Hi,
i have a really weird problem here.
I got some Debian Etch 4.0 running with bonding ifenslave.
There around 30 other linux Etch 4 servers in the same Subnet, and 10 with bonding enabled.
They are connected to 2xNortel 5510 Stack and running good with MultiLinkTrunk.
Everything ok here.
Now i changed the Ip-Adress of a bonding-box to another ip from the same subnet. Lets call it
WWW1
All ok, i can ping all Servers, also can route out.
BUT:
Now i have ONE non-bonding server which can NOT ping nor communicate in any way with the changed bonding-server and vice versa. Lets call it
INT
I start "ping INT" on WWW1
I start "tcpdump host WWW1" on INT
WWW1 tells me:
Code:
www1:~# ping INT
PING INT (89.250.xxx.xxx) 56(84) bytes of data.
From www1 (89.250.xxxxxx) icmp_seq=2 Destination Host Unreachable
From www1 (89.250.xxxxxx) icmp_seq=3 Destination Host Unreachable
From www1 (89.250.xxxxxx) icmp_seq=4 Destination Host Unreachable
INT tells me:
Code:
int:~# tcpdump host 89.250.xxxxxxx
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 96 bytes
14:39:19.060533 arp who-has INT.... tell www1
14:39:19.061574 arp reply INT is-at 0x:e0:81:74:xx:xx (oui Unknown)
14:39:20.060447 arp who-has INT tell www1
14:39:20.060451 arp reply INT is-at 0x:e0:81:74:xx:xx (oui Unknown)
14:39:21.060403 arp who-has INT tell www1
14:39:21.060414 arp reply INT is-at 0x:e0:81:74:xx:xx (oui Unknown)
also the ARP tables shows following on INT:
Code:
www1 (unvollständig) eth0
where unvollständig should be the MAC from the bond0 of www1...
Everything worked like a charme b4 i changed the ip-adress of WWW1.
Everything is still working between all other Servers, just those 2 make problems and cant communicate.
I dont know if its Switch problem, Network Problem, ARP problem?
Please help and sorry for my english
Sascha
