High % of process switched packets on 7206-VXR.

jkeeffe · ‎07-28-2008

What could cause a high percentage of packets being processed switched on a 7206-VXR NPE-400 router? This is our internet router with a DS3 link to our provider. We do receive the entire BGP table, but I think we have enough memory to accommodate that.

Here is a portion of 'sh ver' and 'sh bgp ipv4 un summary':

ROC-7206D-Qwest uptime is 2 years, 15 weeks, 7 hours, 46 minutes

System returned to ROM by power-on

System restarted at 00:25:39 pdt Sat Apr 15 2006

System image file is "disk0:c7200-ik9s-mz.123-11.T.bin"

Cisco 7206VXR (NPE400) processor (revision A) with 491520K/65535K bytes of memory.

Processor board ID 31648384

R7000 CPU at 350MHz, Implementation 39, Rev 3.3, 256KB L2 Cache

6 slot VXR midplane, Version 2.7

Last reset from power-on

4 FastEthernet interfaces

1 Serial interface

125K bytes of NVRAM.

125952K bytes of ATA PCMCIA card at slot 0 (Sector size 512 bytes).

8192K bytes of Flash internal SIMM (Sector size 256K).

Configuration register is 0x2102

ROC-7206D-Qwest#sh bgp ipv4 un summary

BGP router identifier 192.254.3.37, local AS number 32974

BGP table version is 230956639, main routing table version 230956639

256489 network entries using 30009213 bytes of memory

277292 path entries using 13310016 bytes of memory

45981/43476 BGP path/bestpath attribute entries using 5333796 bytes of memory

39583 BGP AS-PATH entries using 1013230 bytes of memory

621 BGP community entries using 42538 bytes of memory

16 BGP extended community entries using 2372 bytes of memory

0 BGP route-map cache entries using 0 bytes of memory

18143 BGP filter-list cache entries using 217716 bytes of memory

BGP using 49928881 total bytes of memory

20999 received paths for inbound soft reconfiguration

BGP activity 9191965/8935476 prefixes, 60598774/60321482 paths, scan interval 60 secs

Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd

6x.xxx.4.xx 4 209 73160381 2405466 230956639 0 0 5d21h 256287

19x.xxx.x.xx 4 32974 24385206 65101478 230956639 0 0 6w3d 3

I've also attached a portion of the config and the output of 'show interfaces switching.

Any idea what may be causing such a high percentage of process switched packets?

Giuseppe Larosa · ‎07-28-2008

Hello Jim,

first of all check the CEF status of physical interfaces with sh cef interface.

I would add ip route-cache cef under all interfaces.

A last thought is about memory issue

sh proc mem | inc Free to check the amount of Free memory.

You receive a BGP Internet full table of 256287 prefixes. BGP table takes 49928881 total bytes of memory.

For each prefix in order to be CEF switched the router needs to allocate a memory block that contains all the information needed to send out packets (outgoing interface, outgoing MAC address if applicable) so the memory used by CEF is greater then the one used by BGP.

So a possible reason could be this: not enough memory to CEF switch all the prefixes

Packets for prefixes not stored in the CEF table can only be process switched.

Hope to help

Giuseppe

jkeeffe · ‎07-28-2008

CEF is enabled on all interfaces, and it looks to be enough memory for CEF.

Here is the output of those commands:

ROC-7206D-Qwest#sh proc mem | include Free

Processor Pool Total: 432205280 Used: 224391136 Free: 207814144

I/O Pool Total: 33554432 Used: 4385800 Free: 29168632

PID TTY Allocated Freed Holding Getbufs Retbufs Process

ROC-7206D-Qwest#sh ip cef summary

IP CEF with switching (Table Version 481811), flags=0x0

256368 routes, 0 reresolve, 0 unresolved (0 old, 0 new), peak 255842

381775 instant recursive resolutions, 99998 used background process

256368 leaves, 13301 nodes, 52791980 bytes, 481822 inserts, 225454 invalidations

1 load sharing elements, 376 bytes, 1 references

universal per-destination load sharing algorithm, id 5E8B55F6

4(1) CEF resets, 0 revisions of existing leaves

Resolution Timer: Exponential (currently 1s, peak 1s)

0 in-place/0 aborted modifications

refcounts: 4385807 leaf, 3405294 node

Table epoch: 0 (256368 entries at this epoch)

ROC-7206D-Qwest#sh ip int

Serial1/0 is up, line protocol is up

!

IP fast switching is enabled

IP fast switching on the same interface is enabled

IP Flow switching is enabled

IP CEF 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 disabled

IP route-cache flags are Fast, Flow cache, CEF, Subint Flow

!

FastEthernet2/0 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 Flow switching turbo vector

IP Flow CEF switching turbo vector

IP multicast fast switching is enabled

IP multicast distributed fast switching is disabled

IP route-cache flags are Fast, Flow cache, CEF, Subint Flow

!

FastEthernet2/1 is up, line protocol is up

!

Address determined by non-volatile memory

MTU is 1500 bytes

!

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 Flow switching turbo vector

IP Flow CEF switching turbo vector

IP multicast fast switching is enabled

IP multicast distributed fast switching is disabled

IP route-cache flags are Fast, Flow cache, CEF, Subint Flow

!

BGP Policy Mapping is disabled

Giuseppe Larosa · ‎07-28-2008

Hello Jim,

this is good news.

Packets with IP options like router alert and so on have to examined by the main cpu and so are process-switched even if a CEF entry exists for them.

try sh cef to see CEF punt counters

Hope to help

Giuseppe

Giuseppe Larosa · ‎07-29-2008

Hello Jim,

I would like to add an example of legitimate packets with IP options set seen in your network:

H.323 endpoints in setting up video streams use IP packets with options we have seen discarded by a FWSM.

Hope to help

Giuseppe