Bond0 looking for an IP address

RouterOS General
1

Hi all

I have bond0 from (Ether1 and Ether2) in broadcast mode.
Ether1 has a mac address of: D4:CA:6D:97:B7:58
Ether2 has a mac address of: D4:CA:6D:97:B7:59
Bond0 has a mac address of: D4:CA:6D:97:B7:58 (same as Ether1, I suppose this is right?)

I have assigned a static IP 192.168.1.105 to bond0.

It’s setup as a DHCP server bound to bond0, with an appropriate pool.

All day, every day, i get these log entries:

dhcp,warning | dhcp-server offering lease 192.168.1.77 for D4:CA:6D:97:B7:58 without success

The problems:

  1. The router actually freezes all network activity for about 1-3 seconds every say 10 or 15 seconds. (if you watch the graphs… you can actually see big gaps!!)

  2. (related) The router is dreadfully unstable. It’s constantly dropping VPN connections, you can ping though it but the ping replies vary wildly, and every 5th or 6th reply fails.

  3. (related) The router is constantly disconnected from winbox.

The Questions:

  1. What on earth is going on?

  2. Why is the bond0 (or Ether1, is not possible to tell the difference) looking for an IP address from its self when i have assigned an IP to bond0 already, and Ether1 is part of the bond, and none of the other ports are active?

Thanks

2

It sounds like you have ‘/ip dhcp-server alert’ configured for either the bonding interface, or for one or both ethernet interfaces; and that whatever you have connected to the ethernet ports is forwarding between them.

So, broadcast packets emitted from ether1 end up being received by ether2, and vice-versa.

The DHCP alert mechanism seems to work by sending a broadcast DHCP discovery packet, and then not following through on the request. Thus the lack of success in completing the lease.

Broadcast bonding replicates every packet being sent out each member interface, and isn’t something you would usually want to use in conjunction with a switch. Generally, it would only be used with isolated cables/circuits/tunnels/links between routers, and even then it is an unusual mechanism.

Are you sure that broadcast bonding is really what you want to do?

–Eric

3

Thanks for the reply,

There are two cisco switched connected to the two Ethernet ports. The entire networks is setup for redundancy. Its a complete A/B network. Every devices is replicated, and sends each packet on both the A and B network. The A and B networks are connected at the switch(s).

I suppose my only choice now would be to change the bonding configuration for this device, or configure DHCP on Ether3 and use a single cable for just that (dhcp is not mission critical).

4

I would suggest taking a different approach to redundancy; or at least actually isolating the ‘A’ and ‘B’ networks, so that packets from one do not end up on the other.

Duplicating every packet is likely to cause problems, and will certainly make troubleshooting more difficult.

Beyond that, you can probably just turn off DHCP alerts to make the errors go away, assuming that is actually where they are coming from.

Or, since those errors don’t really indicate a functionality problem, you could filter/ignore them.

–Eric