The Dell MD storage allows a maximum of 64TB virtual disk. We have some imaging staff that require more than a 100TB of storage preferably in one location.

I'll create 2 x 64TB vdisk on the same dell storage. It is then SAS connected to the server and I'll be using LVM to create a volume group consisting of 2 virtual disk to form a 128TB logical volume.

Say few years down the line. They run out of space and we need to SAS connect a separate dell storage block to the server. Now the server has two separate blocks of storage. I then add more disks from the second block of storage to volume group.

Now the volume group consist of two virtual disk from the first block and two virtual disk on the second block of storage. What would happen when one of the blocks goes offline? Would I then immediately corrupt my volume?


The volume (LV) will go into partial mode (see p flag in lvs output) but you may be still able to read and write to the disk unless the missing parts are accessed which will result in I/O errors (I am not saying it is a good idea to continue using filesystem in such state.)

Some applications or filesystems may not handle I/O failures well and you may loose some writes which have not made it to the disk but with journalling FS (like ext4) it is unlikely you would get FS corrupted beyond repair.

You will not be able to activate or modify partial logical volume (e.g. resize it) and it is fine. In general you do not want activating it.

The worst thing you could do at this moment is to run fsck. Do not. Not until the volume is back. Otherwise you may as well say good bye to a large part of your data.

If other LVs were added/removed while the disk was missing, you will need to run vgextend --restoremissing VG PV which will make the Volume group whole again (see m flag in pvs output.)

The mounted FS may not fully recover and you may need to umount first, (optionally running fsck now) and mount it back.

You may also want to consider setting up multipath (even with one path), which is able to hide short term outages from the system, as I/O will be queued.


That depends on how you create your logical volumes...

The lvcreate has the --type option to select among others a mirror and a number RAID level that can handle the loss of an underlying physical volume at the LVM level.

If you use RAID the raid_fault_policy determines what happens at the loss of physical volume.

If you don't set up RAID, would you lose everything when you have a physical volume failure in a LVM that does NOT use RAID? Yes.

  • The physical disks (virtual disk) are already raided with what I believe is Dell proprietary raid. It is very unlikely that I'll create those vdisk from two separate storage blocks. But just in case we have done that and we didn't select to create a logical volume with raid. If the one of the disk was offline and brought back online at a later stage. Would the logical volume now be corrupted? Or that LVM is smart enough to know one of the physical device is gone and would not corrupt it? – lbanz Mar 9 '16 at 18:34
  • This is incorrect. In case of transient error it is likely there will be some lost data, but it is not such a disaster, see my answer. – Martian Mar 11 '16 at 15:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.