[c-nsp] snmp counters wrong for SVI with multicast

[Artyom Viklenko]

On 05.04.2012 01:36, Tony wrote:
> Hi,
>
> I recently had similar problems and logged a case with TAC. After discussing/arguing with them for quite some time (2 cases over a period of 4-5 months) they were finally able to reproduce my issue and the net result was:
>
> Filed a bug
> : CSCtj63596&
> the fix is available in 12.2(33)SRE3
> I don't know if this is the same problem as you, but the symptoms are similar (the bug description is fairly vague). I was getting twice the real bandwidth usage in certain monitoring intervals, not continually though. It didn't seem to be specific to multicast in my case, but the box in question does have other multicast traffic through it.
>
> We are currently running 12.2(33)SDR4 and I haven't yet had a chance to upgrade to the version indicated above so I can't tell you whether it has fixed the problem as TAC suggests.
>
>
> If you do upgrade, it would be good to know if it fixes your issue (as it could be related to similar code about SNMP stats).
>

Hi!

Today I rebooted one router and upgrade IOS to 12.2(33)SRD7.
Problem still there. Can confirm that it present on SRD5 and SRD6
also. And as with previous version - delta is equal to total
bandwidth of multicast streams more or less. Router is one of
MPLS core nodes. IPTV traffic lives in MVPN vrf. It comes through
one of uplinks and then exit router via one physycal interface
acting as router port. Output load of this interface is exactly
the delta between physical uplink port and interface vlan carrying
this inbound multicast.

My previous experiments with SRE2 was little bit negative with
multicasting. In our case with SRE2 we got 100% cpu utilization
caused by incoming multicast streams on the router receiving all
100+ streams. So we decide to stay with SRD. Now we have SRE6.
May be it's time to try again. But need deep testing.

Thanks!


>
> regards,
> Tony Miles.
>
>
>
>
> ________________________________
>   From: Artyom Viklenko<artem at aws-net.org.ua>
> To: "cisco-nsp at puck.nether.net"<cisco-nsp at puck.nether.net>
> Sent: Wednesday, 4 April 2012 5:33 PM
> Subject: [c-nsp] snmp counters wrong for SVI with multicast
>
> Hi, List!
>
> Apparently I found some strange situation. We use Cacti to
> draw interfaces utilization on our 7600 routers w/RSP720-3CXL-GE.
> The problem is that cacti draws more traffic on vlan interface
> that it is really forwards and this is only in cases when
> multicast traffic inbound or outbound is present. More over
> if some interface vlan used for only multicast delivery
> (iptv in our case) cacti showes bandwidth near two times more
> than it actually forwards. If some interface vlan receives
> multicast and unicast - cacti also draws more input. and
> delta typically equals to total multicast stream bandwidth.
>
> I try to draw some graphs using anoter host and anoter software
> MRTG. And it also reveals this problem. F.e. I have physical
> port say Gi2/1 configured as trunk and only one vlan say 333.
> if I grapth Gi2/1 - it showes 280Mbps input. At the same time
> graph for vlan333 showes 400Mbps input!
>
> 'show interface xxx' displays relatively equal load on both
> interfaces.
>
> IOS is 12.2(33)SRD6 ADVIPSERVICES. Try to search release notes
> and didn't found any relevant.
>
> Does anybody know what is this and how to fix it.
> I'm preparing to upgrade to SRD7 but still dont know is this
> fix the problem. And in release notes still nothing about this.
>
> Thanks in advance!
>
> --            Sincerely yours,
>                              Artyom Viklenko.
>


-- 
            Sincerely yours,
                             Artyom Viklenko.
-------------------------------------------------------
artem at aws-net.org.ua | http://www.aws-net.org.ua/~artem
artem at viklenko.net   | JID: artem at jabber.aws-net.org.ua
FreeBSD: The Power to Serve   -  http://www.freebsd.org