt/0.
...
DUMP: DUMP IS DONE
[Remove the bkup-volume:]
# vxedit -rf rm bkup-vol
[Synchronize the disk group:]
# scconf -c -D name=schost-1,sync
|
Restoring Cluster Files Overview
The ufsrestore(1M) command copies files to disk, relative
to the current working directory, from backups created using the ufsdump(1M) command.
You can use ufsrestore to reload an entire file system
hierarchy from a level 0 dump and incremental dumps that follow it, or to
restore one or more single files from any dump tape. If ufsrestore is run as superuser, files are restored with their original owner,
last modification time, and mode (permissions).
Before you start to restore files or file systems, you need to know
the following information.
-
Which tapes you need
-
The raw device name on which you want to restore the file
system
-
The type of tape drive you will use
-
The device name (local or remote) for the tape drive
-
The partition scheme of any failed disk, because the partitions
and file systems must be exactly duplicated on the replacement disk
Restoring Cluster Files
Table 9–2 Task Map: Restoring Cluster Files
How to Restore Individual Files Interactively (Solstice DiskSuite/Solaris Volume Manager)
Use this procedure to restore one or
more individual files. Be sure the cluster is running problem-free before
performing the restore procedure.
-
Become superuser on the cluster node you want to restore.
-
Stop all the data services that are using the files to be restored.
# scswitch -z -g resource-group[,...] -h ““
|
-
Restore the files using the ufsrestore command.
How to Restore the root (/) File System (Solstice DiskSuite/Solaris Volume Manager)
Use this procedure to restore the root
(/) file systems to a new disk, such as after replacing
a bad root disk. The node being restored should not be booted. Be sure the
cluster is running problem-free before performing the restore procedure.
Note –
Since you must partition the new disk using the same format as
the failed disk, identify the partitioning scheme before you begin this procedure,
and recreate file systems as appropriate.
-
Become superuser on a cluster node with access to the metaset, other than the node you want to restore.
-
Remove from all metasets the hostname of the node being restored.
Run this command from a node in the metaset other than the node you
are removing.
# metaset -s setname -f -d -h nodelist
|
-
-s setname
-
Specifies
the diskset name.
-
-f
-
Force.
-
-d
-
Deletes from
the diskset.
-
-h nodelist
-
Specifies the name of the node to delete from the diskset.
-
Replace the failed disk on the node on which the root (/)
file system will be restored.
Refer to disk replacement procedures in the documentation that came
with your server.
-
Boot the node that you want to restore.
-
Create all the partitions and swap on the root disk using the format(1M)
command.
Recreate the original partitioning scheme that was on the failed disk.
-
Create the root (/) file system and other file
systems as appropriate, using the newfs(1M) command.
Recreate the original file systems that were on the failed disk.
Note –
Be sure to create the /global/.devices/node@nodeid
file system.
-
Mount the root (/) file system on a temporary mount
point.
# mount device temp-mountpoint
|
-
Use the following commands to restore the root (/)
file system.
# cd temp-mountpoint
# ufsrestore rvf dump-device
# rm restoresymtable
# cd /
# umount temp-mountpoint
# fsck raw-disk-device
|
The file system is now restored.
-
Install a new boot block on the new disk.
# /usr/sbin/installboot /usr/platform/`uname -i`/lib/fs/ufs/bootblk raw-disk-device
|
-
Reboot the node in single-user mode.
-
Replace the disk ID using the scdidadm(1M) command.
-
Use the metadb(1M)
command to recreate the state database replicas.
# metadb -c copies -af raw-disk-device
|
-
-c copies
-
Specifies the number of replicas
to create.
-
-f raw-disk-device
-
Raw disk device on which to create replicas.
- -a
-
Adds replicas.
-
Reboot the node in cluster mode.
-
Start the reboot.
During this boot you might see an error or warning message, ending with
the following instruction:
Type control-d to proceed with normal startup,
(or give root password for system maintenance):
|
-
Press Control-d to boot into multiuser mode.
-
From a cluster node other than the restored node, use the metaset command to add the restored node to all metasets.
phys-schost-2# metaset -s setname -a -h nodelist
|
-
-a
-
Creates and
adds the host to the diskset.
The node is rebooted into cluster mode. The cluster is
ready to use.
Example—Restoring the root (/) File System
(Solstice DiskSuite/Solaris Volume Manager)
The following example shows the root (/) file system
restored to the node phys-schost-1 from the tape device /dev/rmt/0. The metaset command is run from
another node in the cluster, phys-schost-2, to remove and
later add back node phys-schost-1 to the diskset schost-1. All other commands are run from phys-schost-1. A new boot block is created on /dev/rdsk/c0t0d0s0,
and three state database replicas are recreated on /dev/rdsk/c0t0d0s4.
[Become superuser on a cluster node other than the node to be restored.]
[Remove the node from the metaset:]
phys-schost-2# metaset -s schost-1 -f -d -h phys-schost-1
[Replace the failed disk and boot the node:]
|
Boot the node from the Solaris CD:
-
SPARC: At the OpenBoot PROM ok
prompt, type the following command:
-
x86: Insert the CD into the system's CD drive
and boot the system by shutting it down and then turning it off and on. On
the Current Boot Parameters screen, type the following command:
<<< Current Boot Parameters >>>
Boot path: /pci@0,0/pci8086,2545@3/pci8086,1460@1d/pci8086,341a@7,1/
sd@0,0:a
Boot args:
Type b [file-name] [boot-flags] <ENTER> to boot with options
or i <ENTER> to enter boot interpreter
or <ENTER> to boot with defaults
<<< timeout in 5 seconds >>>
Select (b)oot or (i)nterpreter: b -s
|
[Use format and newfs to recreate partitions and file systems.]
[Mount the root file system on a temporary mount point:]
# mount /dev/dsk/c0t0d0s0 /a
[Restore the root file system:]
# cd /a
# ufsrestore rvf /dev/rmt/0
# rm restoresymtable
# cd /
# umount /a
# fsck /dev/rdsk/c0t0d0s0
[Install a new boot block:]
# /usr/sbin/installboot /usr/platform/`uname \
-i`/lib/fs/ufs/bootblk /dev/rdsk/c0t0d0s0
[Reboot in single-user mode:]
# reboot -- "-s"
[Replace the disk ID:]
# scdidadm -R /dev/dsk/c0t0d0
[Recreate state database replicas:]
# metadb -c 3 -af /dev/rdsk/c0t0d0s4
# reboot
Press Control-d to boot into multiuser mode.
[Add the node back to the metaset:]
phys-schost-2# metaset -s schost-1 -a -h phys-schost-1
|
How to Restore a root (/) File System That Was
on a Metadevice (Solstice DiskSuite/Solaris Volume Manager)
Use this procedure to restore a root (/) file system that was on a metadevice when the backups were performed.
Perform this procedure under circumstances such as when a root disk is corrupted
and replaced with a new disk. The node being restored should not be booted.
Be sure the cluster is running problem-free before performing the restore
procedure.
Note –
Since you must partition the new disk using the same format as
the failed disk, identify the partitioning scheme before you begin this procedure,
and recreate file systems as appropriate.
-
Become superuser on a cluster node with access to the metaset, other than the node you want to restore.
-
Remove from all metasets the hostname of the node being restored.
# metaset -s setname -f -d -h nodelist
|
-
-s setname
-
Specifies the metaset name.
-
-f
-
Force.
-
-d
-
Deletes from
the metaset.
-
-h nodelist
-
Specifies the name of the node to delete from the metaset.
-
Replace the failed disk on the node on which the root (/)
file system will be restored.
Refer to disk replacement procedures in the documentation that came
with your server.
-
Boot the node that you want to restore.
-
Create all the partitions and swap on the root disk using the format command.
Recreate the original partitioning scheme that was on the failed disk.
-
Create the root (/) file system and other file
systems as appropriate, using the newfs command
Recreate the original file systems that were on the failed disk.
Note –
Be sure to create the /global/.devices/node@nodeid
file system.
-
Mount the root (/) file system on a temporary mount
point.
# mount device temp-mountpoint
|
-
Use the following commands to restore the root (/)
file system.
# cd temp-mountpoint
# ufsrestore rvf dump-device
# rm restoresymtable
|
-
Install a new boot block on the new disk.
# /usr/sbin/installboot /usr/platform/`uname -i`/lib/fs/ufs/bootblk raw-disk-device
|
-
Remove the lines in the /temp-mountpoint/etc/system file for MDD root information.
* Begin MDD root info (do not edit)
forceload: misc/md_trans
forceload: misc/md_raid
forceload: misc/md_mirror
forceload: misc/md_hotspares
forceload: misc/md_stripe
forceload: drv/pcipsy
forceload: drv/glm
forceload: drv/sd
rootdev:/pseudo/md@0:0,10,blk
* End MDD root info (do not edit)
|
-
Edit the /temp-mountpoint/etc/vfstab file to change
the root entry from a metadevice to a corresponding normal slice for each
file system on the root disk that is part of the metadevice.
Example:
Change from—
/dev/md/dsk/d10 /dev/md/rdsk/d10 / ufs 1 no -
Change to—
/dev/dsk/c0t0d0s0 /dev/rdsk/c0t0d0s0 / ufs 1 no -
|
-
Unmount the temporary file system, and check the raw disk device.
# cd /
# umount temp-mountpoint
# fsck raw-disk-device
|
-
Reboot the node in single-user mode.
-
Replace the disk ID using the scdidadm command.
-
Use the metadb command to recreate the state database
replicas.
# metadb -c copies -af raw-disk-device
|
-
-c copies
-
Specifies the number of replicas to create.
-
-af raw-disk-device
-
Creates initial state database replicas on the named raw disk
device.
-
Reboot the node in cluster mode.
-
Start the reboot.
During this boot you will see error or warning messages, ending with
the following instruction:
Type control-d to proceed with normal startup,
(or give root password for system maintenance):
|
-
Press Control-d to boot into multiuser mode.
-
From a cluster node other than the restored node, use the metaset command to add the restored node to all metasets.
phys-schost-2# metaset -s setname -a -h nodelist
|
-
-a
-
Adds (creates)
the metaset.
Set up the metadevice/mirror for root (/)
according to the Solstice DiskSuite documentation.
The node is rebooted into cluster mode. The cluster is ready to use.
Example—Restoring a root (/) File System
That Was on a Metadevice (Solstice DiskSuite/Solaris Volume Manager)
The following example shows the root (/) file system
restored to the node phys-schost-1 from the tape device /dev/rmt/0. The metaset command is run from
another node in the cluster, phys-schost-2, to remove and
later add back node phys-schost-1 to the metaset schost-1. All other commands are run from phys-schost-1. A new boot block is created on /dev/rdsk/c0t0d0s0,
and three state database replicas are recreated on /dev/rdsk/c0t0d0s4.
[Become superuser on a cluster node with access to the metaset,
other than the node to be restored.]
[Remove the node from the metaset:]
phys-schost-2# metaset -s schost-1 -f -d -h phys-schost-1
[Replace the failed disk and boot the node:]
|
Boot the node from the Solaris CD:
-
SPARC: At the OpenBoot PROM ok
prompt, type the following command:
-
x86: Insert the CD into the system's CD drive
and boot the system by shutting it down and then turning it off and on. On
the Current Boot Parameters screen, type the following command:
<<< Current Boot Parameters >>>
Boot path: /pci@0,0/pci8086,2545@3/pci8086,1460@1d/pci8086,341a@7,1/
sd@0,0:a
Boot args:
Type b [file-name] [boot-flags] <ENTER> to boot with options
or i <ENTER> to enter boot interpreter
or <ENTER> to boot with defaults
<<< timeout in 5 seconds >>>
Select (b)oot or (i)nterpreter: b -s
|
[Use format and newfs to recreate partitions and file systems.]
[Mount the root file system on a temporary mount point:]
# mount /dev/dsk/c0t0d0s0 /a
[Restore the root file system:]
# cd /a
# ufsrestore rvf /dev/rmt/0
# rm restoresymtable
[Install a new boot block:]
# /usr/sbin/installboot /usr/platform/`uname \
-i`/lib/fs/ufs/bootblk /dev/rdsk/c0t0d0s0
[Remove the lines in /temp-mountpoint/etc/system file for MDD root information:]
* Begin MDD root info (do not edit)
forceload: misc/md_trans
forceload: misc/md_raid
forceload: misc/md_mirror
forceload: misc/md_hotspares
forceload: misc/md_stripe
forceload: drv/pcipsy
forceload: drv/glm
forceload: drv/sd
rootdev:/pseudo/md@0:0,10,blk
* End MDD root info (do not edit)
[Edit the /temp-mountpoint/etc/vfstabfile]
Example:
Change from—
/dev/md/dsk/d10 /dev/md/rdsk/d10 / ufs 1 no -
Change to—
/dev/dsk/c0t0d0s0 /dev/rdsk/c0t0d0s0 /usr ufs 1 no -
[Unmount the temporary file system and check the raw disk device:]
# cd /
# umount /a
# fsck /dev/rdsk/c0t0d0s0
[Reboot in single-user mode:]
# reboot -- "-s"
[Replace the disk ID:]
# scdidadm -R /dev/dsk/c0t0d0
[Recreate state database replicas:]
# metadb -c 3 -af /dev/rdsk/c0t0d0s4
# reboot
Type Control-d to boot into multiuser mode.
[Add the node back to the metaset:]
phys-schost-2# metaset -s schost-1 -a -h phys-schost-1
|
SPARC: How to Restore a Non-Encapsulated root (/) File
System (VERITAS Volume Manager)
Use this
procedure to restore a non-encapsulated root (/) file
system to a node. The node being restored should not be booted. Be sure the
cluster is running problem-free before performing the restore procedure.
Note –
Since you must partition the new disk using the same format as
the failed disk, identify the partitioning scheme before you begin this procedure,
and recreate file systems as appropriate.
-
Replace the failed disk on the node where the root file system will
be restored.
Refer to disk replacement procedures in the documentation that came
with your server.
-
Boot the node that you want to restore.
-
If you are using the Solaris CD, at the OpenBoot PROM ok prompt, type the following command:
-
If you are using a Solaris JumpStartTM server, at the OpenBoot PROM ok prompt, type
the following command:
-
Create all the partitions and swap on the root disk using the format command.
Recreate the original partitioning scheme that was on the failed disk.
-
Create the root (/) file system and other file
systems as appropriate, using the newfs command.
Recreate the original file systems that were on the failed disk.
Note –
Be sure to create the /global/.devices/node@nodeid
file system.
-
Mount the root (/) file system on a temporary mount
point.
# mount device temp-mountpoint
|
-
Restore the root (/) file system from backup, and
unmount and check the file system.
# cd temp-mountpoint
# ufsrestore rvf dump-device
# rm restoresymtable
# cd /
# umount temp-mountpoint
# fsck raw-disk-device
|
The file system is now restored.
-
Install a new boot block on the new disk.
# /usr/sbin/installboot /usr/platform/`uname -i`/lib/fs/ufs/bootblk raw-disk-device
|
-
Reboot the node in single-user mode.
-
Start the reboot.
During this boot you will see error or warning messages, ending with
the following instruction:
Type control-d to proceed with normal startup,
(or give root password for system maintenance):
|
-
Type the root password.
-
Update the disk ID using the scdidadm command.
# scdidadm -R /dev/rdsk/disk-device
|
-
Press Control-d to resume in multiuser mode.
The node reboots into cluster mode. The cluster is ready to use.
SPARC: Example—Restoring a Non-Encapsulated root (/)
File System (VERITAS Volume Manager)
The following example shows a non-encapsulated root (/)
file system restored to the node phys-schost-1 from the
tape device /dev/rmt/0.
[Replace the failed disk and boot the node:]
|
Boot the node from the Solaris CD. At the OpenBoot PROM ok prompt, type the following command:
ok boot cdrom -s
...
[Use format and newfs to create partitions and file systems]
[Mount the root file system on a temporary mount point:]
# mount /dev/dsk/c0t0d0s0 /a
[Restore the root file system:]
# cd /a
# ufsrestore rvf /dev/rmt/0
# rm restoresymtable
# cd /
# umount /a
# fsck /dev/rdsk/c0t0d0s0
[Install a new boot block:]
# /usr/sbin/installboot /usr/platform/`uname \
-i`/lib/fs/ufs/bootblk /dev/rdsk/c0t0d0s0
[Reboot in single-user mode:]
# reboot -- "-s"
[Update the disk ID:]
# scdidadm -R /dev/rdsk/c0t0d0
[Press Control-d to resume in multiuser mode]
|
SPARC: How to Restore an Encapsulated root (/) File System
(VERITAS Volume Manager)
Use this procedure
to restore an encapsulated root (/) file system to a
node. The node being restored should not be booted. Be sure the cluster is
running problem-free before performing the restore procedure.
Note –
Since you must partition the new disk using the same format as
the failed disk, identify the partitioning scheme before you begin this procedure,
and recreate file systems as appropriate.
-
Replace the failed disk on the node where the root file system will
be restored.
Refer to disk replacement procedures in the documentation that came
with your server.
-
Boot the node that you want to restore.
-
If you are using the Solaris CD, at the OpenBoot PROM ok prompt, type the following command:
-
If you are using a Solaris JumpStartTM server, at the OpenBoot PROM ok prompt, type
the following command:
-
Create all the partitions and swap on the root disk using the format command.
Recreate the original partitioning scheme that was on the failed disk.
-
Create the root (/) file system and other file
systems as appropriate, using the newfs command.
Recreate the original file systems that were on the failed disk.
Note –
Be sure to create the /global/.devices/node@nodeid file system.
-
Mount the root (/) file system on a temporary mount
point.
# mount device temp-mountpoint
|
-
Restore the root (/) file system from backup.
# cd temp-mountpoint
# ufsrestore rvf dump-device
# rm restoresymtable
|
-
Create an empty install-db file.
This puts the node in VxVM install mode at the next reboot.
# touch /temp-mountpoint/etc/vx/reconfig.d/state.d/install-db
|
-
Remove or comment out the following entries from the /temp-mountpoint/etc/system file.
* rootdev:/pseudo/vxio@0:0
* set vxio:vol_rootdev_is_volume=1
|
-
Edit the /temp-mountpoint/etc/vfstab file and replace all VxVM mount points with the
standard disk devices for the root disk, such as /dev/dsk/c0t0d0s0.
Example:
Change from—
/dev/vx/dsk/rootdg/rootvol /dev/vx/rdsk/rootdg/rootvol / ufs 1 no -
Change to—
/dev/dsk/c0t0d0s0 /dev/rdsk/c0t0d0s0 / ufs 1 no -
|
-
Unmount the temporary file system and check the file system.
# cd /
# umount temp-mountpoint
# fsck raw-disk-device
|
-
Install the boot block on the new disk.
# /usr/sbin/installboot /usr/platform/`uname -i`/lib/fs/ufs/bootblk raw-disk-device
|
-
Reboot the node in single-user mode.
-
Update the disk ID using scdidadm(1M).
# scdidadm -R /dev/rdsk/c0t0d0
|
-
Run vxinstall to encapsulate the disk and reboot.
-
If there is a conflict in minor number with any other system, unmount
the global devices and reminor the disk group.
-
Shut down and reboot the node in cluster mode.
SPARC: Example—Restoring an Encapsulated root (/)
File System (VERITAS Volume Manager)
The following example shows an encapsulated root (/)
file system restored to the node phys-schost-1 from the
tape device /dev/rmt/0.
[Replace the failed disk and boot the node:]
|
Boot the node from the Solaris CD. At the OpenBoot PROM ok prompt, type the following command:
ok boot cdrom -s
...
[Use format and newfs to create partitions and file systems]
[Mount the root file system on a temporary mount point:]
# mount /dev/dsk/c0t0d0s0 /a
[Restore the root file system:]
# cd /a
# ufsrestore rvf /dev/rmt/0
# rm restoresymtable
[Create an empty install-db file:]
# touch /a/etc/vx/reconfig.d/state.d/install-db
[Edit /etc/system on the temporary file system and
remove or comment out the following entries:]
# rootdev:/pseudo/vxio@0:0
# set vxio:vol_rootdev_is_volume=1
[Edit /etc/vfstab on the temporary file system:]
Example:
Change from—
/dev/vx/dsk/rootdg/rootvol /dev/vx/rdsk/rootdg/rootvol / ufs 1 no-
Change to—
/dev/dsk/c0t0d0s0 /dev/rdsk/c0t0d0s0 / ufs 1 no -
[Unmount the temporary file system, then check the file system:]
# cd /
# umount /a
# fsck /dev/rdsk/c0t0d0s0
[Install a new boot block:]
# /usr/sbin/installboot /usr/platform/`uname \
-i`/lib/fs/ufs/bootblk /dev/rdsk/c0t0d0s0
[Reboot in single-user mode:]
# reboot -- "-s"
[Update the disk ID:]
# scdidadm -R /dev/rdsk/c0t0d0
[Run vxinstall:]
# vxinstall
Choose to encapsulate the root disk.
[If there is a conflict in minor number, reminor the rootdg disk group:]
# umount /global/.devices/node@nodeid
# vxdg reminor rootdg 100
# shutdown -g0 -i6 -y
|
Where to Go From Here
For instructions about how to mirror the encapsulated root disk, see
the Sun Cluster
Software Installation Guide for Solaris OS.
Scarica il manuale in formato PDF (1037 KB)