[c-nsp] c6k msfc - hsrp flapping

[Yuri Lukin]

Lee,

lee.e.rian at census.gov wrote ..
> Hi Yuri,
> 
> The assumption I've been making is that enough multicast traffic hitting
> the msfc to cause hsrp and eigrp to flap would also cause very high cpu
> busy and/or lots of input queue drops.  Maybe that's a bad assumption,
> but
> that's why I think the traffic is being dropped before it gets to the msfc.
> But I don't know what commands to run to prove or disprove that theory.
> 

"show mls multicast statistics" on the SP side is probably the best command for this. 

I think you can look at output from "show counters 15/1" from the SP side
(assuming your active MSFC shows up as module 15 in the switch). Also, 
"show mac 15/1" would show how many unicast/multicast/broadcast frames
are going to/from the msfc. The command "show mls rlog l2" also shows
the messages that the switch is receiving from the active MSFC. 

> 
> > How are they interconnected?
> 
> L3 point-to-point links between the routers.
> 
> > Can you check the various counters to see which interfaces are
> > showing excessive multicast traffic?
> 
> None of the multicast counts look excessive to me
>   (counters haven't been cleared & uptime is 2 days, 16 hours)
> I hadn't thought about it until now, but unless hsrp keepalives aren't
> counted the multicast counts are impossibly low.
> 

The counters I was referring to were the individual port counters from the SP side. 
For example, "show mac" and "show counters mod/port". Also, the various 
"show mls" commands can help...

> 
> 
> > Perhaps you can "debug standby" as the problem is happening and paste
> > the output to the list?
> 
> Probably not unless I can tell my boss that I've got a pretty good idea
> of
> what the problem is & how to fix it.
> I think we're just going to leave multicast turned off & upgrade the core
> sup1a/msfcs to sup720s.
> 
> Thanks,
> Lee

Did it ever work or was the problem there from day 1? If multicast worked fine 
at this site without any ill side effects, then something must've changed
recently. Perhaps the amount of multicast traffic went up significantly, 
enough to cause a problem elsewhere. Don't give up yet, the smoking gun
is there somewhere..you just have to find it ;)

-Yuri