Hi there!

I need some advice about storage :-) I just recently tried to expand my raid5 (ext4) on a Debian 8.5 box only to find out that it will not accept a size over 16TB (32bit file system). I am buying disks to backup my data and start over.
I have 8 3TB disks and I want to use them in a raid5 system, preferably all appearing in one single volume.

What I wonder is if it is possible to make 2 raid5(4 disks each) with LVM on top and make it appear as 1 volume (and mount it all in the same dir)?
Or would it be better to use a 64bit file system (liks XFS) and put it all in 1 raid5?

Any advice and help greatly appreciated.

ext4 these days should build for 64-bit automagically - if you created the filesystem a while ago that may not have been true at the time. Check /etc/mke2fs.conf for "auto_64-bit_support = 1". So, just mkfs.ext4 on your new disks should work "out of the box".
Edit: build just the one RAID5.

Depending on how recent the Debian LVM support is, you can save all the angst and just allow LVM handle it all - don't predefine the RAID. Works really well, and allows you to define a failure policy so you can have spares re-assigned at the time of failure automagically. My preferred usage on Arch/Fedora.

2 members found this post helpful.

Thank you for your reply. The system was built about 3 years ago on Ubuntu OS, and for sure 64-bit was not automatically build at the time. I guess it should do that now with latest stable version of Debian installed, but I also read somewhere http://serverfault.com/questions/604...-64bit-support that even if the flag for auto_64-bit_support was set to 1 the file system was 32bit. It takes a long time to build a 21 TB array so I just wanna make sure I don't have to do it again.

Nor sure what you meant about letting LVM handle it all, but then again I have not used LVM before so really need to read up on it.
Edit: If there is a tutorial somewhere you would recommend, I would be grateful for a link.

Current ext4 will apparently automatically configure itself as "64bit" if the device you are doing the mkfs on is over 16TiB, so you should be ok. And for ext4 the mkfs is (relatively) quick. I haven't access to a device that large, so I haven't actually tested this - however, I just created a 5 Gig file that I mkfs'd on with "-O 64bit", and that worked according to tune2fs. This is supported for when a filesystem is allocated less than 16 TiB but may later be enlarged over that limit (without reformatting).
See "man ext4" for the only sensible mention of 64bit I could find.

As for LVM I only mentioned it as you had - the Redhat LVM Admin Guide is probably as good a reference as any. I always have a downloaded copy around.
Edit: (always forget something ...) see "man lvcreate" for description of the RAID support keywords, but go get the manual anyway.

1 members found this post helpful.

OK, great:-) I will leave this open for now so I can get back to it when I have backup up my data and configured my system. Thanks for your help.

Why do you need to backup ?.
The old disks are the backup - just mkfs the new (RAID) device and see if it satisfies your needs; the data copy can be done any time.

I am marking this as solved now. I needed to back up because the old disks was part of the new array and to create the raid with all the disks in mdadm (without adding disks afterwards) I had to back everything up. I ended up putting it all in one raid5 because that gives the most use of the space. LVM was not really needed in my case. I formatted the array in XFS after reading around about what would be most useful. Thanks again for help.