0

I've got 2x3750 switches with 4 physical links in between. I've bundled all the physical links using PAgP, but I want ONLY 2 links to be active at any single time. How do I go about this?

  • 2
    You could create two separate channel groups out of the four links, but why do you want to limit the number of links? – Ron Trunk Oct 31 '19 at 11:52
  • @RonTrunk That's the specs I've received, asked for clarification but haven't received any answer yet. LACP's standby sounds like the closest thing to what's been requested, but I wasn't sure how to achieve it with PAgP. Could you please elaborate on your solution? I couldn't understand how creating two groups would fix the number of active physical links to 2. – Zac Oct 31 '19 at 12:15
  • 2
    Each group will have two links, so you end up with two etherchannels, each with two links. You can use spanning tree or routing protocols to use one or the other. – Ron Trunk Oct 31 '19 at 12:19
2

That's not possible with PAgP. Traffic is always aggregated across the whole LAG trunk.

Splitting the trunk in two groups of two ports plus spanning tree would work (switching over to the spare trunk when the default trunk's bandwidth decreases due to link loss), but when one physical link in each group fails you end up with two single links and one of them blocked by STP.

LACP could - in theory - support this scenario by limiting the number of active links. While possible in the protocol, I don't think any hardware actually supports that. Usually, the port number limit is hardcoded on a device.

If you think about it there's extremely little point in trying to limit the number of active trunk members.

| improve this answer | |
  • Sometimes it's worth having some capacity in waiting. Logic tells you that if you have 4x1gig available, then get the most from your investment and use all of them. One fails, we just drop down to 3gig right? But if your applications are designed to expect 4gig and it drops they don;t perform "like normal". Same with users - if you're wanging 4 gigs and it drops a couple links your users will complain. But if you design for 2 gig and one link dies, then you bring one of the spares online, you're back to 2 gig - your original throughput design. – A J Rossington Nov 4 '19 at 9:38
  • @AJRossington ... and if we design for 2 Gb, connect 4 Gb and monitor the actual load (to scale up in time) we're less bound to hit a problem like the 2x 1 Gb links I pointed out. ;-) – Zac67 Nov 4 '19 at 16:37
  • I don't follow. Are you saying you design for 2gig, have 4 available but only use 2 until you need a spare? I'm not suggseting we simply don't have all 4 connected, I'm saying keep 2 online and 2 in standby somehow. You could do this programmatically by monitoring the links – A J Rossington Nov 5 '19 at 17:05
  • 1
    So connect, configure all 4 but "put 2 to sleep" until needed. You could do it with SNMP. Configure all 4 ports into the bundle, shutdown 2. Have the monitoring system listen for and interface down event from either of the "live" links and respond by sending an SNMP message to bring one of the standby's online – A J Rossington Nov 5 '19 at 17:07
  • 2
    Logic tells you that if you have 4x1gig available, then get the most from your investment and use all of them. One fails, we just drop down to 3gig right? But if your applications are designed to expect 4gig and it drops they don;t perform "like normal". If you really need 4gig performance, then consider upgrading to 10Gbps links. 4 x 1Gbps LAG gives you good redundancy, not performance. You still limited of 1 Gbps max per session. – Andrey Prokhorov Dec 2 '19 at 9:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.