1

I have a CentOS 6 server with two bonded ethernet ports. Both ports are connected to a switch which I then connect to a laptop.

If I flood-ping the laptop and disconnect one of the bonded ports, some packets (~20) are lost before things stabilise. This happens in a fraction of a second. However, if I then plug the cable back in, I drop packets for 30 seconds before things go back to normal.

Any ideas?

Improve this question

3 Answers 3

Reset to default
2

Bonded how ? Are you using LACP between the server and the switch ?

If not, the devices will have to do normal ARP discovery to re-establish the bonded link; this may well take that long.

Improve this answer
1
  • It's currently using mode 0 / round-robin. I'll try tinkering with different modes if it's more likely to be a server thing than a switch thing. The switch is a Dell PowerConnect, is there likely to be any tweaks needed for that or should it just work?
    – Gaspode
    Feb 20, 2012 at 9:23
0

I had a similar problem using balance-alb, Debian 6.0 and ZyXEL ES-2024A switches. Meaning: way too many variables. If you are using a switch-agnostic approach like balance-alb, can you try using balance-tlb?

That was enough for my purposes, so I stopped my debugging at that point. If not, we'll need MOAR details, as @adaptr said.

Improve this answer
0

I suspect that your switch ports are not configured with spanning-tree portfast. When your server port recovers it will immediately be placed back in the bond but the switch will need to go through 2 x forward delay (30s) before it will resuming forwarding on that port.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .