Cisco 2911 512Mb DRAM BGP Full Table Packet Loss Problem

[jeliasoncisco]

Hello,

I have recently installed a Cisco 2911 ISR G2 with the default 512 Mb DRAM intending to eBGP peer. I ordered the 2GB upgrade RAM however due to time constraints on backordered parts, I fired up this router and eBGP peered without it. The Peer advertised the whole route table with 400,000+ routes. The BGP session came up then the router crashed due to not enough memory. The router disabled IP CEF due to insufficient memory. I disabled IP CEF permanently and have been running the router in this condition for 3-weeks with a stable eBGP session. This resulted in no CEF, 25% CPU during light traffic, 89% memory, and 50% CPU when traffic is around 30 Mbps through the router.

The problem I am looking for help on is this: I am experiencing a hit to the throughput resulting in a lost packet and practically a brief traffic stall roughly every minute. This hit is so quick that it does not always result in packet loss and IP traffic sessions are not reset. I do see this on my live bandwidth graphs that the traffic takes a dive every so often, roughly 1-minute.

I initially thought this problem could be L2 to the upstream eBGP peer but all interfaces are clear of errors. I also thought this could be the BGP session going down, however, It is always up. I thought this could be duplex mismatch on L2, however its solid and no logs on either end. Funny thing is pinging thr router from both the LAN side and the WAN side results in the same packet lost every minute or so.

Even though the CPU and memory always stays the same at under 20-50% CPU and less than 89% memory, do you think this could be the BGP Scanner walking the routing table every minute?

Thanks for your thoughts.

Jordan

[jeliasoncisco]

I just wanted to add that this router is only functioning as an BGP AS Border Router. I am not running any other services like NAT, VPN, or otherwise. Only have ACLs on interfaces and eBGP configured.

[patrick.preuss]

Hi Jordan do you have The Need for a Full table or will it be sufficent for The Moment to work with a Default Route from the isp.

Bgp Full table has High demand in Memory and CPU.


Sent from Cisco Technical Support iPhone App

[jeliasoncisco]

Hi Patrick,

Thanks for your advice. I was not intending to run with this little RAM, however was stuck due to backordered parts. A spike in CPU is not registering on my NMS or within the show proc cpu history, so I am thinking when the bgp scanner runs it is so momentary that it causes the packet loss.

My thoughts is to trim the routing table down to Peer and Their Peers' prefixes. Then review the result. This will reduce the routes in the table to roughly 35,000 routes vs the 437,000 full table.

Will let you know what happens

[Richard Burts]

I agree that if something is creating impact every minute that it sounds like the BGP scanner is what is causing it. And I agree that reducing the size of the BGP table is the best way to deal with this issue.

HTH

Rick

[Joseph W. Doherty]

Disclaimer

The    Author of this posting offers the information contained within this    posting without consideration and with the reader's understanding that    there's no implied or expressed suitability or fitness for any  purpose.   Information provided is for informational purposes only and  should not   be construed as rendering professional advice of any kind.  Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In    no event shall Author be liable for any damages whatsoever  (including,   without limitation, damages for loss of use, data or  profit) arising  out  of the use or inability to use the posting's  information even if  Author  has been advised of the possibility of such  damage.

Posting

Since you're pulling Internet routes from your ISP, including perhaps moving to just their prefixes, you have multiple ISPs?

Assuming you do, another possible option would be to consider PfR, if it's supported on your ISP connected routers.

Years ago, I had a pair of 3660s using DS3s, each with two ISP full BGP tables, suffering from CPU loading, especially from the BGP scanner.  Converted them to OER over default route, CPU load dropped very, very much and network performance actually was as good as or better than it was.

[patrick.preuss]

Hi


I know ...

Currently we are Running Part table (50 %) roundabout with 7200 g2 Engines and 1g RAM . From my experience 2 Gb plus and a Good CPU are required. X800 Series kann Handel only Part Off The table Caused By Hardware Design.

Patrick

Sent from Cisco Technical Support iPhone App