Re: Reciving multiple RAs causes NetworkManager to storm the network with RSes



On Thu, 2019-02-14 at 10:29 -0500, Brian J. Murrell wrote:
On Thu, 2019-02-14 at 15:16 +0100, Thomas Haller via networkmanager-
list wrote:
Hi,

Hi.

From the logs it looks like to me the IPv6 address of the sender is
indeed the instance managed by NetworkManager.

Oh yes, I am sure it's the NM machine that is sending the RSes.

Is the issue hard to reproduce?

Not hard to reproduce but I don't have a "lab" to reproduce it in and
so have to do it on my production network and doing so brings the
router/switch that everything is connected to to it's knees so it's
pretty catastrophic to reproduce.

I'm also concerned with the disk load that disabling the rate
limiting
of journald is going to cause on the NM machine given that it's
suppressing thousands of messages at a time.

But I will give it another go.

Just so that I only have to do this once and as briefly as possible,
what level of debugging do I need to set to get the messages logged
by:

https://cgit.freedesktop.org/NetworkManager/NetworkManager/tree/src/ndisc/nm-ndisc.c?id=2c881b8064e1cb5b227e5a3c61abfe95c6ddd05a#n711


Hi,

Best way is:

- first disable ratelimiting in journald. For that, set 
  RateLimitIntervalSec=0 in /etc/systemd/journald.conf and
  `systemctl restart systemd-journald.service`

- enable level=TRACE in /etc/NetworkManager/NetworkManager.conf
  (level=DEBUG would be enough for these particular loglines,
  but as you go throgh the effort of collecting logs, there is no need 
  to restrict what to collect).

- restart NM, reproduce, and provide logs.


In particular, loglines with the word "solicitation" seem relevant.


Thanks,
Thomas

Attachment: signature.asc
Description: This is a digitally signed message part



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]