Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I want to use LVM to have a partition that spans over multiple drives. I know that there is a down side in that if one drive fails all the data in the array is lost. Is it possible to have LVM not stripe the data and in case a drive fails then only data that has parts of it on the fail drive will be lost? So theoretically i lose only the data which is on the failed drive

share|improve this question
1  
I don't think this is possible -- the metadata has to be stored somewhere, and whichever disks that happen to be storing any metadata could take down the entire partition's structure with them, leaving you only with raw data on the intact drives (which is pretty difficult to recover without metadata). –  Mehrdad Nov 29 '12 at 7:26
    
I was thinking that maybe each drive can store it's own metadata. I don't need redundancy but would like to have the ease of use of not worrying that i don't have enough space on one drive and copy on the other and want to have the data centralized under one entry for easy navigation. –  Alex Linte Nov 29 '12 at 7:29
    
Ah, you'd need a specially-written file system for that. I'm not aware of any that do this but maybe someone else does... good luck in your search! –  Mehrdad Nov 29 '12 at 7:34
add comment

2 Answers

based on this QA you are SOL with LVM and any current/stable file system

http://serverfault.com/q/208155/6625

I looked into btrfs but no luck there yet either!!!

Since you want a "logical volume" spanning 2 or more disks... for the volume to be larger then a single physical disk.. then you have an alternative to using LVM to achive this.

read up on the linux "mdadm" tool which is also know as the Linux Software/Kernel Raid sub-system.

If you have at least 3 drives, then you have the option of using the Linux Software Raid system (mdadm; Multi-Disk ADMin) to create a 3 disk RAID-5 (striping with parity) "disk-array" volume so that if one drive fails, then the array will continue work in a degraded performance level until you can "replace" the physical failed drive with a new one, re-partition with the same size as last, and use mdadm tool to add the new volume to the array and rebuild.. all while working with the system.

RAID-5 "disk-arrays" reduce your 3 disk capacity to the same as a 2 disk volume though.

so, for example.. 3 x 1TB drives in RAID-5 can give you a big (and fast) 2TB volume that spans all 3 drives with striping for performance as well as the stripping of the "parity" data that allows the volume to still work read&write even when one disk fails completely.

None of the above requires LVM either... However.. nothing stops you from using LVM on top of mdadm "disk-arrays"..

share|improve this answer
    
Hi Heston, I am aware of RAID-5 (and all other RAID modes). As i said in a comment to the initial request, i am not interested in redundancy. This is going to be implemented on a home server to host media. I want to get the most of space out of the drives used. So for RAID 5 the total space is n-1 which at this point i'm not interested since the redundancy is not a concern for me. –  Alex Linte Nov 29 '12 at 8:46
    
I checked your last edit and it seams that if a drives fail all LV that had any data on the failed drive will be lost. Since i wanted to use one LV that spans on all the drives, this means i lose everything if one of the drives fail. I guess i'll stick with each drive being mounted individually. –  Alex Linte Nov 30 '12 at 16:46
add comment

LVM does not do striping by default.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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