I had a problematice router will always hit cpu utilisation of 70% - 80%. is it normal for a router with series 3600 to encounter such issues and what is the best recommendated usage of a router cpu threshold? is there any official recommendation from cisco?
worry - lim
70% to 80 % is on alarming stage and u really neeed to find out which process is hogging up your CPU.
You can make use of show processes cpu and show processes cpu sorted to have sorted display of process using cpu...
Can you please paste the output of "show processes cpu Command"
Refer link below for further assistance: (Troubleshooting High CPU utilization on Cisco Routers):
PLS RATE if Helps
Guru Prasad R
hi.. sir , thanks for the quick response ... in fact, i am going to implement a higher version router to replace the 3600 series routers. however, i cant do it if i did not have a more concrete recommendation or even a formula from cisco to support for me to introduce to my management whereby a higher version router such as 3700 to 7200 series shall be able to fulfill the needs. moreover, we had 2 routers with the same series been with the organisation for 3 yrs.
therefore, until now there is no link from cisco side to verify that 3600 series router with such high cpu recommended to change to a higher routers. so if there is anyone is able to help me to verify this. i will be more than happier.
Cisco does sometimes publish recommendations about the capacity of different routers, but you often have to dig for them. They will describe either how much bandwidth the router can usually handle and/or it PPS rate.
For the 3600 series, see http://www.cisco.com/en/US/products/hw/routers/ps274/products_tech_note09186a00801e1155.shtml
Take special note about the performance difference between process switching and fast switching. Roughly the 3660 could handle 50 to 60 Mbps in its fast path but only 5 or 6 Mbps in its process path.
As just shown, the capacity of most routers can vary greatly depending on what you're doing with it. This too is sometimes mentioned but often without specifics beyond some notation of using service features.
What can be a killer is when a service feature causes the router to process switch instead of fast switch. This could be part of your problem, but you haven't provided enough detail information.
Also, not using all the routers features to their fullest can impact performance. E.g. not using the best caching mechanism.
Also, because of differences in usage, it's almost impossible to create a concrete recommendation for what size box you need. However, if you have experience with an existing Cisco router running your traffic, then if another Cisco router offer 2x the performance, that would probably be close to being true for you too, everything else being equal.
We need you to post the "show proc cpu" and "show proc mem" at the time the CPU util is in 70-80%
Replacing the router with higher model will not help if you have serious problem in your network. The new router however big it is may inherit the same problem.
As for router recommendation, both 3600 and 3700 are EOS.
HI Charles, [Pls Rate if Helps]
In your output posted, i found no process having CPU utilization greater than 10%
WARNING: Interrupt CPU Utilization is 86%, which is very high (>60%)
CPU interrupts are primarily caused by fast switching of traffic. Interrupts are
also generated any time a character is output from the console or auxiliary ports
of a router.
PLS RATE if HELS
Guru Prasad R
Hi.. guru .. thanks for the advise. we were also suspecting that there might be some unforseen interruption that indirectly causes CPU Utilization hits between 70% - 80%. we will take note of your suggestion but as for interruption due to console. i have to say that it might not be possible cause the cpu alarm hits furiously was like couple of months back. however, as for auxiliary ports.. i think i needed to verify further.
thanks for the reply...
In addition to other comments you can monitor few things with syslog capability available with IOS.
By doing so you could import the logs back to an external server and find out any malfunction or any irratic behavior happening with your box.
Hi.. Guru ,
i had confirmed and verify that we did not plug-in any form of cable or console to auxiliary and console.
What model 3600 is this and can you give us the bit rate of the active interfaces and their type (Ethernet, serial) when you get the type of stats you've attached. Are you using this router for LAN routing?
Hi.. the model that we used is 3640 and the attached file is the bit rate of the active interface as we are using fastethenet for toward input and output is about the followings:
5 mintue input rate 15147000 bits/sec, 2483 packets/sec
5 mintue output rate 4895000 bits/sec, 2224 packets/sec
in additional, we are not using lan routing
If you total bandwidth is only under 10 Mbps, I would expect the box to show a busy CPU (25 to 50%) but not what you're seeing. If you haven't already, take a look at:
and see if any of it helps.
Can you try enabling ip route-cache flow and check based on the flow stats.
you can view the same using show ip cache flow which will give you the transactions happening and the source/destination ip address as well port numbers involved in transactions.
This will give you an idea about any irratic behavior happening with the traffic pattern and traffic being processed by the router..
The port numbers in the result will be displayed in Hex which you can decrypt the transactions and the traffic destined to specific ports...
can you please post running configuration of your router? It could be some complex functions that are run through HW, that are not really required for operation.
We had similar problems on Catalyst switches. Optimizing configuration reduced average switch load from 85 to 35 percent.
Hi..all Cisco Expert.. thank for the help, in the meantime, we will to a higher version of the router to resolve this issues. cause it is confirm that it might be due to a lot of interruption in the route-cache that causes the cpu to shoot to that kinda of extended.
last but not least , i thank all the inputs and suggestion given.. really appreciated as networking was not a standalone job afterall..