strange BGP problem - CPU and BGP updates

Konstantin Dunaev · ‎03-17-2010

Hello,

we have some strange problem on our c7200 IOS 12.2(31) with an eBGP peer.

the eBGP session begins as usuall and without any problem, all prefixes (Full BGP table) are announced to the neighbour, of course during this time the CPU goes high up to 80%. Then the both peers are in sync and traffic run as usuall, but after some minutes the CPU goes again upto 80% and the debug on the neighbour side shows that it gets again the BGP updates for all BGP prefixes from our 7200 with "rejected - duplicated" comments for every recieved prefix, it means that it gets the updates for prefixes which are allready exists in BGP table.

And it heppens every 3-5 minutes. Trully said I can't explain why our 7200 always tries to send the Full bgp table again and again.

The work-around was to configure the peer with "soft inbound", it's OK and we'd like to use it anyway , but from my opinion BGP shouldn't send Full BGP table very 5 minutes without "soft inbound" as well.

Can somebody try to explain why it couls happend?

Regards,

Konstantin

milan.kulik · ‎03-17-2010

Hi,

are those eBGP neighbors directly connected or multihop?

If the later, I remember similar issue in my lab while having default route only pointing to the neighbor.

Fixed by adding a static route to the neighbour address.

HTH,

Milan

Konstantin Dunaev · ‎03-17-2010

Thank you for the answer!

they are ebgp-multihop connected, but there is an exact route in routing table (from iBGP table ) for that eBGP peer.

milan.kulik · ‎03-18-2010

Hi,

just guessing without knowing your topology details, but you might be in some kind of this problem:

http://www.cisco.com/en/US/customer/tech/tk365/technologies_tech_note09186a00800945fe.shtml#topic1

I'd recommend configuring a static route to your BGP peer IP address.

HTH,

Milan

Konstantin Dunaev · ‎01-19-2012

Hello,

if somebody still interested in, we've found the problem - it was BGP related "bugs" on c7200p-a3jk91s-mz.122-31.SB13.bin

after upgrating it to the last version in SB train c7200p-a3jk91s-mz.122-31.SB21.bin

all CPU related problems are gone.

Kishore Chennupati · ‎01-19-2012

Konstantin,

Its not recommened to use "soft-configuration inbound". as it actually increases the memoory and CPU.

These days with the advancements in BGP, the devices use "route refresh" capability

http://www.cisco.com/en/US/products/ps6599/products_data_sheet09186a0080087b3a.html#19969

Konstantin Dunaev · ‎01-19-2012

Hi Kishore,

thank you for advise, on most BGP session I've removed it already.

but our problem was a bit different - I was observing th "sh ip bgp summary" on the NPE-G2 router and could see then "BGP Path" was constantly growing up to 600K, all other routers (NPE-G1 and smaller) don't have such behaviour. And of cause the processing of 600K BGP takes too much CPU time, after IOS update, "BGP Path" stays stable as on all other routers.