high cpu due to STP

[spacut]

I have a problems with 5500 - it keeps high CPU util:

sh proc cpu

CPU utilization for five seconds: 61.36%

one minute: 60.16%

five minutes: 60.04%

PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process

--- ----------- ---------- -------- ------- ------- ------- --- ---------------

(...)

20 11003174 1066284 1463000 13.38% 13.92% 13.98% -2 SptBpduRx

21 29144774 296791 1196000 41.98% 40.84% 40.96% -2 SptBpduTx

Moreover I can see:

sh spantree debug bpdu

Reception of BPDUs is currently enabled.

BPDU received/processed/dropped: 3756864 / 3756864 / 0

BPDU dropped from Invalid Source Port: 3180768

How can I detect on which port I get BPDUs which are not supposed to be there?

Take into account the CPU util - I can't type any resource-exhaustive commands.

[deilert]

at thie time your problem is happening try issuing a sh top , this will show you the utilization over a 30 sec span for the 10 busiest sw ports .

[spacut]

Ok, How to interprete these data?

5/9 is atrunk to the second switch, 5/5, 5/8, 5/7 are the trunks to the routers.

sh top

Start Time: Sep 18 2003 16:29:03

End Time: Sep 18 2003 16:29:35

PortType: all

Metric: util

Port Band- Uti Bytes Pkts Bcst Mcst Error Over

width % (Tx + Rx) (Tx + Rx) (Tx + Rx) (Tx + Rx) (Rx) flow

----- ----- --- -------------------- ---------- ---------- ---------- ----- ----

5/9 1000 42 3068567653 5835879 85 4112 0 0

5/5 1000 18 1308864430 2460062 86 4066 0 0

5/8 1000 13 994568430 1878640 85 4074 0 0

5/7 1000 9 705974104 1303358 85 4045 0 0

5/4 1000 1 76690878 148819 55 130 0 0

5/1 1000 0 49376600 77163 55 1037 0 0

5/6 1000 0 38580770 48764 85 4044 0 0

2/10 a-100 0 2031744 10679 1 21 0 0

2/6 100 0 1232518 11840 79 2673 0 0

2/2 a-100 0 1199685 11667 79 2675 0 0

2/3 a-100 0 1190782 11555 79 2674 0 0

2/5 100 0 1190246 11548 79 2674 0 0

2/21 a-100 0 790996 5415 0 20 0 0

2/11 a-10 0 26473 173 0 57 0 0

3/5 a-10 0 2990 33 0 19 0 0

2/12 a-10 0 1616 21 0 18 0 0

2/13 a-10 0 1616 21 0 18 0 0

[deilert]

The broadcasr & multicast traffic over 5/9 is double than the other ports , What does the remote end look like ?

[spacut]

The second side look like this (port 1/2 connects to 5/9 on this second, overloaded switch):

sh top

Start Time: Sep 18 2003 18:51:30

End Time: Sep 18 2003 18:52:00

PortType: all

Metric: util

Port Band- Uti Bytes Pkts Bcst Mcst Error Over

width % (Tx + Rx) (Tx + Rx) (Tx + Rx) (Tx + Rx) (Rx) flow

----- ----- --- -------------------- ---------- ---------- ---------- ----- ----

1/2 1000 35 2586591407 5190173 44 4845 0 0

3/46 10 32 23121650 75761 0 56 0 0

2/3 1000 31 2302267235 4718980 19 1958 0 0

2/4 1000 26 1910230389 3641661 44 3371 0 0

[deilert]

If you so a sh spantree statistics x/x it will give you some good info re bpdu's

[spacut]

It gave no clue where the problem could be.

BTW What is the way to block sending BPDU frames to the trunk ports leeading to routers? I've typed "spantree portfast enable" on it but it haven't prevent from sending BPDUs to this router/port.

Any clue?

[deilert]

you really do not want to do this (enable port fast )on ports that are connected to other switches or routers, this makes the port go immediately to forwarding. This is what could be causing your problem

[spacut]

I don't follow why it can cause a problem.

My task is to decrease an impact on CPU util done by STP so why it can be problematic?

[deilert]

the switch ports will start forwarding BPDU's immediately bypassing the STP algorithm, the algorithm takes about 40 seconds , this was intended for ports that are connected to pc's server etc. not devices that participate in STP, By enabling you will not get any ports that will be blocking , this could lead to looping frames on the wire .