Attached is a switching layout that is in production. I'm having an issue with ttl expired responses that terminate at 2 newly installed Dell 8024s. When I ping an offline device with a valid ip address from my workstation I get ttl expired errors coming from these switches. The switch/router is the default gateway for the network so I'm curious to why the error is not originating from the router. The first response will come from the .18 switch and then the final 3 from the .17???? EE8024SwitchConfig.pdf
Switches / HubsNetworkingDell
Last Comment
mat1458
8/22/2022 - Mon
Don Johnston
What is the source IP address and destination IP address of the pings?
cs2009
ASKER
both in same subnet:
Host 10.10.1.100
destination 10.10.12.33
SM 255.255.0.0
Don Johnston
If what you're describing is correct then the behavior is not really possible. The TTL field is only decremented by a router (layer 3 device). If the source is 10.10.1.100/16 and the destination is 10.10.12.33/16, they hosts are on the same network and no router is needed. Which means the TTL isn't being decremented so there could be no TTL expired.
Unless the sending host is transmitting with a TTL of zero which isn't likely either.
I suspect there's more to your network than meets the eye. Can you provide configs of the network devices and indicate which ports the 10.10.1.100 and 10.10.12.33 hosts are connected to.
exactly my thought... there is no hop on the same subnet. Both hosts are connected to ports on Vlan 10 on separate switches. I expected to get a request timed out but not so. All switches have routing on Vlan 10 enabled. The dell 8024 is a new, different animal to me compared to the HPs. The previous image shows the basic topology. 8024 startconfig attached. startup-config
Don Johnston
I have no idea why a switch (or router) would respond with a ICMP TTL exceeded message for a device on the same network as the sender.
Also from the target host. To me it looks like one of them may have a subnet mask other than 255.255.0.0.
giltjr
You may also want to look at all the routing tables in all of the devices. There may be a device with a entry for 10.10.0.0/24 or 10.10.1.0/24. Which could also cause weird results.
The printer is connected to a switch down stream. I cleared cache on both switches. After I ping host an entry is written back to the cache. As before the ping request will go to the .17 switch on the first attempt and then .18 for the final 3 attempts?
I found parameters on routing interface and proxy arp and local proxy arp are indeed checked. If I uncheck and ping host I now get ttl timeout on last resort gateway on the actual router, which is where I thought it should be.
On the dell 8024, the 2 arp settings are enabled by default. Should I just keep the defaults and go with lesson learned? Thanks guys.
Weird. I personally would not use proxy arp. Since so many network monitoring products use ping as one way to see if a device is up and to measure latency issues, I would leave it off.
Otherwise your monitoring software will show the device is up when in fact it may not be.
cs2009
ASKER
thanks giltjr
mat1458
To be clear: this seems to be a bug of Dell. The switches must never proxy ARP for a request in the same subnet. Proxy ARP is only useful across subnet boundaries.