IP Networking

I have a Linux server connected to the Internet like so:

Server (66.77.88.130) -> /29 VLAN Gateway (66.77.88.129) -> Core Router (64.0.0.1) -> Internet

If I run a traceroute from my server to an arbitrary site on the internet I get a result like this:

Code:

# traceroute some-other-server
traceroute to some-other-server (1.2.3.4), 30 hops max, 40 byte packets
 1 66.77.88.129  0.468 ms  0.622 ms  0.806 ms
 2 64.0.0.1  0.329 ms  0.343 ms  0.362 ms
 3 edge2.NewYork1.Level3.net (4.71.230.9)  0.401 ms 0.400 ms  0.4 03 ms
<<snip>>

But if I run a traceroute to my server from another location I get a result like this:

Code:

# traceroute my-server
traceroute to my-server (66.77.88.130), 30 hops max, 40 byte packets
<<snip>>
 10 edge2.NewYork1.Level3.net (4.71.230.9)  20.401 ms 20.400 ms
 11 64.0.0.1  21.329 ms  21.343 ms  21.362 ms
 12 66.77.88.130  21.468 ms  21.622 ms  21.806 ms

You'll notice the vlan gateway (66.77.88.129) is missing from the second traceroute.

Can anyone tell me why that is the case or explain how it works out like that?

BTW: let me make it clear that my network works fine - this is a question out of curiosity.

This also sounds curious to me, the gateway has to be there, unless some funky rules are placed somewhere. This is an ICMP packet, and it has to go through the default gateway first. The "issue" is definitely located inside of your network, and as it seems, it's valid only for outgoing requests. I would suggest to try different tools like "mtr" or "tcptraceroute" and post the results, while once again having checked the rules on various firewalls / routes and / or different network equipment.
HTH

No change - similar results - missing the default gateway hop on ingress traffic.

I suspect it must be something to do with the vlan since the default gateway is one of those enterprise managed switches (think Cisco/Juniper - this one's an older Foundry) - quite possibly the switch is changing between routed and bridged mode or something based on which way the packets travel.

Like I say, it's one hop less so I'm not really concerned about "fixing" this issue.

It does seem strange that a packet could appear to bypass the physical network layer tho