[c-nsp] Problem with Microsoft NLB on Server 2008 running in Multicast mode

Peter Rathlev peter at rathlev.dk
Thu Mar 25 16:57:33 EDT 2010


On Thu, 2010-03-25 at 21:11 +0100, Asbjorn Hojmark - Lists wrote:
> On Thu, 25 Mar 2010 20:01:55 +0100, you wrote:
> > C6k(config)# arp vrf X 10.0.0.1 03bf.0abc.def0 ARPA
> 
> If it is in fact a 6500, that alone is a pretty bad idea, as traffic
> bound for the cluster will be process-switched. It is recommended to
> add at static MAC entry as well.
> 
>  mac-address-table static 03bf.0abc.def0 vlan X int Y disable-snoop

Hmm, I can't seem to reproduce this.

With just a static ARP like above and a flood ping of about 1000 pps the
"show interface VlanX stats" doesn't show the "processor" part rising.
At least not more than a single packet or two at most.

! C1#sh run part common | incl arp
! arp vrf SomeVRF 10.0.0.1 03bf.0a55.151e ARPA
! arp vrf SomeVRF 10.0.0.2 03bf.0a55.151d ARPA
! C1#sh ip route vrf SomeVRF 10.0.0.1
! Routing entry for 10.0.0.0/24
!   Known via "connected", distance 0, metric 0 (connected, via interface)
!   Advertised by bgp 65432
!   Routing Descriptor Blocks:
!   * directly connected, via Vlan121
!       Route metric is 0, traffic share count is 1
! 
! C1#clear count vlan 121
! Clear "show interface" counters on this interface [confirm]
! C1#
! 006022: Mar 25 21:46:53.762 CET: %CLEAR-5-COUNTERS: Clear counter on interface Vlan121 by me on vty0 (a.b.c.d)
! C1#sh int vlan 121 stats
! Vlan121
!           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
!                Processor         12       9592          4        264
!              Route cache          0          0          0          0
!        Distributed cache          0          0          0          0
!                    Total         12       9592          4        264
! C1#

Then I send 1000 ICMP packets towards the target:

# [21:46:45 root at einstein ~]# ping -f -c 1000 10.0.0.1
# PING 10.0.0.1 (10.0.0.1) 56(84) bytes of data.
#  
# --- 10.0.0.1 ping statistics ---
# 1000 packets transmitted, 1000 received, +992 duplicates, 0% packet loss, time 1644ms
# rtt min/avg/max/mdev = 1.514/1.862/17.698/1.220 ms, pipe 11, ipg/ewma 1.646/1.651 ms
# [21:47:05 root at einstein ~]# 

Counters afterwards:

! C1#sh int vlan 121 stats
! Vlan121
!           Switching path    Pkts In   Chars In   Pkts Out  Chars Out
!                Processor         48      39630         14        924
!              Route cache          0          0          0          0
!        Distributed cache       9210    7655923      10213    2692260
!                    Total       9258    7695553      10227    2693184
! C1#

Am I looking in the wrong place? I can't seem to find anything in
release notes or via Bug Toolkit that mentions that mentions the process
switching.

-- 
Peter




More information about the cisco-nsp mailing list