In this article we will learn
- How to recover LVM2 partition (Restore deleted LVM)
- How to restore PV (Physical Volume) in Linux
- How to restore VG (Volume Group) in Linux
- How to restore LVM metadata in Linux
Earlier we had a situation wherein the LVM metadata from one of our CentOS 8 node was missing. Due to this all the logical volumes, volume groups and physical volumes mapped to that LVM metadata was not visible on the Linux server. So we had to restore LVM metadata from the backup using vgcfgrestore. I will share the steps to reproduce the scenario i.e. manually delete the LVM metadata and then steps to recover LVM2 partition, restore PV, restore VG and restore LVM metadata in Linux using vgcfgrestore.
vgcfgbackup can be used to manually create LVM backups, as these backups are very helpful and can also be used in LVM Disaster Recovery.
Prepare Lab Environment
Before we go ahead with the steps to recover LVM2 partition in Linux, we must first prepare Lab Environment with logical volumes. Next we will manually delete lvm metadata to reproduce the issue scenario.
I have created a Virtual Machine with CentOS 8 OS using Oracle VirtualBox which is installed on a Linux server. Next I added an additional virtual disk to this VM which is mapped to /dev/sdb.
Create Physical Volume
The first step is to create physical volume using pvcreate
[root@centos-8 ~]# pvcreate /dev/sdb
Physical volume "/dev/sdb" successfully created.
Create Volume Group
Next create a new Volume Group, we will name this VG as test_vg.
[root@centos-8 ~]# vgcreate test_vg /dev/sdb
Volume group "test_vg" successfully created
List the available volume groups using vgs. I currently have two volume groups wherein rhel volume group contains my system LVM2 partitions
[root@centos-8 ~]# vgs
VG #PV #LV #SN Attr VSize VFree
rhel 1 2 0 wz--n- <14.50g 0
test_vg 1 0 0 wz--n- <8.00g <8.00g <-- new VG
Create Logical Volume
Create a new logical volume test_lv1 under our new volume group test_vg
[root@centos-8 ~]# lvcreate -L 1G -n test_lv1 test_vg
Logical volume "test_lv1" created.
Create File System on the Logical Volume
Create ext4 file system on this new logical volume
[root@centos-8 ~]# mkfs.ext4 /dev/mapper/test_vg-test_lv1
mke2fs 1.44.6 (5-Mar-2019)
Creating filesystem with 262144 4k blocks and 65536 inodes
Filesystem UUID: c2d6eff5-f32f-40d4-88a5-a4ffd82ff45a
Superblock backups stored on blocks:
32768, 98304, 163840, 229376
Allocating group tables: done
Writing inode tables: done
Creating journal (8192 blocks): done
Writing superblocks and filesystem accounting information: done
List the available volume groups along with the mapped storage device. Here as you see test_vg is mapped to /dev/sdb
[root@centos-8 ~]# vgs -o+devices VG #PV #LV #SN Attr VSize VFree Devices rhel 1 2 0 wz--n- <14.50g 0 /dev/sda2(0) rhel 1 2 0 wz--n- <14.50g 0 /dev/sda2(239) test_vg 1 1 0 wz--n- <8.00g <7.00g /dev/sdb(0)
Similarly you can see the new logical volume test_lv1 is mapped to /dev/sdb device
[root@centos-8 ~]# lvs -o+devices
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert Devices
root rhel -wi-ao---- 13.56g /dev/sda2(239)
swap rhel -wi-ao---- 956.00m /dev/sda2(0)
test_lv1 test_vg -wi-a----- 1.00g /dev/sdb(0) <-- new Logical Volume
Add some data to Logical Volume
We will put some data into our logical volume to make sure there are no data loss after we recover LVM2 partition, restore PV and restore VG using LVM metadata in the next steps.
[root@centos-8 ~]# mkdir /test [root@centos-8 ~]# mount /dev/mapper/test_vg-test_lv1 /test/
Create a dummy file and note down the md5sum value of this file
[root@centos-8 ~]# touch /test/file [root@centos-8 ~]# md5sum /test/file d41d8cd98f00b204e9800998ecf8427e /test/file
Next un-mount the logical volume
[root@centos-8 ~]# umount /test/
How to manually delete LVM metadata in Linux?
To manually delete LVM metadata in Linux you can use various tools such as wipefs, dd etc. wipefs can erase filesystem, raid or partition-table signatures (magic strings) from the specified device to make the signatures invisible for libblkid. wipefs does not erase the filesystem itself nor any other data from the device.
In this example we will use wipefs to delete LVM metadata from /dev/sdb device. Since the device in question /dev/sdb is in use by Volume Group, we have to use -f to forcefully wipe the LVM metadata
[root@centos-8 ~]# wipefs --all --backup -f /dev/sdb
/dev/sdb: 8 bytes were erased at offset 0x00000218 (LVM2_member): 4c 56 4d 32 20 30 30 31
We have used --backup so that before deleting the LVM metadata, wipefs will create a backup of the ext4 signature containing LVM metadata under the home folder of the user who is executing the command. Since we used root user, our LVM metadata backup is stored under root user's home folder.
[root@centos-8 ~]# ls -l /root/wipefs-sdb-0x00000218.bak -rw------- 1 root root 8 Apr 5 13:45 /root/wipefs-sdb-0x00000218.bak
dd if=~/wipefs-sdb-0x00000218.bak of=/dev/sdb seek=$((0x00000218)) bs=1 conv=notruncNext you can verify that all the logical volumes, volume groups and physical volume part of /dev/sdb is missing from the Linux server
[root@centos-8 ~]# lvs -o+devices
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert Devices
root rhel -wi-ao---- 13.56g /dev/sda2(239)
swap rhel -wi-ao---- 956.00m /dev/sda2(0) <--Our Logical volume no more visible
[root@centos-8 ~]# vgs
VG #PV #LV #SN Attr VSize VFree
rhel 1 2 0 wz--n- <14.50g 0 <-- test_vg no more visible
[root@centos-8 ~]# pvs
PV VG Fmt Attr PSize PFree
/dev/sda2 rhel lvm2 a-- <14.50g 0 <-- /dev/sdb no more visible
Similarly with lsblk also we can verify that there are no LVM2 partitions under /dev/sdb
[root@centos-8 ~]# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 15G 0 disk
├─sda1 8:1 0 512M 0 part /boot
└─sda2 8:2 0 14.5G 0 part
├─rhel-root 253:0 0 13.6G 0 lvm /
└─rhel-swap 253:1 0 956M 0 lvm [SWAP]
sdb 8:16 0 8G 0 disk
sr0 11:0 1 1024M 0 rom
sr1 11:1 1 1024M 0 rom
Step 1: List backup file to restore LVM metadata in Linux
- LVM metadata backups and archives are automatically created whenever there is a configuration change for a volume group or logical volume, unless this feature is disabled in the
lvm.conffile. - By default, the metadata backup is stored in the
/etc/lvm/backupfile and the metadata archives are stored in the/etc/lvm/archivefile. - How long the metadata archives stored in the
/etc/lvm/archivefile are kept and how many archive files are kept is determined by parameters you can set in thelvm.conffile. - A daily system backup should include the contents of the
/etc/lvmdirectory in the backup. - You can manually back up the LVM metadata to the
/etc/lvm/backupfile with thevgcfgbackupcommand. - You can restore LVM metadata with the
vgcfgrestorecommand.
To list the available backups of LVM metadata use vgcfgrestore --list. Currently we have three backup stages where the last backup was taken after we created test_lv1 logical volume.
[root@centos-8 ~]# vgcfgrestore --list test_vg File: /etc/lvm/archive/test_vg_00000-1327770182.vg VG name: test_vg Description: Created *before* executing 'vgcreate test_vg /dev/sdb' Backup Time: Sun Apr 5 13:43:26 2020 File: /etc/lvm/archive/test_vg_00001-1359568949.vg VG name: test_vg Description: Created *before* executing 'lvcreate -L 1G -n test_lv1 test_vg' Backup Time: Sun Apr 5 13:44:02 2020 File: /etc/lvm/backup/test_vg VG name: test_vg Description: Created *after* executing 'lvcreate -L 1G -n test_lv1 test_vg' Backup Time: Sun Apr 5 13:44:02 2020
So we will use the last backup i.e. /etc/lvm/backup/test_vg to restore LVM metadata till the stage where test_lv1 was created.
Step 2: Restore PV (Physical Volume) in Linux
You must perform proper pre-checks and take backup of your file system before executing these steps in production environment to prevent any data loss.
- It is very important that to restore PV, you create the new PV using the same UUID as it was earlier or else restore VG and recover LVM2 partition will fail in the next steps.
- You can get the UUID of your Physical Volume from backup file "
/etc/lvm/backup/test_vg" - Below is a sample content of physical_volumes from the backup file. If you have more than one physical volumes then you need to search for the missing PV's UUID
- In my case
SBJi2o-jG2O-TfWb-3pyQ-Fh6k-fK6A-AslOg1is the UUID of the missing PV so I will use this to restore PV in Linux
physical_volumes { pv0 { id = "SBJi2o-jG2O-TfWb-3pyQ-Fh6k-fK6A-AslOg1" device = "/dev/sdb" # Hint only status = ["ALLOCATABLE"] flags = [] dev_size = 16777216 # 8 Gigabytes pe_start = 2048 pe_count = 2047 # 7.99609 Gigabytes } }
Next again it is important that you test the physical volume restore. We use --test mode to verify the operation. With --test commands will not update LVM metadata. This is implemented by disabling all metadata writing but nevertheless returning success to the calling function.
So here I have provided the same UUID of /dev/sdb as we collected earlier, followed by the backup file we want to use to restore PV and then the device name using which we will perform pvcreate. The pvcreate command overwrites only the LVM metadata areas and does not affect the existing data areas.
[root@centos-8 ~]# pvcreate --test --uuid "SBJi2o-jG2O-TfWb-3pyQ-Fh6k-fK6A-AslOg1" --restorefile /etc/lvm/backup/test_vg /dev/sdb TEST MODE: Metadata will NOT be updated and volumes will not be (de)activated. WARNING: Couldn't find device with uuid SBJi2o-jG2O-TfWb-3pyQ-Fh6k-fK6A-AslOg1. Physical volume "/dev/sdb" successfully created.
With --test mode we know that the command execution is successful. So we will run the same command without --test to restore PV in real.
[root@centos-8 ~]# pvcreate --uuid "SBJi2o-jG2O-TfWb-3pyQ-Fh6k-fK6A-AslOg1" --restorefile /etc/lvm/backup/test_vg /dev/sdb
WARNING: Couldn't find device with uuid SBJi2o-jG2O-TfWb-3pyQ-Fh6k-fK6A-AslOg1.
Physical volume "/dev/sdb" successfully created.
Next verify the list of available Physical Volumes
[root@centos-8 ~]# pvs
PV VG Fmt Attr PSize PFree
/dev/sda2 rhel lvm2 a-- <14.50g 0
/dev/sdb lvm2 --- 8.00g 8.00g <-- Now /dev/sdb is visible
Step 3: Restore VG to recover LVM2 partition
- After we restore PV, next step is to restore VG which will further recover LVM2 partitions and also will recover LVM metadata.
- Similar to
pvcreate, we will execute vgcfgrestore with--testmode to check the if restore VC would be success or fail. - This command will not update any LVM metadate
[root@centos-8 ~]# vgcfgrestore --test -f /etc/lvm/backup/test_vg test_vg TEST MODE: Metadata will NOT be updated and volumes will not be (de)activated. Restored volume group test_vg.
As we see that the command execution in --test mode was successful so now we can safely execute our command to restore VG and recover LVM2 partition in Linux using vgcfgrestore.
[root@centos-8 ~]# vgcfgrestore -f /etc/lvm/backup/test_vg test_vg
Restored volume group test_vg.
Using vgs your can check if restore VG was successful.
[root@centos-8 ~]# vgs
VG #PV #LV #SN Attr VSize VFree
rhel 1 2 0 wz--n- <14.50g 0
test_vg 1 1 0 wz--n- <8.00g <7.00g <-- test_vg is not visible
Next verify the if you were able to restore deleted lvm and recover LVM2 partition using lvs.
[root@centos-8 ~]# lvs
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
root rhel -wi-ao---- 13.56g
swap rhel -wi-ao---- 956.00m
test_lv1 test_vg -wi------- 1.00g <-- our logical volume is also visible
Step 4: Activate the Volume Group
Next activate the volume group test_vg
[root@centos-8 ~]# vgchange -ay test_vg 1 logical volume(s) in volume group "test_vg" now active
Step 5: Verify the data loss after LVM2 partition recovery
The most crucial part, make sure there was no data loss in the entire process to restore PV, restore VG, restore LVM metadata and recover LVM2 partition.
[root@centos-8 ~]# mount /dev/mapper/test_vg-test_lv1 /test/
If we are able to mount the logical volume so it means our ext4 file system signature is intact and not lost or else the mount would fail.
[root@centos-8 ~]# ls -l /test/
total 16
-rw-r--r-- 1 root root 0 Apr 5 13:45 file
drwx------ 2 root root 16384 Apr 5 13:44 lost+found
Our test file exists and the md5sum matches the value of what we had taken before deleting the LVM metadata
[root@centos-8 ~]# md5sum /test/file
d41d8cd98f00b204e9800998ecf8427e /test/file <-- same as earlier
So overall restore PV, restore VG, restore LVM metadata and recover LVM2 partition was successful.
Lastly I hope the steps from the article to recover LVM2 partition using vgcfgrestore on Linux was helpful. So, let me know your suggestions and feedback using the comment section.
References:
Restore deleted LVM metadata
Performing Logical Volume Backup and restore using vgcfgbackup and vgcfgrestore
Related Searches: lvm backup and restore in linux, lvm disaster recovery, how to restore vg in linux, how to restore pv in linux, how to restore lvm metadata in linux
/etc/lvm/archive/ backup available 10 days back now /root files stored new data then can we vgcfgrestore 10 backup recovery then my 10 days data will loss ?
The steps provided only restore the metadata of LVMs and not the actual data but still it is strongly recommended to test before making any changes in production.
Thanks. You saved my day with this article
Nice. But what if the header wipe happens while the device is still mounted? Then you will see something like (on the
pvcreate --test):Can’t open
/dev/sdbexclusively. Mounted filesystem?!! STOP HERE and DO NOT PROCEED and read on !!
The perhaps most puzzling thing is, that the mounted FS still seems to be fully accessible, but LVM no more sees it. Hence you are at danger to lose everything.
To get back on track (getting rid of the warning above, such that you can proceed) following steps are necessary:
First, quiesce the FS and unmount the FS. (Read: Stop all processes which hog on the FS, then normally umount.) If you are paranoied, do a backup of the FS before the umount, just in case. If the umount fails, this can be because something still uses it which is not a process. DO NOT USE lazy or forced umount (umount -l does not help here), because then it becomes very difficult to find the culprit. Like a swapfile, or loop file (losetup) or binding mount.
Now STILL the device is used by the system. Why? Because it is registered in devicemapper, but LVM cannot deregister it (vgchange -an) because it is no more known to LVM. But it is still known to dmsetup. Hence you must do it yourself.
this lists all devices. You will see something like
Be sure to identify the right one. It is named after the VG dash LV (with dashes in VG’s or LV’s name doubled). Then run
This should free the device such that pvcreate -test will no more show this warning and you can proceed.
However if this fails again, then something still hogs on the device! This can be, again, in things like swap, losetup etc., so if it is not found in
dmsetup, then look elsewhere. If you are completely desperate you can reboot the machine, but in that case devices might become renumbered, so your sdb can show up as sdca.Hi Team,
disk partition is not extending, due to the below-mentioned error.
How to recover LVM.
Can do a vgchange on rootvg