Large File System out of lots of small chunks of free space.....
Posted on Thu, Dec 05, 2024 at 10:03 by Hubertus A. Haniel
DON'T TRY THIS AT WORK!
Just because it is possible it does
not mean it is good practice!
I would not endorse this
trickery/hack in a commercial environment.
This would not be supported in an enterprise environment and is probably not the safest way of keeping your data so I would not recommend this in a critical environment with critical data. On top of this not being the safest way to store your data it will also not be very efficient and I would expect a performance impact as a result of this. So this is just a bit of fun and may just help you out with a temporary fix to get you over a hurdle. I have done this on RHEL9 but it will work in the same way on other Linux distributions
Lets say you have a system with lots of file systems of which the size does not really matter but there are a few gigabytes here and there and you may even be able to add NFS mounted stuff although if there is a network failure you may end up with corruptions when the NFS parts fail. You can not shrink or rearrange the file systems to free up enough to store a larger file. In this example:
/filesystem-01
/filesystem-02
/filesystem-03
/filesystem-04
It is irrelevant of how big these file systems are but let say each of these have only about 25gb free but you want to create a file that is in the region of 80gb to 100gb for what ever reason. So lets go ahead and in each of these file systems we will create a sparse file (We can just create a full size file but a sparse file is faster to create. You will find that after that step the real space is not actually being used yet but with ls -al you will see the file size is shown as what it could grow to.)
dd if=/dev/zero of=/filesystem-01/filestore.img bs=1 count=0 seek=25G
dd if=/dev/zero of=/filesystem-02/filestore.img bs=1 count=0 seek=25G
dd if=/dev/zero of=/filesystem-03/filestore.img bs=1 count=0 seek=25G
dd if=/dev/zero of=/filesystem-04/filestore.img bs=1 count=0 seek=25G
Next we will create loop back devices that point to these files:
losetup --show -f /filesystem-01/filestore.img
losetup --show -f /filesystem-02/filestore.img
losetup --show -f /filesystem-03/filestore.img
losetup --show -f /filesystem-04/filestore.img
These will most likely end up being /dev/loop0 through to /dev/loop3 but if you have other loop back stuff mounted it may differ. The command "losetup -a" will list them for you.
We can now create a raid device on top of these loop devices with level raid 0 to have on continuous device:
mdadm -C /dev/md/filestore -l 0 -n 4 /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3
You can now treat /dev/md/filestore like a normal disk device. So you can partition it create one large filesystem on it.
When you want to unmount it (before rebooting for example although Linux may do it for you the steps are:
- Unmount the file system
- Stop the raid device ("mdadm --stop /dev/md/filestore"
- Remove the loop devices ("losetup -d <for each device created>")
To remount the device again the steps are:
- "losetup --show -f" for each of the image files as above
- "mdadm --assemble /dev/md/filestore /dev/loop0 /dev/loop1 /dev/loop2 /dev/loop3"
- mount the device as previously
If you are going to keep this setup for a longer time you may want to script the above to ensure it gets done on boot etc. - The same is probably archievable just with plain LVM but I have not attempted that and I suspect it may be more intrusive as you may have to modify lvm.conf to scan the loop devices. You also run in the danger of messing up your lvm meta data so I did not want to take the risk on creating volume groups on files that are already on top of a volume group which is the case on my system. You also would not want to extend existing logical volumes on to files bearing in mind that you may not be able to shrink stuff down when you want to remove it again.
Edited on: Thu, Dec 05, 2024 13:42Posted in HowTo (RSS), System - Linux (RSS)
