Hi Tobias
Here are perhaps some hints...
-channel-group NUMBER mode on: Is a static channel with NO channel protocol on it. If you speak LACP on the other side the channel won't come up because the neighbors don't agree to a channel. You need a "channel-group NUMBER mode active" on the switch side so it speaks LACP. Set it to this mode and run a debug on the switch so you can see more: debug lacp ? See: (http://www.cisco.com/en/US/docs/switches/lan/catalyst3560/software/release/1...) You can also fix the channel only to speak LACP with "channel-protocol lacp" in combination with "channel-group NUMBER active". See: (http://www.cisco.com/en/US/docs/switches/lan/catalyst3560/software/release/1...)
-switchport nonegotiate: disables DTP (dynamic trunking protocol) which is used to negotiate a trunk between to dtp speaking switches. If you set the port to "switchport mode trunk" on the switch side then you make a trunk on this port. See: (http://www.cisco.com/en/US/docs/switches/lan/catalyst3560/software/release/1...) . So in my point of view you don't need the "switchport nonegotiate" command.
-speed 1000: if you fix the speed to 1Gbit then its always full duplex. There is no 1Gbit with halfduplex as far as I know. So the duplex full is also useless on this port. Generally speaking: If you fix the duplex mode and speed one side please watch out that you also fix it on the other side.
So I would frist watch out that the channel comes up. For trunking and vlans on the trunks you can troubleshoot later.
Cheers
Luki
Am 05.04.2012 16:21, schrieb Tobias Brunner:
Hi everyone,
I've "compiled" all your input, here is the output:
The cables are all cat5e, which should be OK. But to be sure I've managed to find a cat6 cable, but as expected, it did not change anything. So the cables are not the problem.
Then I added "speed 1000" and "duplex full" to the two port configuration, but: no change.
Next step: Try "mode on" -> Does not work because "on" is PAGP which is not compatible with LACP.
Go on: Try "switchport nonegotiate" -> No change in behaviour.
When the servers are booting (BIOS, not yet OS) then there is no bonding/etherchannel available. Both ports are independent. That should be no problem for the "mode active/passive" and it works as expected, the server can do DHCP/PXE during boot. But it takes a long time until the server gets an IP from DHCP because it takes so long to get the protocol up.
Is your server supporting LACP?
During BIOS boot: no While running OS: yes
How are is the bonding configured on the interfaces?
During BIOS boot: not configured While running OS: configured with linux bonding (that works fine)
Maybe this will stay a mysterium =)
I wish everyone good luck in finding the easter-eggs. (Maybe this behavior is a cisco-easter-egg and solved after easter, who knows =)
Cheers, Tobias