Details
-
Bug
-
Resolution: Fixed
-
Major
-
Lustre 2.16.0
-
None
-
3
-
9223372036854775807
Description
On our fat Numa nodes, problem/crash occurs randomly but quite often, and the stack-trace looks always like following :
[246065.156962] LustreError: 789506:0:(class_obd.c:839:obdclass_exit()) obd_memory max: 16059942, leaked: 40 ................ [246065.484593] Call trace: [246065.487180] percpu_counter_add_batch+0x2c/0x188 [246065.491987] 0xffffc0a0cf97869c [246065.495282] rcu_do_batch+0x184/0x450 [246065.499120] rcu_core+0x174/0x3e8 [246065.502593] rcu_core_si+0x18/0x38 [246065.506156] __do_softirq+0x134/0x460 [246065.509988] ____do_softirq+0x18/0x38 [246065.513818] call_on_irq_stack+0x24/0x30 [246065.517914] do_softirq_own_stack+0x24/0x48 [246065.522275] irq_exit_rcu+0xa0/0xf8 [246065.525930] el1_interrupt+0x4c/0xd0 [246065.529674] el1h_64_irq_handler+0x18/0x38 [246065.533947] el1h_64_irq+0x7c/0x80 [246065.537508] cpuidle_enter_state+0xd4/0x828 [246065.541870] cpuidle_enter+0x40/0x70 [246065.545615] cpuidle_idle_call+0x150/0x208 [246065.549892] do_idle+0xa8/0x128 [246065.553188] cpu_startup_entry+0x3c/0x50 [246065.557282] secondary_start_kernel+0xf0/0x150 [246065.561911] __secondary_switched+0xb8/0xc0 .................. [246065.600422] Kernel panic - not syncing: Oops: Fatal exception in interrupt [246065.607542] SMP: stopping secondary CPUs
interesting is the fact there seems to always be this same preceeding message claiming that some OBD memory has been leaked, and apparently always as a multiple of 40 bytes...