[c-nsp] CPU utilization for handling interrupt
Hi, I'm seeing the following output for show interface gig x/y switching ROUTER#sh inter gig x/y switching GigabitEthernet x/y Throttle count 0 Drops RP 0 SP 0 SPD Flushes Fast 0 SSE 0 SPD Aggress Fast 0 SPD Priority Inputs 46670 Drops 0 Protocol Path Pkts In Chars In Pkts Out Chars Out Other Process 1078379 69124236 1 96 Cache misses 0 Fast 0 0 0 0 Auton/SSE 0 0 0 0 IP Process 3594269215 341714357335 162336237 18154131440 Cache misses 0 Fast 395280896627 35724688800466 406469605169 44781968153216 Auton/SSE 1220084333084 240117721335247 1899837692532 1757256129434539 ARP Process 28158607 1689516436 31556627 3029436192 Cache misses 0 Fast 0 0 0 0 Auton/SSE 0 0 0 0 The IP Process and IP Fast are accumulating . The config. of the interface is as follows: -- interface GigabitEthernet x/y ip address x.x.x.x x.x.x.x ip verify unicast source reachable-via rx allow-default no ip redirects no ip proxy-arp ip tcp adjust-mss 1400 speed nonegotiate no cdp enable service-policy input POLICY service-policy output POLICY end Regard, Waseem ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
On 05/10/11 09:31, Waseem wrote: Hi, I'm seeing the following output for show interface gig x/y switching What platform? What IOS version? And what is your question? ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
On 05/10/11 12:05, Waseem wrote: 7600+RSP720-3C-GE 12.2(33)SRB2 why I'm seeing 10% CPU utilization by interrupt handling? Try using a SPAN of the CPU to see what traffic is hitting the CPU; this is by far the quickest way to find the cause. ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
It is a regular internet traffic to port 80, from our customers, which should be CEF switched. From: Phil Mayers p.may...@imperial.ac.uk To: Waseem waseem_alir...@yahoo.com Cc: cisco-nsp@puck.nether.net cisco-nsp@puck.nether.net Sent: Wednesday, October 5, 2011 2:07 PM Subject: Re: [c-nsp] CPU utilization for handling interrupt On 05/10/11 12:05, Waseem wrote: 7600+RSP720-3C-GE 12.2(33)SRB2 why I'm seeing 10% CPU utilization by interrupt handling? Try using a SPAN of the CPU to see what traffic is hitting the CPU; this is by far the quickest way to find the cause. ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
On 05/10/11 12:15, Waseem wrote: It is a regular internet traffic to port 80, from our customers, which should be CEF switched. Port 80 traffic to where? Can you show some? There must be something wrong with the traffic or your config for the 7600 to be process switching it. You need to give more details instead of giving the minimum information in each reply you make. In your original post, you gave a small config snippet; can you describe the topology in more detail? Rather than replacing IP addresses with x.x.x.x can you replace them with corresponding private IPs? Or show the real config? I am assuming the traffic is ingressing on the gig interface whose config you listed; what is the egress interface? What is the config for the service-policy you list? What does: sh int GiX/X sh ip int GiX/X sh tcam int GiX/X acl in ip ...say? ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
7600+RSP720-3C-GE 12.2(33)SRB2 why I'm seeing 10% CPU utilization by interrupt handling? From: Phil Mayers p.may...@imperial.ac.uk To: cisco-nsp@puck.nether.net Sent: Wednesday, October 5, 2011 1:41 PM Subject: Re: [c-nsp] CPU utilization for handling interrupt On 05/10/11 09:31, Waseem wrote: Hi, I'm seeing the following output for show interface gig x/y switching What platform? What IOS version? And what is your question? ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
On 05/10/2011 12:15, Waseem wrote: It is a regular internet traffic to port 80, from our customers, which should be CEF switched. Sounds like your router is punting all traffic. Are you seeing the following errors in your logs? %CFIB-SP-7-CFIB_EXCEPTION : FIB TCAM exception, Some entries will be software switched If this is the case, you need to drop the number of routes that the box is handling, and then reboot the system. Once the FIB limits are exceeded on this platform, rebooting is the only way to revert to hardware forwarding. Nick ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
TCP Adjust-mss causes the 6k to punt the SYN to SW. I'm not sure if this will be process switched or CEF switched (interrupt), but I don't see a reason why we couldn't do it in software CEF. -Pete On Wed, Oct 5, 2011 at 8:20 AM, Nick Hilliard n...@foobar.org wrote: On 05/10/2011 12:15, Waseem wrote: It is a regular internet traffic to port 80, from our customers, which should be CEF switched. Sounds like your router is punting all traffic. Are you seeing the following errors in your logs? %CFIB-SP-7-CFIB_EXCEPTION : FIB TCAM exception, Some entries will be software switched If this is the case, you need to drop the number of routes that the box is handling, and then reboot the system. Once the FIB limits are exceeded on this platform, rebooting is the only way to revert to hardware forwarding. Nick ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
I'm not receiving that log, I have nearly 600Mbps on this link, nearly 3 - 6 Mbps is being process switched from this link only, I tried to disable it, the CPU due to interrupt got 0%. please check the following packets. -- interface Gi1/9, routine process_rx_packet_inline dbus info: src_vlan 0x406(1030), src_indx 0x8(8), len 0x42(66) bpdu 0, index_dir 0, flood 0, dont_lrn 0, dest_indx 0x380(896) B8020401 0406 0008 4200 00060530 0E40 0380 destmac 00.1E.13.E4.A2.00, srcmac 00.25.9E.20.7A.D0, protocol 0800 protocol ip: version 0x04, hlen 0x05, tos 0x00, totlen 48, identifier 6637 df 1, mf 0, fo 0, ttl 126, src 109.127.86.37, dst 209.85.145.105 tcp src 63915, dst 80, seq 2253251144, ack 0, win 16384 off 7 checksum 0xBA29 syn --- interface Gi1/9, routine naboo_fastsend dbus info: src_vlan 0x406(1030), src_indx 0x380(896), len 0x46(70) bpdu 0, index_dir 0, flood 0, dont_lrn 0, dest_indx 0x380(896) 0002 04062800 0380 4600 00060560 0040 0380 destmac 00.25.9E.20.7A.D0, srcmac 00.1E.13.E4.A2.00, protocol 0800 layer 3 data: 4534 5DCA4000 3706F051 57F8D9C0 6D7F5670 005052A5 845F754B EC784102 8012 00E2 02040514 01030304 0402 001E688A 0413 0340 interface Gi1/9, routine process_rx_packet_inline dbus info: src_vlan 0x406(1030), src_indx 0x8(8), len 0x42(66) bpdu 0, index_dir 0, flood 0, dont_lrn 0, dest_indx 0x380(896) E0020401 0406 0008 4200 00060520 0E40 0380 destmac 00.1E.13.E4.A2.00, srcmac 00.25.9E.20.7A.D0, protocol 0800 protocol ip: version 0x04, hlen 0x05, tos 0x00, totlen 48, identifier 7783 df 1, mf 0, fo 0, ttl 126, src 109.127.86.8, dst 95.211.87.169 tcp src 29827, dst 80, seq 2269663441, ack 0, win 8192 off 7 checksum 0x9B2E syn - interface Gi1/9, routine process_rx_packet_inline dbus info: src_vlan 0x406(1030), src_indx 0x8(8), len 0x4E(78) bpdu 0, index_dir 0, flood 0, dont_lrn 0, dest_indx 0x380(896) 10020401 0406 0008 4E00 00060550 0E40 0380 destmac 00.1E.13.E4.A2.00, srcmac 00.25.9E.20.7A.D0, protocol 0800 protocol ip: version 0x04, hlen 0x05, tos 0x00, totlen 60, identifier 28750 df 1, mf 0, fo 0, ttl 126, src 109.127.86.29, dst 207.66.182.20 tcp src 58557, dst 80, seq 2150691164, ack 0, win 8192 off 10 checksum 0xD911 syn -- those are captured from dumping the CPU. do you find anything that make them need special handling? regards, Waseem From: Nick Hilliard n...@foobar.org To: Waseem waseem_alir...@yahoo.com Cc: NSP cisco-nsp@puck.nether.net Sent: Wednesday, October 5, 2011 3:20 PM Subject: Re: [c-nsp] CPU utilization for handling interrupt On 05/10/2011 12:15, Waseem wrote: It is a regular internet traffic to port 80, from our customers, which should be CEF switched. Sounds like your router is punting all traffic. Are you seeing the following errors in your logs? %CFIB-SP-7-CFIB_EXCEPTION : FIB TCAM exception, Some entries will be software switched If this is the case, you need to drop the number of routes that the box is handling, and then reboot the system. Once the FIB limits are exceeded on this platform, rebooting is the only way to revert to hardware forwarding. Nick ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
On 05/10/11 14:15, Pete Lumbis wrote: TCP Adjust-mss causes the 6k to punt the SYN to SW. I'm not sure if this will be process switched or CEF switched (interrupt), but I don't see a reason why we couldn't do it in software CEF. Ah, well spotted; I didn't see that. FWIW I have used adjust-mss on our default route to work around temporary MTU problems; it performed quite well, but I'm not sure what traffic rate the OP is facing. ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] CPU utilization for handling interrupt
Hi, TCP adjust-mss is the key, you were right. Thanks Waseem From: Pete Lumbis alum...@gmail.com To: Nick Hilliard n...@foobar.org Cc: Waseem waseem_alir...@yahoo.com; NSP cisco-nsp@puck.nether.net Sent: Wednesday, October 5, 2011 4:15 PM Subject: Re: [c-nsp] CPU utilization for handling interrupt TCP Adjust-mss causes the 6k to punt the SYN to SW. I'm not sure if this will be process switched or CEF switched (interrupt), but I don't see a reason why we couldn't do it in software CEF. -Pete On Wed, Oct 5, 2011 at 8:20 AM, Nick Hilliard n...@foobar.org wrote: On 05/10/2011 12:15, Waseem wrote: It is a regular internet traffic to port 80, from our customers, which should be CEF switched. Sounds like your router is punting all traffic. Are you seeing the following errors in your logs? %CFIB-SP-7-CFIB_EXCEPTION : FIB TCAM exception, Some entries will be software switched If this is the case, you need to drop the number of routes that the box is handling, and then reboot the system. Once the FIB limits are exceeded on this platform, rebooting is the only way to revert to hardware forwarding. Nick ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
