Meanwhile I found a workaround, though no real explanation.
The solution was to configure all servers using static IPs instead of DHCP *and* to remove the entries from the router DHCP reservation list. I think I already tried it earlier without success, but maybe I forgot to remove the systems from the router DHCP list. Anyway, this way - static IP and no entries for the MACs on the router - it works. As soon as I switch the servers to dynamic IPs (with or without DHCP reservations on the router), it doesn't work anymore - though the servers *do* get IP addresses and *can* communicate with all other systems, except the router. This static IP setup is ok for me, but still it would be nice to have an explanation. Closest thing I can think of is that the router - though it has the IP/MAC relationship in its static list (DHCP reservation) - for some reason doesn't know/notice that those IPs are online and so never react on packets from them. This might also be an issue with the Vigor router, this device already gave me headaches earlier (I still didn't find the ideal router .-).
Thanks for your help anyway, I hope the "solution" helps others.
I'm not sure if this is an issue with the Cisco WET2000, or with CentOS 6.0...
We recently got a Cisco WET200 bridge (firmware 126.96.36.199-ETSI) to connect some servers with only ethernet interfaces to our LAN where we don't have a wired connection between the roomn with the servers and the LAN (powerline doesn't work neither btw). The LAN is connected to the internet (cable provider) using a Vigor2110Vn (firmware 188.8.131.52). The Vigor serves as router and access point to the LAN (IP 192.168.0.1). There are wired and wireless systems connected to the Vigor.
The servers on the seperate segment conencted via the Cisco bridge are all running CentOS 6.0. To connect the servers to the bridge, we have a switch (3com) between them so all servers are on the same port 2 of the bridge. I set the servers up using DHCP using the Vigor as DHCP server, and their MAC addresses are bound to fix IP addresses in the Vigor (currently 192.168.0.211 - 192.168.0.213).
The whole setup works, the servers receive their IP addresses correctly, they have the Vigor router also correclty in their arp table. I can also without any problems reach the servers from the LAN and vice versa, I can for example mount a filesystem from our NAS (IP 192.168.0.17) over the bridge. Performance is quite nice.
Only one thing does not work: I can't reach the router from the CentOS servers. Pings from server to router don't succeed, and of course no connection to the internet is possible at all (because the Vigor is the default gateway). But, as I mentioned, the DHCP setup works; when I change the assigned IP on the Vigor, the router gets the new IP at the next reboot, so server and Vigor can communicate with each other using DHCP. But obviously not in any other way. Every other system on the LAN is visible from CentOS though, both wired and wireless systems. Just not the router itself. I can even see the router's (correct) MAC and IP in the servers arp table (arp -a). Firewall on the servers are turned off.
The real strange thing comes here: when I shutdown the server and connect it to the router directly (wired), skipping the bridge, everything works perfectly. This would be an indication that the bridge has an issue.
If I connect my standard laptop (Windows 7) to the Cisco bridge though, on the same switch where the servers are located, this *also* works perfectly! This would be an indication that the issue is with CentOS and contradicts the first test. In further tests I verified MacOS Lion (directly), Suse Linux inside a VMWare (bridged), and even CentOS 5.2 in a VMWare (bridged) in the seperated segment; they all work flawlessly. Only CentOS 6.0 (both native and inside a VMWare) don't work over the bridge to the router.
Even if I install VMWare with a Windows XP guest OS on the CentOS 6.0 server as host, this guest OS works perfectly if I set its ethernet interface up as bridged (so it gets an IP address in the 192.168.x.x range), while - as said - the CentOS 6 host itself doesn't get through. This excludes (I think) hardware issues with the network interface. If I set up the Windows XP guest as NAT, of course it doesn't work.
I also tried sniffing the traffic. Unfortunately I can't sniff directly at the Vigor to see what reaches it, because there is only a wireless connection there. So I tried to mirror the wired (and also wireless) port on the Cisco to port 1 and sniffed there. Unfortunately I only see packets from the servers to the LAN and no reply packets, not even for working connections between servers and LAN where there *must* be packets in both directions, and yes, of course I tried "Ingress", "Egress" and "Both". I seem to do something wrong there. So sniffing didn't give me any indications what could be wrong.
Any ideas or suggestions? It seems to be something related to the interaction Centos6+Cisco+Vigor, because with only 2 of those components there are no issues: servers directly on LAN work (Centos6+Vigor), servers to other LAN systems work (Centos6+Bridge), and laptop on bridged segment works (Bridge+Vigor). No idea what else I could try... Possibly some MTU size issue? Fragmentation?Thanks for any advice,