I have two switches (switchA & switchB) that are connected with a trunked link (VLANs 12, 22 transversing the trunk). HSRP is configured on both of the switches, and SwitchA is the HSRP active for both VLANs.
VLAN 12:
10.166.12.0 /24
SwitchA - 10.166.12.2
SwitchB - 10.166.12.3
HSRP VIP - 10.166.12.1
SwitchA is Active, Switch B is Standby
VLAN 22:
10.166.22.0 /24
SwitchA - 10.166.22.2
SwitchB - 10.166.22.3
HSRP VIP - 10.166.22.1
SwitchA is Active, Switch B is Standby
Now there is a host in VLAN 22, on switchB - IP address 10.166.22.123. If I ping this host from any other host within VLAN 22, I see a 50% packet loss (!.!.!.!.!.). However, if I ping the host from any host outside of the network (for instance, ping from
10.166.12.1), I do not see any packet loss whatsoever. There is another host in VLAN 22, 10.166.22.126 that does not see any packet loss. 22.126 is off of switchB as well.
I see packet loss even if I ping .123 from 22.1 and 22.3:
switcha#ping
Protocol [ip]:
Target IP address: 10.166.22.123
Repeat count [5]: 10
Datagram size [100]:
Timeout in seconds [2]:
Extended commands [n]: y
Source address or interface: 10.166.22.1
Sending 10, 100-byte ICMP Echos to 10.166.22.123, timeout is 2 seconds:
!.!.!.!.!.
Success rate is 50 percent (5/10), round-trip min/avg/max = 1/1/4 ms
Protocol [ip]:
Target IP address: 10.166.22.123
Repeat count [5]: 10
Datagram size [100]:
Timeout in seconds [2]:
Extended commands [n]: y
Source address or interface: 10.166.22.1
Sending 10, 100-byte ICMP Echos to 10.166.22.123, timeout is 2 seconds:
!.!.!.!.!.
Success rate is 50 percent (5/10), round-trip min/avg/max = 1/1/4 ms
So as I'm troubleshooting this, I guess I have some remedial questions. Host 10.166.22.123 receives an ICMP request packet from 10.166.22.3. If host .123 wants to send an ICMP reply packet to 22.3 (which is the switchB VLAN22
interface), it would send an ARP request out to the 255.255.255.255 address (if it doesn't already know the MAC address for the host - which it should since it just received a packet from the host), and should receive a response back with the correct mac-address. Then the .123 host should create a frame with the destination MAC address that it received in the ARP request and will forward the ICMP packet to switchB, which should then notice that the packet is destined for itself. Am I missing something in this process? Since the communications remain within the single network, the CEF adjacency table would not come in to play, correct? I apologize if this is remedial -- I'm just trying to get an idea of why this packet loss would occur - and only involve one host. The most baffling part, is that there is no packet loss when pinging the host from a different network.
BTW - The network configuration of the
host machine has been verified.
I have sniffed ICMP on the trunk between the two switches and this is what I saw:
Sniff Test #1 - When sending 10 pings from 10.166.22.2 (switchA) to host 10.166.22.123 (connected to switchB) - I see 10 requests and 10 replies in Ethereal.
Sniff Test # 2When sending 10 pings from 10.166.22.3 (switchB) to host 10.166.22.123 (connected to switchB) - I see 4 ping replies in Ethereal.
I would have assumed that in Sniff Test#2, I should not have seen any requests or replies. I'm planning on sniffing ARP in addition to ICMP to see what else is going on.
I would appreciate any kind of input or ideas for add'l troubleshooting. I'm finding that I have some misconceptions about how this should be working. Has anyone seen this kind of behavior before
THanks in Advance.
-Tim
How low will we go? Check out Yahoo! Messengers low PC-to-Phone call rates.
