On Mon, Apr 1, 2019 at 10:12 AM Zirconium Hacker <jared.e...@gmail.com> wrote: > > Hi, > > I recently started a balance on a BTRFS filesystem. > At least one process failed to write files before or during the > balance due to a supposed out-of-space condition. > Yes, I know it's not good to allow that to happen. > > Here is the relevant dmesg output: > [694828.101677] BTRFS info (device sdh1): balance: start -dusage=60 > [694828.103543] BTRFS info (device sdh1): relocating block group > 2933242462208 flags data > [694883.648385] BTRFS info (device sdh1): found 76 extents > > At that point, disk I/O according to iotop dropped to almost zero. > After waiting a while, I attempted to cancel the balance, but that > just resulted in a hung cancel process. > > Now, any process that attempts to write to or read from the filesystem hangs. > I suppose that means there's a kernel deadlock. > The output of dmesg after running echo w >/proc/sysrq-trigger is attached. > > kernel: 5.0.3-arch1-1-ARCH x86_64 > btrfs-progs version: v4.20.2 >
Are qgroups enabled? -- Chris Murphy
