A quick question of the upcoming kernel switch.

[lazlo.vii]

With today's announcement of the upcoming switch to debian-sources-lts I have started thinking about stuff on my server that might be affected.  While I want to track as much of the  stable Funtoo Portage tree as can I am unsure what affects downgrading the kernel will have.  I use btrfs for my RAID10 and that is my main concern.  I don't really know what else might be affected because of the huge amount of work that goes into the kernel.  I try to keep a my server simple and just install what I need or really want so I don't think it will be a lot.

 

Short of downgrading to debian-sources-lts and testing it right now, is there an easy way to predict some the changes before hand so I be on the look out for trouble?

[palica]

https://btrfs.wiki.kernel.org/index.php/Changelog#By_feature

don't know how much debian backports btrfs fixes (or new features)

changelog here:

https://salsa.debian.org/kernel-team/linux/blob/046a5da6eb748c59f7402d5aae84ab7543dbb396/debian/changelog?expanded=true&viewer=simple

downgrade could be problematic if you use ZSTD as your compression for example, but not limited to this probably.

IMHO - test the downgrade on a separate pc or disk. (and never touch a running system)

at least for those running btrfs this kernel switch should be done with extreme caution.

[lazlo.vii]

Thank you for the information.  I don't really own another system that fully compares with my server, but I can install Funtoo to a new btrfs partition on my desktop pc, copy over the  world file from the server, build the same software, and then try the kernel down grade.  It will take some time but might give me a good idea of the kinds of problems I might encounter.

[palica]

it is not that much about software becoming incompatible with older release, it is more of any patches and fixes that didn't make it into 4.9-lts branch that you probably used, at the time of creation of the system, some BTRFS features that were in 4.14 branch and are not in 4.9.

so creating a loop device btrfs system with 4 disks and creating a raid10 with 4.14 and then turning on all the flags like compress and such on the fs and putting some dummy data on there, and afterwards downgrading the kernel and mounting the filesystem and comparing for example the checksums of the dummy data or performing a scrub could give you information about if this downgrade is safe for your server, but also don't take this for granted and have backups!!!

couple of ideas how to test without 4 harddrives can be found here: https://www.funtoo.org/BTRFS_Fun

[lazlo.vii]

This morning I gave it some thought and decided I will set up a VM with 4x 4GB LV's on my ssd.  I still have my copy of SRCD 5.2.2 so I will install in the same manner I used for the host server, which was to just use one drive formatted in btrfs for the install and then convert to RAID10 once the system had been rebooted and the profile set. Once that is done I can switch kernels and see what happens.  If I run into any problems because of the switch I will not lose data because it will be a VM and because it is VM I can walk away and leave it paused if need be. 

 

If using different versions of the kernel is going to cause issues I think is good way to test for it because I'll end up using one kernel (4.14.32) for the creation of the initial btrfs partition, a second (4.14.12-2) for the creation of the RAID10, and a third (4.9.130) to test scrubbing, compression, rebalancing, and anything else I will have time for. 

 

The different SRCD versions and the kernel used in them are listed here if anyone needs them:

http://www.system-rescue-cd.org/Changes-x86/

 

After this test I do not think I will use SRCD any more for installing btrfs based Funtoo systems.  A Devuan or Debian livecd seems like a much safer choice.

[palica]

good plan. let us know what you find out.

[lazlo.vii]

I have changed my mind about the size of LV's and decided to go with 8GB instead of 4GB so I don't have to guess about having enough space in /*/tmp to compile a new kernel.  I'll edit this post to add my results when I get done.  Hopefully that will be within the next 3 hours.

 

Well, I have run basic tests with defrag and zlib compression and changed RAID back and forth from 0,1, and 10, added large amounts of data, deleted some of it, did it all again, made a few snap shops, deleted them, and so on.

 

I can't find any obvious problems with the usage or functionality of kernel 4.9 on a btrfs filesystem created with newer kernel.  I know that data point doesn't make a statistically useful sample, but it does put my mind at ease.