Nikolay Borisov writes: > > > > Unfortunately the cfq scheduler did not help. The system wedged. > > > > I did notice this for the first time... > > > > [Wed Jan 9 06:03:41 2019] BTRFS info (device sda1): the free space cache file (83320273633280) is invalid, skip it > > What you could do is mount btrfs with -o clear_cache to make btrfs > rebuild the freespace cache. > > > > > anything I should do about that? > > > > The messages were similar... > > > > [Wed Jan 9 23:52:04 2019] INFO: task nfsd:2997 blocked for more than 120 seconds. > > [Wed Jan 9 23:52:04 2019] Not tainted 4.17.14-custom #1 > > [Wed Jan 9 23:52:04 2019] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. > > [stuff deleted] > > [Wed Jan 9 23:54:07 2019] RBP: 00007f0c3f348c60 R08: 00000000000000ff R09: 0000000000001000 > > [Wed Jan 9 23:54:07 2019] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000a6672f > > [Wed Jan 9 23:54:07 2019] R13: 00007f0c918affe0 R14: 0000000000000000 R15: 00007f0c918affb0 > > > > These don't tell the full story, what seems to be happening is that > stuff is waiting for transaction to finish but it's not evident which > thread is holding the transaction. Please, paste the output of > "echo w > /proc/sysrq-trigger" so we have full picture of what's blocked > where. > Thanks. I have not seen the free space cache message with the new kernel. Next time the hang pops up I'll echo w > /proc/sysrq-trigger and send it along. I am hoping that the new kernel will magically fix the problem. Thanks again, sb. Scott Blomquist
