https://bugzilla.kernel.org/show_bug.cgi?id=200835 Dave Chinner (david@xxxxxxxxxxxxx) changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |david@xxxxxxxxxxxxx --- Comment #2 from Dave Chinner (david@xxxxxxxxxxxxx) --- Memory reclaim is stuck waiting for log forces to complete. The log force is stuck in both cases waiting for iclog IO completion here: Aug 04 04:33:45 hpmicroserver kernel: INFO: task kworker/1:2:30142 blocked for more than 120 seconds. Aug 04 04:33:45 hpmicroserver kernel: Not tainted 4.14.52-1-lts #1 Aug 04 04:33:45 hpmicroserver kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. Aug 04 04:33:45 hpmicroserver kernel: kworker/1:2 D 0 30142 2 0x80000000 Aug 04 04:33:45 hpmicroserver kernel: Workqueue: xfs-cil/dm-4 xlog_cil_push_work [xfs] Aug 04 04:33:45 hpmicroserver kernel: Call Trace: Aug 04 04:33:45 hpmicroserver kernel: ? __schedule+0x284/0x870 Aug 04 04:33:45 hpmicroserver kernel: ? wake_up_q+0x70/0x70 Aug 04 04:33:45 hpmicroserver kernel: schedule+0x28/0x80 Aug 04 04:33:45 hpmicroserver kernel: xlog_state_get_iclog_space+0x105/0x2c0 [xfs] Aug 04 04:33:45 hpmicroserver kernel: ? wake_up_q+0x70/0x70 Aug 04 04:33:45 hpmicroserver kernel: xlog_write+0x163/0x6e0 [xfs] Aug 04 04:33:45 hpmicroserver kernel: xlog_cil_push+0x21d/0x3c0 [xfs] Aug 04 04:33:45 hpmicroserver kernel: process_one_work+0x1e9/0x3b0 Aug 04 04:33:45 hpmicroserver kernel: worker_thread+0x21f/0x3f0 Aug 04 04:33:45 hpmicroserver kernel: ? pwq_unbound_release_workfn+0xc0/0xc0 Aug 04 04:33:45 hpmicroserver kernel: kthread+0x119/0x130 Aug 04 04:33:45 hpmicroserver kernel: ? __kthread_parkme+0xa0/0xa0 Aug 04 04:33:45 hpmicroserver kernel: ret_from_fork+0x22/0x40 This indicates the log force is in progress, and it is waiting for log buffer space to become available to continue writing the in-memory journal to disk. This means that all log buffers are currently under IO, and the journal is waiting for IO completion to occur before it can issue more IO to the journal to make further progress. This typically occurs when there is an IO problem in the layers below XFS. Without seeing the full logs, or knowing any other details about your system, hardware or storage I can't say any more than this. -Dave. -- You are receiving this mail because: You are watching the assignee of the bug.
