[c-nsp] high cpu usage due to interrupts
Rodney Dunn
rodunn at cisco.com
Tue Feb 13 17:06:31 EST 2007
> But only one interface is shown!
>
> # show ip mds interface
> FastEthernet3/0/0 is up, line protocol is up
>
> With "show ip interfaces" all are shown as
> "IP multicast distributed fast switching is enabled" [2]
I don't know why that is.
Make sure ip mroute-cache distributed is on everywhere.
>
> But other show command shows all slots as being supported supported:
>
> # show ip mds stats linecard
>
> Slot Status IPC(seq/max) Q(high/route) Reloads
> 1 active 1552/1652 0/0 2
> 2 active 1552/1652 0/0 2
> 3 active 1552/1652 0/0 2
>
> # show ip mds stats switching
>
> Slot Total Switched Drops RPF Punts Failures
> (switch/clone)
> 1 0 0 0 0 0 0/0
> 2 0 0 0 0 0 0/0
> 3 793372 0 69694 0 723678 0/0
>
>
> Next, I will show interface stats while forwarding only unicast traffic
>
> EB-LasPalmas1#show interfaces fastethernet2/0/0 stat
> FastEthernet2/0/0
> Switching path Pkts In Chars In Pkts Out Chars Out
> Processor 42212 4263733 600462 794226267
> Route cache 1164 1657027 362382562 1310489808
> Distributed cache 348137292 4278774324 1742274 579066040
> Total 348180668 4284695084 364725299 2683782169
>
> [... 20 seconds interval between consecutive "show" ...]
>
> EB-LasPalmas1#show interfaces fastethernet2/0/0 stat
> FastEthernet2/0/0
> Switching path Pkts In Chars In Pkts Out Chars Out
> Processor 42220 4264288 600469 794226717
> Route cache 1164 1657027 362496468 1398057522
> Distributed cache 348245714 43293681 1742275 579066287
> Total 348289098 49214996 364839214 2771350654
Yeah..notice the numbers. You are dCEF switching almost ALL traffic coming
in the FE2/0/0 but are punting most of the traffic going out the interface
to Route Cache which means the RP has to switch it. I'd have to see
the config and see if you have features on causing the punts.
And chekc just" sh cef linecard
>
> And the result is:
>
> Pkts In: almost all through the distributed cache
>
> (348245714-348137292)/20 secs = 5421 pkts/sec
>
> Pkts Out: almost all through the route cache:
>
> (362496468-362382562)/20 secs = 5695 pkts/sec
Yep..that's causing the high CPU on the RSP.
>
> witch matches the interface statistics:
>
> 5 minute input rate 25178000 bits/sec, 5680 packets/sec
> 5 minute output rate 37065000 bits/sec, 5815 packets/sec
Those should be the same regardless of the switching path taken.
>
> So at least for unicast seems that is working, I assume that the output
> path cannot be forwarded with the VIPs...
No, that's not right. In a well setup dCEF 75xx 99% of all packets would
be dCEF switched that are transit through the box.
>
> So I don't understand why iterfaces are not shown in "show ip mds" commands,
> IMHO seems contradictory with the output of "show ip interfaces"... [2]
Yes I agree. Check 'sh cef line' and also 'sh diag' for both slots to
see if they have the same amount of memory in them.
>
> Tomorrow I will try to use only VIP enabled interfaces for multicast.
>
> Following the instructions of
>
> http://www.cisco.com/en/US/products/sw/iosswrel/ps1835/products_configuration_guide_chapter09186a00800ca800.html
>
> I have verified that everything is ok, but I'm using ip route-cache flow instead
> of distributed (in order to enable netflow information on that interface)
That doesn't change the switching path. That just enables netflow. It's
confusing.
If you can't do "ip cef distributed" globally then you may be doing
distributed fast switching. I think you enable that via:
ip route-cache distributed under all the interfaces.
>
> I will try that too
>
> Thanks so much for your comments, other than talking about obsolete hardware
> I think this can be applied to Cisco in general, not CSR-1 of course ;-)
>
> Again Thanks
>
> regards
>
> Ulisses
>
> [1]
>
> egress card:
>
> Slot 2:
> VIP2 controller, HW rev 2.14, board revision B0
> Serial number: 25378622 Part number: 73-1684-05
> PA Bay 0 Information:
> Fast-Ethernet PA, 1 ports, 100BaseTX-ISL
> Serial number: 26248560 Part number: 73-1688-05
>
> ingress card:
>
> Slot 3:
> VIP2 controller, HW rev 2.14, board revision B0
> Serial number: 25378613 Part number: 73-1684-05
> PA Bay 0 Information:
> Fast-Ethernet PA, 1 ports, 100BaseTX-ISL
> Serial number: 25234923 Part number: 73-1688-05
>
>
> [2]
>
> > > > > FastEthernet2/0/0 is up, line protocol is up
> > > > > IP fast switching is enabled
> > > > > IP fast switching on the same interface is disabled
> > > > > IP Flow switching is enabled
> > > > > IP CEF switching is enabled
> > > > > IP Distributed switching is enabled
> > > > > IP Flow switching turbo vector
> > > > > IP Flow CEF switching turbo vector
> > > > > IP multicast fast switching is enabled
> > > > > IP multicast distributed fast switching is enabled
> > > > > FastEthernet3/0/0 is up, line protocol is up
> > > > > IP fast switching is enabled
> > > > > IP fast switching on the same interface is disabled
> > > > > IP Flow switching is enabled
> > > > > IP CEF switching is enabled
> > > > > IP Distributed switching is enabled
> > > > > IP Flow switching turbo vector
> > > > > IP Flow CEF switching turbo vector
> > > > > IP multicast fast switching is enabled
> > > > > IP multicast distributed fast switching is enabled
>
> _______________________________ __ ___________________________________
> /_/
> Network Operation Center __ __ Email: noc at rediris.es
> RedIRIS/Red.es /_/ RedIRIS /_/
> Edificio Bronce __ Tel: + 34 912127625
> Pza. Manuel Gómez Moreno, s/n /_/ Urgencias: + 34 629148201
> E-28020 Madrid Fax: + 34 915568864
>
> _________________ Spanish Academic & Research Network _______________________
More information about the cisco-nsp
mailing list