Problem with 6509-E router and bgp full route

[JGFR]

Hello

I currently have a 6509-E, I tried to set up on this last two freight forwarders in full-road except that this last had a concern of memory full.

The logs :

.Jan 12 10:54:57.944: %BGP-3-NOTIFICATION: sent to neighbor IP.ADD.RE.SS-Transit2 3/1 (update malformed) 0 bytes

.Jan 12 10:54:57.944: %BGP-4-MSGDUMP: unsupported or mal-formatted message received fromIP.ADD.RE.SS:

FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF 005C 0200 0000 2E40 0101 0040 020E 0203

0000 00AE 0000 34E0 0000 0F5B 4003 04CC 44FC 7580 0404 0001 2138 C008 0800 AE52

0900 AE55 FD14 D8E6 A011 D117 0012 D014 4014 D00D F012 CC61 8010 A4C3

.Jan 12 10:54:57.952: %BGP-4-BGP_OUT_OF_MEMORY: BGP resetting because of memory exhaustion.

Example BGP down logs : 

.Jan 12 10:55:48.813: %BGP-5-ADJCHANGE: neighborIP.ADD.RE.SS-TransitCustomer1 Down No memory

I know the 6500 only supports 768k IPv4 on my configuration, but the routes are normally unique. 

Is it a problem ? Is this normal for a 6500? Do you have a solution?

For information I have a SUP7203BXL.

Thanks you for your help. 
Jo

[Georg Pauwen]

Hello,

the SUP7203BXL has 1GB of RP DRAM, if you haven't upgraded it. I am not sure about the first part of your post, are you receiving the full routing table from two neighbors ??

Either way, a 'show processes memory | i bgp' should show you the amount of memory used by BGP. Two full routing tables would likely exceed the 1GB you have installed.

[JGFR]

Hello,

Yes i need receive the full routing table from two neighbors to get best path.

Currently i think have 2x1Gb of memory on my 6500 router.

SUP7203BXL support 2Gb for routing processor ?

Jordan

[Georg Pauwen]

Hello,

the default seems to be the max:

WS-SUP720-3BXL supports 1GB DRAM for route the processor and 1GB DRAM for the switch processor.

I would consider the options Joseph mentioned above.

[Georg Pauwen · Accepted solution]

You could also configure an AS_PATH filter to accept only routes from the ISP and directly coonnected ASs, as described in this document. Unless you are an ISP yourself, you would never really need the full BGP table anyway.

http://www.cisco.com/c/en/us/support/docs/ip/border-gateway-protocol-bgp/12512-41.html#bgpreceivedtable

[Joseph W. Doherty]

If you have any line cards with DFCs, they are all the XL variant too?

You're only taking IPv4 or IPv6 too?

Assuming you've indeed exhausted your XL memory, one possible solution would be to "cull" the routes you're accepting.  For example, any destinations more than than one or two AS hops, you might use a default route.  (Yea, I know, not as optimal, but if that's a real concern you either need a router that can accept your full route tables, or you might, if supported, consider using PfR.  [Years ago, I used the original version of PfR, OER, with two ISPs, and replaced using full BGP route tables with default routes.  OER provided even better path selection.])