[f-nsp] Odd CPU spikes on Jetcore BI4K, IP level
Bardo Cornelissen
b.cornelissen at caveo.nl
Wed Oct 15 06:57:33 EDT 2008
Hi Jeroen,
I'm going to watch that ip cache, I see it's little above 100k now so I
assume it could grow above de default 140k and that would explain and we
should be able to solve this. Many thanks already!
I'm not sure how to look at this cam-partition, can you tell if anything
looks like an issue or exhausted here:
#sh cam-part det
==== SLOT 1 CAM PARTITION ====
IGC: 0 (0x00)
Number of CAM devices per IGC: 1
Number of hw entries per CAM: 0x08000
Total size of CAM = 2Mbits
complete CAM index range per IGC:
(sw) 1 - 49151 (1 - 0x0bfff), total entries: 49151 (0x0bfff)
(hw) 0 - 32767 (0 - 0x07fff), total entries: 32768 (0x08000)
Percentage of CAM hardware entries for each partition:
Layer3 = 16384 (1Mbits) (50%)
Level3 = 1023 (0.062438Mbits) (3.121948%)
Level2 = 1024 (0.0625Mbits) (3.125%)
Layer2 = 8192 (0.5Mbits) (25%)
Layer4 = 8192 (0.5Mbits) (25%)
Layer 3 sw index range:
L3 L3 1 - 2047 (0x00001 - 0x007ff), free 2021 (0x007e5)
L3 L2 2048 - 4095 (0x00800 - 0x00fff), free 1803 (0x0070b)
L3 4096 - 32767 (0x01000 - 0x07fff), free 23357 (0x05b3d)
layer 3 hw index range (inversely mapped):
16383 - 0 (0x03fff - 0x00000)
L2 index range:
(sw) 32768 - 40959 (0x08000 - 0x09fff), free 7806 (0x01e7e)
(hw) 16384 - 24575 (0x04000 - 0x05fff)
L4 pool 3 index range:
(sw) 40960 - 43519 (0x0a000 - 0x0a9ff), free 2560 (0x00a00)
(hw) 24576 - 27135 (0x06000 - 0x069ff)
L4 pool 2 index range:
(sw) 43520 - 46079 (0x0aa00 - 0x0b3ff), free 2560 (0x00a00)
(hw) 27136 - 29695 (0x06a00 - 0x073ff)
L4 pool 1 index range:
(sw) 46080 - 48639 (0x0b400 - 0x0bdff), free 2560 (0x00a00)
(hw) 29696 - 32255 (0x07400 - 0x07dff)
L4 pool 0 index range:
(sw) 48640 - 49151 (0x0be00 - 0x0bfff), free 504 (0x001f8)
(hw) 32256 - 32767 (0x07e00 - 0x07fff)
IGC: 1 (0x01)
Number of CAM devices per IGC: 1
Number of hw entries per CAM: 0x08000
Total size of CAM = 2Mbits
complete CAM index range per IGC:
(sw) 1 - 49151 (1 - 0x0bfff), total entries: 49151 (0x0bfff)
(hw) 0 - 32767 (0 - 0x07fff), total entries: 32768 (0x08000)
Percentage of CAM hardware entries for each partition:
Layer3 = 16384 (1Mbits) (50%)
Level3 = 1023 (0.062438Mbits) (3.121948%)
Level2 = 1024 (0.0625Mbits) (3.125%)
Layer2 = 8192 (0.5Mbits) (25%)
Layer4 = 8192 (0.5Mbits) (25%)
Layer 3 sw index range:
L3 L3 1 - 2047 (0x00001 - 0x007ff), free 1951 (0x0079f)
L3 L2 2048 - 4095 (0x00800 - 0x00fff), free 1100 (0x0044c)
L3 4096 - 32767 (0x01000 - 0x07fff), free 13870 (0x0362e)
layer 3 hw index range (inversely mapped):
16383 - 0 (0x03fff - 0x00000)
L2 index range:
(sw) 32768 - 40959 (0x08000 - 0x09fff), free 7878 (0x01ec6)
(hw) 16384 - 24575 (0x04000 - 0x05fff)
L4 pool 3 index range:
(sw) 40960 - 43519 (0x0a000 - 0x0a9ff), free 2560 (0x00a00)
(hw) 24576 - 27135 (0x06000 - 0x069ff)
L4 pool 2 index range:
(sw) 43520 - 46079 (0x0aa00 - 0x0b3ff), free 2560 (0x00a00)
(hw) 27136 - 29695 (0x06a00 - 0x073ff)
L4 pool 1 index range:
(sw) 46080 - 48639 (0x0b400 - 0x0bdff), free 2560 (0x00a00)
(hw) 29696 - 32255 (0x07400 - 0x07dff)
L4 pool 0 index range:
(sw) 48640 - 49151 (0x0be00 - 0x0bfff), free 504 (0x001f8)
(hw) 32256 - 32767 (0x07e00 - 0x07fff)
Kind regards,
Bardo Cornelissen.
Caveo Internet BV
> -----Oorspronkelijk bericht-----
> Van: Jeroen Wunnink [mailto:jeroen at easyhosting.nl]
> Verzonden: woensdag 15 oktober 2008 12:32
> Aan: Bardo Cornelissen
> CC: foundry-nsp at puck.nether.net
> Onderwerp: Re: [f-nsp] Odd CPU spikes on Jetcore BI4K, IP level
>
> Hi Bardo,
>
> Except for the CAM usage, also look on how saturated your IP cache is.
> 'sh ip cache' and look at the top at the total number of cache entries.
>
> The default max is 140000, if it exceeds this maximum it'll start
> throwing packets over the CPU, if it's at or near it's max, increase it
> with 'system-max ip cache 400000' (reload required)
>
>
> Bardo Cornelissen wrote:
> > Hi folks,
> >
> > I'm facing an odd issue causing high CPU load on a Jetcore BI4K, with
> for
> > what seems to be caused in a ip transit uplink.
> >
> > We are running BGP on multiple jetcore BigIrons. Now there are 3
> border
> > routers which each have an ip transit uplink. After some time (a day
> or
> > sometimes more like a week) we see a high CPU load and it doesn't
> stop until
> > we shutdown the BGP session with one of these carriers. So far we
> have seen
> > 2 crashes which may have been caused by this.
> >
> > So we moved that carrier to a different router (same model), replaced
> the
> > used SFP, upgraded the software on the router but nothing solves the
> > problem. Also our carrier relocated the port to a complete different
> switch
> > and router, but still no luck. My carrier double checked everything
> over and
> > over and with their reputation it's most unlikely they would be
> failing
> > here.
> >
> > Now the load is on IP level when issuing a 'sh proc cpu' command and
> NOT on
> > the BGP level.
> >
> > I hope someone somehow has ever seen a situation alike and found the
> issue,
> > or maybe has some recommendations on how to debug this. Maybe someone
> knows
> > any additional configuration that may help me out here.
> >
> > The concerned config is:
> > !
> > vlan 20 name TRANSIT-BGP by port
> > untagged ethe 1/1
> > router-interface ve 20
> > !
> > interface ethernet 1/1
> > route-only
> > no spanning-tree
> > sflow forwarding
> > !
> > interface ve 20
> > ip address ???.???.???.??? 255.255.255.252
> > no ip redirect
> > ip arp-age 120
> > !
> >
> > Thanks in advance for any help and efforts!
> >
> > Kind regards,
> >
> >
> > Bardo Cornelissen.
> > Caveo Internet BV
> >
> >
> > _______________________________________________
> > foundry-nsp mailing list
> > foundry-nsp at puck.nether.net
> > http://puck.nether.net/mailman/listinfo/foundry-nsp
> >
>
>
> --
>
> Met vriendelijke groet,
>
> Jeroen Wunnink,
> EasyHosting B.V. Systeembeheerder
> systeembeheer at easyhosting.nl
>
> telefoon:+31 (035) 6285455 Postbus 48
> fax: +31 (035) 6838242 3755 ZG Eemnes
>
> http://www.easyhosting.nl
> http://www.easycolocate.nl
>
>
More information about the foundry-nsp
mailing list