server lockup

Hi there

I've been getting a few hard locks a week.  I finally got around to getting logs and a monitor picture of a crash and there is this "Bad page map in process khugepaged" that happened twice in the server's last few hours before the crash.  Monitoring tells me it happened around 12:22 today (Aug 17)

I've seen suggestions of doing this:

echo never > /sys/kernel/mm/transparent_hugepage/enabled

But Google did not manage to explain what that does in a way I was able to understand.

Trace snippet:

Aug 17 12:02:27 Tower kernel: BUG: Bad page map in process khugepaged  pte:80000803dad9f035 pmd:38bded067
Aug 17 12:02:27 Tower kernel: addr:000055f1fdd0c000 vm_flags:00100073 anon_vma:ffff880002c604d0 mapping:          (null) index:55f1fdd0c
Aug 17 12:02:27 Tower kernel: file:          (null) fault:          (null) mmap:          (null) readpage:          (null)
Aug 17 12:02:27 Tower kernel: CPU: 5 PID: 323 Comm: khugepaged Tainted: G    B D W       4.14.40-unRAID #1
Aug 17 12:02:27 Tower kernel: Hardware name: Supermicro X10SL7-F/X10SL7-F, BIOS 3.0 04/24/2015
Aug 17 12:02:27 Tower kernel: Call Trace:
Aug 17 12:02:27 Tower kernel: dump_stack+0x5d/0x79
Aug 17 12:02:27 Tower kernel: print_bad_pte+0x20e/0x22b
Aug 17 12:02:27 Tower kernel: _vm_normal_page+0x73/0x8b
Aug 17 12:02:27 Tower kernel: khugepaged+0x80f/0x1610
Aug 17 12:02:27 Tower kernel: ? wait_woken+0x68/0x68
Aug 17 12:02:27 Tower kernel: ? collapse_shmem+0xa41/0xa41
Aug 17 12:02:27 Tower kernel: kthread+0x111/0x119
Aug 17 12:02:27 Tower kernel: ? kthread_create_on_node+0x3a/0x3a
Aug 17 12:02:27 Tower kernel: ? call_usermodehelper_exec_async+0x116/0x11d
Aug 17 12:02:27 Tower kernel: ret_from_fork+0x35/0x40

Can someone make sense of this?

tower-diagnostics-20180817-1203.zip

FCPsyslog_tail.txt