[c-nsp] Sup720 - FIB full, software switching

[Lobo]

One other thing I noticed from your email and something that we've 
experienced in the past as well.  I think it may also be related to 
hitting the TCAM limit but check to see if you have this command enabled:

mls rate-limit unicast cef receive 10000 255

According to Cisco, that command will automatically get added to your 
config when the tables get full.  That command will start to drop 
packets and unless you look for it you wouldn't know it's there because 
generally it's not.  All BGP sessions appear normal and none of your 
interfaces show full yet you're still dropping packets.  Cisco advised 
us to increase the receive to 1000000 to avoid any possible issues in 
the future.

Thanks to the other replies about having to reload the switch to clear 
the TCAM exception.  I didn't know that once you hit it that the only 
way to fix it was to completely reload the box.

Jose

On 2/3/2014 9:09 AM, "Rolf Hanßen" wrote:
> Hi,
>
> today I saw 2x Sup720-3B (default 192K IPv4 routes) that received a full
> table.
> After FIB was filled IOS gave a warning that it now may forward in
> software (and resetted all BGP sessions because of memory issues). I don't
> have the exact messages.
>
> The real problem occured after that. I shut the full table BGP session and
> cleared the others, the system now had a few routes only again.
>
> But it started to drop packets, I saw no pattern, it looked nearly random.
> I needed to reboot both boxes to resolve that issue.
>
> IOS was s72033-advipservicesk9_wan-mz.122-33.SXJ.bin
>
> Is there a way to avoid those issues by let it just ignoring routes not
> matching into the FIB?
> Is there a command to reset the routing mode/routes back to CEF without
> reloading the box?
>
> kind regards
> Rolf
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/