[quagga-dev 8958] Re: Strange logs

Stig stig at ubnt.com
Sun Nov 27 20:26:47 GMT 2011


On Sun, Nov 27, 2011 at 2:39 AM, Sami Halabi <sodynet1 at gmail.com> wrote:
> Hi Stug,
> how do I do that?

Well, my home router isn't the best example since it doesn't have
multi-queue nics, but if you look at the nics in /proc/interrupts I
have:

root at v600:/home/vyatta# cat /proc/interrupts
           CPU0       CPU1       CPU2       CPU3
 43:          1          1        619          1   PCI-MSI-edge      eth1
 44:          3     104898          1          2   PCI-MSI-edge      eth2
 45:          0          0          1        576   PCI-MSI-edge      eth3
 46:        575          1          1          1   PCI-MSI-edge      eth4
 47:          7          9      16016          7   PCI-MSI-edge      eth5


You can see the interrupts are spread across the 4 cpu's.  To see the
smp_affinity for each interrupt:

root at v600:/home/vyatta# cat /proc/irq/43/smp_affinity
4
root at v600:/home/vyatta# cat /proc/irq/44/smp_affinity
2
root at v600:/home/vyatta# cat /proc/irq/45/smp_affinity
8
root at v600:/home/vyatta# cat /proc/irq/46/smp_affinity
1
root at v600:/home/vyatta# cat /proc/irq/47/smp_affinity
4

To change the smp affinity you can echo what ever cpu bit mask you
want into the same proc entry.

stig




> Thanks in advance,
> Sami
>
>
> On Sun, Nov 27, 2011 at 1:33 AM, Stig <stig at ubnt.com> wrote:
>>
>> If you have a multi-core system, you could change the interrupt
>> smp-affinity so that some core/thread is not used for handling network
>> interrupts.
>>
>> On Sat, Nov 26, 2011 at 11:35 AM, Sami Halabi <sodynet1 at gmail.com> wrote:
>> > Hi,
>> > Lately i see strange messages on my /var/log, i use FBSD-8.1-R with
>> > Quagga
>> > 0.99.17
>> > Nov 26 19:28:50 bgpServer watchquagga[1394]: zebra state -> unresponsive
>> > :
>> > no response yet to ping sent 10 seconds ago
>> > Nov 26 19:28:57 bgpServer zebra[88251]: SLOW THREAD: task
>> > zebra_client_read
>> > (102aae0) ran for 21313ms (cpu time 1ms)
>> > Nov 26 19:28:57 bgpServer watchquagga[1394]: zebra: slow echo response
>> > finally received after 16.449412 seconds
>> > Nov 26 19:29:12 bgpServer watchquagga[1394]: zebra state -> unresponsive
>> > :
>> > no response yet to ping sent 10 seconds ago
>> > Nov 26 19:29:12 bgpServer watchquagga[1394]: bgpd state -> unresponsive
>> > : no
>> > response yet to ping sent 10 seconds ago
>> > Nov 26 19:29:15 bgpServer zebra[88251]: SLOW THREAD: task
>> > zebra_client_read
>> > (102aae0) ran for 17260ms (cpu time 0ms)
>> > Nov 26 19:29:15 bgpServer watchquagga[1394]: zebra: slow echo response
>> > finally received after 13.707425 seconds
>> > Nov 26 19:29:15 bgpServer bgpd[79850]: SLOW THREAD: task bgp_read
>> > (10675d0)
>> > ran for 17095ms (cpu time 5ms)
>> > Nov 26 19:29:15 bgpServer watchquagga[1394]: bgpd: slow echo response
>> > finally received after 13.700411 seconds
>> > Nov 26 19:29:30 bgpServer watchquagga[1394]: bgpd state -> unresponsive
>> > : no
>> > response yet to ping sent 10 seconds ago
>> > Nov 26 19:29:30 bgpServer watchquagga[1394]: zebra state -> unresponsive
>> > :
>> > no response yet to ping sent 10 seconds ago
>> > Nov 26 19:29:34 bgpServer bgpd[79850]: SLOW THREAD: task bgp_read
>> > (10675d0)
>> > ran for 16281ms (cpu time 0ms)
>> > Nov 26 19:29:34 bgpServer zebra[88251]: SLOW THREAD: task
>> > zebra_client_read
>> > (102aae0) ran for 16276ms (cpu time 0ms)
>> > Nov 26 19:29:34 bgpServer watchquagga[1394]: bgpd: slow echo response
>> > finally received after 13.585467 seconds
>> > Nov 26 19:29:34 bgpServer watchquagga[1394]: zebra: slow echo response
>> > finally received after 13.496174 seconds
>> > Nov 26 19:29:49 bgpServer watchquagga[1394]: bgpd state -> unresponsive
>> > : no
>> > response yet to ping sent 10 seconds ago
>> > Nov 26 19:29:49 bgpServer watchquagga[1394]: zebra state -> unresponsive
>> > :
>> > no response yet to ping sent 10 seconds ago
>> > Nov 26 19:29:54 bgpServer watchquagga[1394]: Forked background command
>> > [pid
>> > 79871]: /usr/local/etc/quagga/service bgpd stop
>> > Nov 26 19:30:00 bgpServer watchquagga[1394]: bgpd state -> down : read
>> > returned EOF
>> > Nov 26 19:30:00 bgpServer watchquagga[1394]: Forked background command
>> > [pid
>> > 79875]: /usr/local/etc/quagga/service zebra restart
>> > Nov 26 19:30:01 bgpServer watchquagga[1394]: Forked background command
>> > [pid
>> > 79879]: /usr/local/etc/quagga/service bgpd start
>> > Nov 26 19:30:01 bgpServer watchquagga[1394]: Phased global restart has
>> > completed.
>> > Nov 26 19:30:01 bgpServer zebra[88251]: SLOW THREAD: task
>> > zebra_client_read
>> > (102aae0) ran for 24850ms (cpu time 0ms)
>> > Nov 26 19:30:01 bgpServer watchquagga[1394]: zebra: slow echo response
>> > finally received after 21.684171 seconds
>> > Nov 26 19:30:01 bgpServer bgpd[79885]: BGPd 0.99.17 starting: vty at 2605,
>> > bgp@<all>:179
>> > Nov 26 19:30:05 bgpServer watchquagga[1394]: bgpd state -> up : connect
>> > succeeded
>> > Nov 26 19:30:06 bgpServer watchquagga[1394]: zebra state -> up : echo
>> > response received after 0.065503 seconds
>> > Nov 26 19:30:21 bgpServer watchquagga[1394]: zebra state -> unresponsive
>> > :
>> > no response yet to ping sent 10 seconds ago
>> > Nov 26 19:30:35 bgpServer zebra[88251]: SLOW THREAD: task
>> > zebra_client_read
>> > (102aae0) ran for 24767ms (cpu time 43ms)
>> > Nov 26 19:30:35 bgpServer watchquagga[1394]: zebra: slow echo response
>> > finally received after 24.512305 seconds
>> > Nov 26 19:30:40 bgpServer watchquagga[1394]: zebra state -> up : echo
>> > response received after 0.004419 seconds
>> >
>> > Lately i pass 2-3GB of traffic daliy maybe its related? moreover I have
>> > 10G
>> > Dual card base 82599-EB card connected to a 10GB Switch.
>> > Why these messages arrive? does that mean i have something wrong?
>> > /etc/sysctl.conf
>> > net.inet.flowtable.enable=0
>> > net.inet.ip.fastforwarding=1
>> > kern.ipc.somaxconn=8192
>> > kern.ipc.shmmax=2147483648
>> > kern.ipc.maxsockets=204800
>> > kern.ipc.maxsockbuf=262144
>> > kern.maxfiles=256000
>> > kern.maxfilesperproc=230400
>> > net.inet.ip.dummynet.pipe_slot_limit=1000
>> > net.inet.ip.dummynet.io_fast=1
>> > #10Gb sysctls
>> > hw.intr_storm_threshold=9000
>> > kern.ipc.nmbclusters=262144
>> > kern.ipc.nmbjumbop=262144
>> > dev.ix.0.rx_processing_limit=4096
>> > dev.ix.1.rx_processing_limit=4096
>> >
>> > Thanks in advance,
>> > --
>> > Sami Halabi
>> > Information Systems Engineer
>> > NMS Projects Expert
>> >
>> > _______________________________________________
>> > Quagga-dev mailing list
>> > Quagga-dev at lists.quagga.net
>> > http://lists.quagga.net/mailman/listinfo/quagga-dev
>> >
>> >
>
>
>
>
> --
> Sami Halabi
> Information Systems Engineer
> NMS Projects Expert
>



More information about the Quagga-dev mailing list