07-29-2009 01:14 AM - edited 03-06-2019 06:59 AM
Hello,
I have a 4900M that is running with constant CPU usage of ~85%. I have 5 other of these switches with the same configuration and placing in the topology that do not have the problem.
sh proc cpu
CPU utilization for five seconds: 82%/0%; one minute: 85%; five minutes: 86%
47 233407012 643237989 362 4.23% 4.27% 4.31% 0 Cat4k Mgmt HiPri
48 1660690936 724232776 2293 76.81% 80.13% 80.72% 0 Cat4k Mgmt LoPri
sh plat heal
%CPU %CPU RunTimeMax Priority Average %CPU Total
K5L3Unicast Adj Tabl 2.00 47.80 15 11 100 500 48 49 37 44108:318
K5 L2 Hardware Addre 2.00 49.79 10 9 100 500 61 55 43 58187:346
Does anybody have any idea from the above data what the problem could be?
As I mentioned above the config is almost identical to that of the other switches that do not have the problem, and all other data looks similar (including sh interfaces switching) although there does seem to be a lot of spanning tree output on all interfaces.
Please let me know if you require furtehr information/output.
Thanks in advance.
07-29-2009 01:27 AM
Hello Christopher,
I've found a document about 4500 high cpu
or
http://www.cisco.com/en/US/products/hw/switches/ps663/products_tech_note09186a00804cef15.shtml
However, if there is an high rate of STP messages there can be too many STP topology changes and you need to investigate this as noted in
or
http://www.cisco.com/en/US/products/hw/switches/ps663/products_tech_note09186a0080094956.shtml
look for flapping links, what STP mode are you using?
Hope to help
Giuseppe
07-29-2009 02:47 AM
Thanks for the quick reply. I've been through the 4500 document, it doesnt really help too much, just instructs to use the usual 'sh proc cpu', 'sh platform health' and 'sh interfaces switching'. There are no STP topology changes just what seems an usually large amount of packets output for the switch uptime.
GigabitEthernet2/1
Throttle count 0
Drops RP 0 SP 0
SPD Flushes Fast 0 SSE 0
SPD Aggress Fast 0
SPD Priority Inputs 0 Drops 0
Protocol Path Pkts In Chars In Pkts Out Chars Out
Other Process 0 0 151844 9110640
Cache misses 0
Fast 0 0 0 0
Auton/SSE 0 0 0 0
PAgP Process 0 0 164099 19363682
Cache misses 0
Fast 0 0 0 0
Auton/SSE 0 0 0 0
Spanning Tree Process 0 0 57274993 3654508948
Cache misses 0
Fast 0 0 0 0
Auton/SSE 0 0 0 0
VTP Process 0 0 14913 1521126
Cache misses 0
Fast 0 0 0 0
Auton/SSE 0 0 0 0
I have noticed there is a lot of flapping interfaces that are assigned to port channels, it flaps between one or the other link. They are connected to teamed network adapters on ESX boxes.
16w2d: %C4K_EBM-4-HOSTFLAPPING: Host 00:50:56:A4:68:F2 in vlan 110 is flapping between port Gi2/1 and port Gi2/2
16w2d: %C4K_EBM-4-HOSTFLAPPING: Host 00:0C:29:DE:86:63 in vlan 110 is flapping between port Gi2/1 and port Gi2/2
interface GigabitEthernet2/1
switchport mode trunk
channel-group 4 mode desirable
!
interface GigabitEthernet2/2
switchport mode trunk
channel-group 4 mode desirable
!
interface Port-channel4
description "ESX trunk"
switchport
switchport mode trunk
!
Im not particularly familar with port channels used in this way to teamed network adapters. It does seem strange to me that the port channel interface has been created in the same way you usually would for a L3 etherchannel, but has no ip assignment, and both the physical and logical interfaces are set to trunk.
Any further assitance would be appreciated.
Thanks again
07-29-2009 10:37 AM
Hello Christopher,
I see a possible problem here:
with mode desirable you are asking the switch to try to use proprietary PAGP to negotiate a bundle.
Towards servers we use LACP that can be enabled with
mode active on switch side.
I suggest you to check with server people.
To change to an LACP bundle I would suggest you to use a different channel number.
This should explain why you see host mac flapping: actually the bundle is not active from the point of view of the switch.
Hope to help
Giuseppe
07-30-2009 04:57 AM
Sounds very likely that this is the cause of the problem.
I did a bit of further reading and it seems Cisco recommend setting the channel-group mode to 'on' so that the bundle is forced and does not use any protocol.
I will implement the changes over the weekend.
Thank you very much for the help.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide