OOM in secondary cgroup leading to networking loss

OOM in secondary cgroup leading to networking loss

[Àbéjídé Àyodélé]

Hi friends,

One one of our machines at work, we observed a sequence of events starting
from an OOM in a secondary cgroup which ends up in the bond interface being
down for a period of up to 12 seconds. Below is some piece of dmesg about when
the bond interface went down:

[Wed Jul 25 19:20:45 2018] Call Trace:
[Wed Jul 25 19:20:45 2018]  <IRQ>
[Wed Jul 25 19:20:45 2018]  ? dev_deactivate_queue.constprop.29+0x60/0x60
[Wed Jul 25 19:20:45 2018]  call_timer_fn+0x30/0x120
[Wed Jul 25 19:20:45 2018]  run_timer_softirq+0x3c8/0x420
[Wed Jul 25 19:20:45 2018]  ? timerqueue_add+0x52/0x80
[Wed Jul 25 19:20:45 2018]  ? enqueue_hrtimer+0x37/0x80
[Wed Jul 25 19:20:45 2018]  ? recalibrate_cpu_khz+0x10/0x10
[Wed Jul 25 19:20:45 2018]  __do_softirq+0xde/0x2b3
[Wed Jul 25 19:20:45 2018]  irq_exit+0xae/0xb0
[Wed Jul 25 19:20:45 2018]  smp_apic_timer_interrupt+0x70/0x120
[Wed Jul 25 19:20:45 2018]  apic_timer_interrupt+0x7d/0x90
[Wed Jul 25 19:20:45 2018]  </IRQ>
[Wed Jul 25 19:20:45 2018] RIP: 0010:cpuidle_enter_state+0xa2/0x2e0
[Wed Jul 25 19:20:45 2018] RSP: 0018:ffffffff9c403eb0 EFLAGS: 00000246
ORIG_RAX: ffffffffffffff10
[Wed Jul 25 19:20:45 2018] RAX: ffff9075c0821a40 RBX: 0006cdb159b6000e RCX:
000000000000001f
[Wed Jul 25 19:20:45 2018] RDX: 0006cdb159b6000e RSI: ffed6f2696159a35 RDI:
0000000000000000
[Wed Jul 25 19:20:45 2018] RBP: ffffd3edc100b900 R08: 0000000000000f48 R09:
0000000000000cfe
[Wed Jul 25 19:20:45 2018] R10: ffffffff9c403e90 R11: 0000000000000f12 R12:
0000000000000003
[Wed Jul 25 19:20:45 2018] R13: ffffffff9c4b03d8 R14: 0000000000000000 R15:
0006cdb159783c8e
[Wed Jul 25 19:20:45 2018]  do_idle+0x181/0x1e0
[Wed Jul 25 19:20:45 2018]  cpu_startup_entry+0x19/0x20
[Wed Jul 25 19:20:45 2018]  start_kernel+0x400/0x408
[Wed Jul 25 19:20:45 2018]  secondary_startup_64+0xa5/0xb0
[Wed Jul 25 19:20:45 2018] Code: 63 8e 60 04 00 00 eb 8f 4c 89 f7 c6 05 79 c7
b8 00 01 e8 00 7c fd ff 89 d9 48 89 c2 4c 89 f6 48 c7 c7 f0 38 28 9c e8 c7 a3
b6 ff <0f> 0b eb bd 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 41 57 41 56
[Wed Jul 25 19:20:45 2018] ---[ end trace 2ad2942fe3431402 ]---
[Wed Jul 25 19:20:45 2018] ixgbe 0000:19:00.1 eno2: initiating reset due to tx
timeout
[Wed Jul 25 19:20:45 2018] ixgbe 0000:19:00.1 eno2: Reset adapter
[Wed Jul 25 19:20:48 2018] ixgbe 0000:19:00.0 eno1: initiating reset due to tx
timeout
[Wed Jul 25 19:20:53 2018] ixgbe 0000:19:00.0 eno1: initiating reset due to tx
timeout

We have observed a similar behavior on a 4.15.11 kernel we were running on a
different machine, the current machine these logs are from runs a 4.14.52
kernel. A more detailed dmesg content can be found here:

https://gist.github.com/bjhaid/49a1c58742ef2458984339503290ef9a

I will appreciate any help in figuring out the cause and fix of this issue, or
what the correct mailing list to post this to.

Extra bit of information is the machines are Kubernetes nodes running docker.

Thanks!

Abejide Ayodele
It always seems impossible until it's done. --Nelson Mandela