Some of you who know me in IRC world may have seen this query before a
few months back, but maybe I may get some good input with all the
facts in one message...

Basically, I'm running an Ubuntu Server 10.04 host that has some weird
ping issues. I've seen this in 9.10 as well.

I have three clients:
- #1: 192.168.62.42
- #2: 192.168.63.151
- #3: 192.168.67.20

I have one server:
- eth0: 192.168.62.105
- eth1: 192.168.63.44

What I know:
- Client #1 (62.0/24) receives a response when pinging eth0 (62.0/24).
- Client #1 (62.0/24) does not receive a response when pinging eth1 (63.0/24).
- Client #2 (63.0/24) does not receive a response when pinging eth0 (62.0/24).
- Client #2 (63.0/24) receives a response when pinging eth1.
- Client #3 (67.0/24) receives a response when pinging both eth0
(62.0/24) and eth1 (63.0/24).

My wording is intentional; i've run tcpdump on both interfaces and
confirmed that the request is a) being received and b) the response is
not going out the wrong interface.

Kernel IP routing table:
Destination     Gateway         Genmask         Flags   MSS Window  irtt Iface
192.168.50.0    192.168.62.51   255.255.255.0   UG        0 0          0 eth0
192.168.64.0    192.168.62.51   255.255.255.0   UG        0 0          0 eth0
192.168.62.0    0.0.0.0         255.255.255.0   U         0 0          0 eth0
192.168.63.0    0.0.0.0         255.255.255.0   U         0 0          0 eth1
0.0.0.0         192.168.63.1    0.0.0.0         UG        0 0          0 eth1
0.0.0.0         192.168.62.1    0.0.0.0         UG        0 0          0 eth0

- I have tried this without a default gateway set for eth1 and have
the same result.
- The problem manifests when the network entry for 192.168.63.0 is
present. If i use 'route delete -net 192.168.63.0/24' client 2 can
ping eth0.

Other notes:
- Each subnet is it's own VLAN and has it's own router. I think we can
best think of these as separate networks that are routed. Also,
Solaris and RHEL5 don't have this problem.
- I have seen this behavior strictly on Ubuntu 9.10 and later.
- This is running in a VM - however, I've replicated this behavior on
a physical machine too.
- I tweaked some sysctl arp settings (arp_ignore, arp_announce,
something with arp_filter) to no avail. ip_forwarding as well.
- Another server of mine that runs Debian has a similar problem but
it's "fixed" by keeping it's ARP entry fresh in the switches that each
interface lives on (pings the router once a minute). Gross solution
but it works.

Any thoughts? I'm out of ideas here.
_______________________________________________
Tech mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to