Reports of (Debian?) Linux kernel 2.6.32 livelocking when notified of leap second
 

On Sat, 30 Jun 2012, Andy Smith wrote:
> Has anyone been seeing this sort of thing in the last 22-ish hours?
>
> http://serverfault.com/questions/403732/anyone-else-experiencing-high-rates-of-linux-server-crashes-today

No problems with the kernel in _up-to-date_ Debian userspace and kernels, at
least in stable. I will know about oldstable on monday.

Now, I've had a few java apps get a bit crazy, might have been this:
https://lkml.org/lkml/2012/6/30/122

And it happened on latest kernel 3.0, so if it indeed is the hrtimers issue,
it is on more kernels than just 3.4.4/3.5.

--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20120701133612.GD2862@khazad-dum.debian.net">http://lists.debian.org/20120701133612.GD2862@khazad-dum.debian.net

Reports of (Debian?) Linux kernel 2.6.32 livelocking when notified of leap second
 

On Sat, 30 Jun 2012 21:52:21 +0000
Andy Smith <andy@strugglers.net> wrote:

> Hello,
>
> Has anyone been seeing this sort of thing in the last 22-ish hours?
>
> http://serverfault.com/questions/403732/anyone-else-experiencing-high-rates-of-linux-server-crashes-today
>
> I've run adjtimex -p on all of my servers and they're all reporting
> status 17, which as far as I can tell means they've already been sent
> adjtimex()by ntpd and will add a leap second at midnight.
>
> As I understand the reports so far, the livelock would happen at the
> point where ntpd sends the adjtime(), soif they haven't locked
> already then they aren't going to.
>

I didn't find out about this until last night, when I shut down ntpd on
the two Squeeze machines I look after. Obviously that was far too late,
as both machines did the leap-second thing at midnight, and neither had
been affected by the problem.

I've since seen it described as the result of a race hazard, so it
would seem to be a matter of chance whether a particular machine had a
problem.

--
Joe


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20120701221815.27e948ef@jretrading.com">http://lists.debian.org/20120701221815.27e948ef@jretrading.com