Hi forum... I have a server running RHEL 4 that has sporadic network connectivity outages. I've completely ruled out network issues, as during the outages I can get to other boxes on the same subnet but still can't ping the machine in question. Load averages and other metrics on the machine also don't show anything obvious. I'm thinking it's te interfaces or the Network services.

I have two questions on this:

1) I can't find anything meaningful in the messages log, and am wondering if there's another log somewhere that might indicate what's going on.

2) The machine has both interfaces bonded together. I think there's a possibility that one interface is going/gone bad, but while bonded I can't say for sure. Shouldn't bonding prevent any loss of connectivity if one interface is bad? I plan on unbonding them and giving each interface it's own IP and monitoring them independantly.

Any more help on this would be greatly appreciated...

I used to have a similar problem at the last place where I worked. The network was crap and was the source of the problem I am quite sure. All the windows boxes on the LAN would reconnect whenever disconnected, but my linux box would not. IN the end I wrote a script to monitor the state of the adapter and restart it if the link was dropped.

1) If you are losing the link you should get a message somewhere, maybe the teaming is preventing it from being reported???

2) Adapter teaming / bonding is overrated IMNSHO. If you have any trouble with the LAN the first thing that you should do is turn teaming off and see if the problem remains.

Adapter teaming / bonding can offer performance improvement, and the potential for redundacy exists, but it should be extensively tested because it can all go to **** really really easily. Further many applications do not like interfacing with teamed adapters.