[j-nsp] Juniper High CPU Process Kqread

Richard A Steenbergen ras at e-gerbil.net
Fri Aug 15 13:02:50 EDT 2008


On Fri, Aug 15, 2008 at 08:59:17AM -0700, Kevin Oberman wrote:
> > PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU COMMAND
> >   2540 root       2   0   223M   220M kqread 268:48 44.34% 44.34% rpd
> 
> While I can't supply any answer, the thing that is eating up the CPU is
> rpd, the main routing daemon. It is reading the kernel queue if it is in
> the kqread state.
> 
> To see the route processor using 44% of the CPU is rather high. Even on
> our core routers I generally  only see 3-5% CPU use by rpd. It runs much
> higher at time when there is routing instability.

Start with the basics:

set task accounting on
show task accounting

And look for the thread with the really high cpu times:

BGP.0.0.0.0+179           18375163 1w5d 23:16:57.823    48:52.877        0.142

If you think there is some kind of persistent oscillation, you can monitor 
the routing updates (from a shell) with:

rtsockmon -t rpd

Could be a bug too (I have a 100% cpu in rpd case that we're still trying 
to diagnose), who knows. :)

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)


More information about the juniper-nsp mailing list