I'm having high CPU utilization due to interrupts on my 7206vxr-npe-g2. on peak hours interrupt is reaching up to 80%. I've check the cef configuration its all enabled on my interfaces, currently i have 2 FE ports one connected to an ISP and on to our LAN. At peak hours, both interfaces is using around 70M each.
But on non-peak hours where traffic on both interfaces are just around 12M each utilization is still quite high on 30% and all due to interrupts.
I can see incrementing unknown protocol drops on my fe5/0 interface what is causing this drops.
Please help me to analyze this problem. I've attached the logs and my isp graph utlization to give you something to start with.
Router#sh proc cpu | exc 0.00
CPU utilization for five seconds: 53%/53%; one minute: 52%; five minutes: 52%
PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
43 26784 99632 268 0.15% 0.10% 0.08% 0 Per-Second Jobs
141 1660 24689294 0 0.15% 0.12% 0.13% 0 HQF Shaper Backg
>> I can see incrementing unknown protocol drops on my fe5/0 interface what is causing this drops.
this is caused by receiving messages of protocols that are not enabled on the interface including: STP BPDUs or ISIS hello messages
however, they are just a few and also there was a bug where at each show interface this counter was incremented by one (your release should not be affected)
These few unknown protocol drops are not causing this issue.
I've noticed that you are not using the built-in GE ports that are present on the NPE-G2.
Also it would be important to know if you have removed the I/O controller (not necessary with NPE-G1 or NPE-G2) I guess you haven't it.
you could use one GE port instead of a PA FE:
specify the media-type with
set speed to 100 and duplex to full or use auto settings
by using GE ports instead of PA you could see if there is any difference
Hope to help
Interrupt level CPU utilization is generally due to traffic forwarding. If you look at the graph of the 'show proc cpu history' and then watch the traffic levels on the interfaces with something like 'show int | i packets/sec|protocol is', this will help you to establish the trend that your router goes through with regards to the amount of traffic that you are forwarding, and the level of the CPU. I am confident that you will find that as the traffic level increases, the CPU also increases. Once this is confirmed, the primary question that should be answered is "what features are configured that would impact (or slow down) the device's ability to forward traffic?" On a processor like the NPE-G2 all forwarding happens with a single multipurpose processor, and therefore each additional feature that is applied to the traffic path will lower the amount of overall throughput that your router can handle.
If you can establish that the two fast ethernet interfaces are the only place that there is traffic, with the 'show interface' command above, then you will want to look at the features that you have configured on each of those interfaces with something like 'show run int f5/0'. With no features configured on the interfaces, just passing routed traffic, the device will forward more, and the CPU will be lower, than with several features configured. When I talk about features I mean things like ACLs, QOS, NAT, FW, etc.
With regards to the unknown protocol drops, as was mentioned previously, these are packets that are coming in that are on protocols that the device does not have configured on that particular interface. If you sent IPX packets to an interface configured for IP only, you would see these increment, for example. If you look at the overall number of unknown protocol drops, in relation to the total number of packets you will quickly notice that these are not the reason behind your increased CPU utilization.
As far as the onboard Gigabit Ethernet interfaces go, you definitely want to make use of these over other Fast Ethernet interfaces installed in the PA slots, however the primary gain with these interfaces is that they will not be restricted by the bus bandwidth limitation on the chassis. This is not going to lower your overall CPU utilization, assuming that the same features are configured on the Gig interfaces as you currently have configured on the Fast Ethernet interfaces.
Hope that helps get you on the right track.