module 3
module3 troubleshooting Non-Bootable LVM Disks and LVM Structures
Recovering Lost or Damaged Structures
1. device files (restore from tape backup or recreate using insf and mknod)
Re-create PV device files:
#ioscan –C disk
#insf –eC disk
Re-create a VG device file:
#mkdir /dev/vg01
#mknod /dev/vg01/group c 64 0x010000
#chmod 755 /dev/vg01
Re-create LV device files:
#mknod /dev/vg01/data b 64 0x010001
#mknod /dev/vg01/rdata c 64 0x010001
#chmod 640 /dev/vg01/*
如果不确定哪个逻辑卷名称和VG中的minor number已经被使用，可以用脚本来创建所有255个devices files
#!/usr/bin/sh
vgname=vg01
vgnum=01
mkdir /dev/$vgname
for m in 0 1 2 3 4 5 6 7 8 9 a b c d e f
do
for n in 0 1 2 3 4 5 6 7 8 9 a b c d e f
do
if [[$m$n=00]] ; then
mknod /dev/$vgname/group c 64 0x$00$m$n
else
mknod /dev/$vgname/lvol$m$n b 64 0x$00$m$n
mknod /dev/$vgname/rlvol$m$n c 64 0x$00$m$n
fi
done
done
then run vgdisplay . Logical volumes with size 0M may be safely removed. Mount the remaining logical volumes to determine their contents and guess their names. Then unmount and rename the device files using the mv command
#mv /dv/vg01/lvol1 /dev/vg01/data
#mv /dev/vg01/rlovl1 /dev/vg01/rdata
Corrupt /etc/lvmtab:
Cp /etc/lvmtab /etc/lvmtab.bkp
Vgscan –p (先预览可能的结果)
vgscan
2. /etc/lvmtab (restore from tape or rebuild using information from PVRAs/VGRAs via vgscan)
3. kernel Structures (Re-scan PVRAs/VGRAs using vgchange –a y)
4. PVRA/VGRA (restore from /etc/lvmconf backup directory)
Vgscan的副作用及修改
1． Vgchange –a to activate all volume groups
2． Use lvlnboot with –R option to correct boot information on disk
3． Use vgreduce to reduce any alternate links that were added to the /etc/lvmtab file by vgscan,but they were not needed
4． If the original primary path of a disk s become an alter nate path after /etc/lvmtab file is reconstructed ,the order can be easily reverted by using vgreduce to remove the primary path and use vgextend to add the path back again.
当卷组激活时，lvm 使用组成卷组的磁盘上的PVRA/VGRA结构来组成LVM的内核结构。然后LVM伪驱动使用这些内核结构来系统的lvm配置。Vgdisplay vgdisplay pv display 等命令使用lvm 内核结构。
l When a pv that is non-responsive at activation time comes back online
l When a failed /corrupted PV is replaced without deactivating/reactivating the VG
Ioscan
Vgchange –a y vg01
Vgdisplay –v vg01
Failed Disks
#tail /var/adm/syslog/syslog.log 包含POWERFAILED信息
如果同时包含pv[#] returned to vg[#]，那么可能是高io 和设置的timeout值比较小
pvchange –t 180 /dev/dsk/disk_device
syslog.log文件中的lbolt信息
#tail /var/adm/syslog/syslog.log
vmunix:SCSI: Request timeout; Abort – lbolt : 137056 , dev: lf070500
lf是十进制的31，lsdev告诉我们31是sdisk SCSI driver的block major#
07代表c7 这里是2位
0代表t0 这里是1位
5 接着的2位数字代表d* ,这里是d5,也有可能是d15等。
00：最后2位数字跟磁带设备相关，磁盘设备始终是00
if vgdisplay and pvdisplay say “couldn’t query disk …”，that typically means that the VGRA/PVRA headers weren’t accessible when the volume group was activated
see if you can access the disk using the dd command ,if you get an error message. Or if the dd command doesn’t successfully read all 1024 records ,th disk may be physically corrupted
# dd if=/dev/rdsk/c0t1d0 of=/dev/null bs=1024k count=1024
1024+0 records in
1024+0 records out
Activating a volume group that contains Failed Disks
覆盖quorum激活卷组
vgchange –a y –q n vg01
这样允许vg激活，但是非冗余的lv可能不可用，要是vg能正常工作，不损失数据，损坏的磁盘必须要进行恢复。
对于启动盘镜像的情况下，在启动时也可以覆盖quorum参数
ISL >hpux –lq (Parisc)
Hpux>boot vmnuix –lq (IPF)
Restoring a Powerfailed Disk
If a disk powerfails afeer activation. No administrator intervention is required when the disk returns
If a disk is powerfailed before activation. Execute ioscan , vgchange vgsync when the disk returns.
Ioscan –fC disk
Vgchange –a y vg01
Vgsync vg01
Replacing a Failed Disk
1. Restore critical LVM structures
2. Recreate FS metadata
3. Restore user data
Vgcfgbackup vg01 (备份PVRA/VGRA到默认的目录 /etc/lvmconf)
Vgcfgbackup –f /tmp/vg01.conf vg01 (备份PVRA/VGRA到指定的文件)
当VG中少disk时，要create ,extend ,reduce logical volume时，需要指定 –A n参数，否则这些命令不会成功（因为physical volume 丢失
