On Mon, Dec 19, 2022 at 2:58 PM Neal Becker <ndbecker2(a)gmail.com&gt; wrote: Something wrong with btrfs filesystem mkswap. If I do it the old-fashioned way: # truncate -s 0 swapfile# chattr +C swapfile# fallocate -l 2G swapfile# chmod 0600 swapfile# mkswap swapfile# swapon swapfile It works correctly

On Mon, Dec 19, 2022 at 2:10 PM Samuel Sieb <samuel(a)sieb.net&gt; wrote: Not quite, without the chattr +C it will still perform COW file operations which you don't want on large files. Thanks, Richard

On Dec 19, 2022, at 15:10, Samuel Sieb <samuel(a)sieb.net&gt; wrote: I totally missed the size discrepancy. I believe in new btrfs versions, you can use: # btrfs filesystem mkswapfile swapfile # swapon swapfile https://btrfs.readthedocs.io/en/latest/Swapfile.html Otherwise you need to set the No_COW attribute with `chattr +C /path/to/swap/directory`. -- Jonathan Billings

On Dec 19, 2022, at 14:59, Neal Becker <ndbecker2(a)gmail.com&gt; wrote: This is pretty much expected behavior. Your zram swap device has a priority of 100, the swap file of -2. See the swap(2) man page: https://linux.die.net/man/2/swapon Basically, you want to have swapped pages in the compressed memory device, it is much faster and essentially is expanding available RAM. If memory pressure gets high enough it will use your swap file, but unlikely unless your system is really thrashing. -- Jonathan Billings

FWIW there are some fixes and enhancements coming in btrfs-progs 6.1, now in koji for Rawhide, including making it easier to get info about hibernation file offset in a swapfile. I haven't messed with the new subcommand, but I personally prefer putting swapfiles in their own subvolume so that I can still snapshot the root subvolume. Snapshotting a subvolume containing a swapfile will render the swapfile invalid for use (or maybe the snapshot fails, not sure, haven't tried recently). The way I do this is mount the top-level of the file system (just do a normal mount without any options), and inside you'll see what appears to be two directories: root and home. Those are the subvolumes the installer creates by default. Create a new subvolume in here and add it to fstab such that: UUID=$fsuuid /var/swap btrfs noatime,subvol=swap 0 0 I use chattr +C on this swap subvolume, that way any new files created inside will inherit. This is something the new subcommand will do for you. An additional entry in fstab: /var/swap/swapfile1 none swap defaults 0 0 You can certainly make a nested /var/swap thereby avoiding the need to create the earlier fstab entry. But note that snapshots still don't have this nested subvolume in them, so if you do a rollback it also won't have the nested swap subvolume or file - thus you boot probably hangs because the fstab is looking for this swapfile to activate and never finds it. So I just do it the way I describe, that way I can more or less forget about it. But an alternative to that, if you really prefer nested, is s/defaults/nofail/ for the swapfile entry and now a missing swap won't cause boot to fail *but* you also may one day forget all this and come to realize that there's no swap activated because you once did a rollback way back when... :D -- Chris Murphy