04-06-2010 07:27 AM - edited 03-04-2019 08:02 AM
please am having high cpu utilisation of over 80% on my GGSN cisco 7600 vxr and 7200.am doing GPRS-GTP on the software loopback interface.please how do i manage this issue.
thanks alot
04-06-2010 12:57 PM
Hello Olankule,
take sh proc cpu sorted
sh proc cpu sorted 1min
from affected devices in order to understand what processes use more resources and how much cpu is used by SW interrupts
To be noted this could be caused by an increase in traffic volume and number of GTP sessions
Hope to help
Giuseppe
04-07-2010 12:51 AM
thanks for the help.
after issuing the command i noted IP INPUT process consuming most memory.so how can i sove the problem.thanks for anticipated response
04-07-2010 01:41 AM
Hello Olankule,
so IP input uses a lot of cpu resources this probably means that the router is process switching a lot of packets.
Can this increase in cpu usage be caused simply by an increase in traffic?
can you compare current traffic levels on router interfaces with that of previuos time periods?
What was the cpu usage before?
My guess is it could be relatively high also before because to encapsulate IP packets into GTP tunnels is cpu intensive.
Have you performed any configuration change or change of IOS image recently?
Examing sh log on router do you see any message related to cpu usage or to cpu processes taking too much cpu time?
Hope to help
Giuseppe
04-07-2010 05:56 AM
04-07-2010 06:21 AM
Hello Olankule,
first of all sorry for the basic note:
but you don'have debug ip pack detail all the time I hope otherwise this can cause the issue you see (debug causes very high cpu usage)
looking at the debug lines I see that there are some fragmented GTP packets
-More-- Apr 3 21:58:00.282: IP: s=10.105.5.41 (FastEthernet1/1), d=41.203.65.114, len 76, rcvd 4
--More-- Apr 3 21:58:00.282: UDP src=2152, dst=2152
--More-- Apr 3 21:58:00.282: IP: tableid=0, s=41.203.64.44 (FastEthernet1/1), d=41.203.65.114 (Loopback6), routed via RIB
--More-- Apr 3 21:58:00.282: IP: s=41.203.64.44 (FastEthernet1/1), d=41.203.65.114, len 88, rcvd 4
--More-- Apr 3 21:58:00.282: UDP src=10651, dst=2152
--More-- Apr 3 21:58:00.282: IP: tableid=0, s=41.203.65.97 (FastEthernet1/0), d=41.203.65.114 (Loopback6), routed via RIB
--More-- Apr 3 21:58:00.282: IP: s=41.203.65.97 (FastEthernet1/0), d=41.203.65.114, len 1500, rcvd 4
--More-- Apr 3 21:58:00.282: IP Fragment, Ident = 1910, fragment offset = 0
--More-- Apr 3 21:58:00.282: UDP src=10651, dst=2152
this can cause an increase of cpu usage, original packets can be of normal size but then there is the GTP overhead and this leads to the need of fragmentation
Note:
My is only a guess that GTP can cause high cpu usage based on the fact that packets needs to be inspected deeply to be placed in the correct session within the GTP tunnel it is wise to open a TAC SR if this is impacting or it is going to impact service
Hope to help
Giuseppe
04-07-2010 06:35 AM
THANKS FOR KIND SUPPORT AND EXPLANATION. WOULD CONTACT TAC ASP.
I must really appreciate your contributions on this forum,your responses are very useful
and precise.
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