3750x2 Disable loop guard and the effect on connectivity

keithsauer507 · ‎04-30-2015

Hello,

I have two building about 500 yards apart. They are connected via a Lightpointe FSO (Free Space Optics) Laser and Airbridge (802.11n RF) point to point bridge. The transport up to the head is done via a multimode fiber optic cable.

When the signal level is good, FSO can achieve the full 1gbps link between buildings. When the signal level falls below a definable threshold (in my case 250 mV), the Airbridge 802.11n RF point to point takes over. This runs in the 5 GHz band on a 40 MHz channel, theoretic speed of 144 to 300mbps.

The issue I have is when this failover occurs, the switch port will go into Loopguard_block and the remote building is dead in the water. I happened to be over there today when this occurred and my quick fix was to unplug the fiber and plug it back in. I since changed the thresholds and widened the window for the Laser to RF failover. If the signal is below 250 mV, it fails over to RF. It does not attempt to return to Laser until the signal hits 350 mV. This "deadband" of 100 mV is to prevent flapping. Previously it was a low of 250 and a high of 300. So hopefully this helps.

However if I were to want to disable Loop Guard, can I do it and what kind of ill effects would that have - if any? Would I do this at both endpoints, or just the remote location? The other end is our HQ and it goes right into the core switch stack (a group of 3750 switches).

sh log on remote switch

001056: Apr 30 18:04:12.261: %SPANTREE-2-LOOPGUARD_BLOCK: Loop guard blocking port GigabitEthernet1/1/1 on VLAN0044.

001057: Apr 30 18:04:12.269: %LINEPROTO-5-UPDOWN: Line protocol on Interface Vlan44, changed state to down

001058: Apr 30 18:07:50.839: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEthernet1/1/1, changed state to down

001059: Apr 30 18:07:52.853: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEthernet1/1/1, changed state to up

001060: Apr 30 18:07:52.920: %LINEPROTO-5-UPDOWN: Line protocol on Interface Vlan44, changed state to up

001061: Apr 30 18:43:32.900: %SPANTREE-2-LOOPGUARD_BLOCK: Loop guard blocking port GigabitEthernet1/1/1 on VLAN0044.

001062: Apr 30 18:43:32.908: %LINEPROTO-5-UPDOWN: Line protocol on Interface Vlan44, changed state to down

001063: Apr 30 18:43:32.925: %SPANTREE-2-LOOPGUARD_UNBLOCK: Loop guard unblocking port GigabitEthernet1/1/1 on VLAN0044.

001064: Apr 30 18:43:33.000: %LINEPROTO-5-UPDOWN: Line protocol on Interface Vlan44, changed state to up

001065: Apr 30 20:06:40.829: %SPANTREE-2-LOOPGUARD_BLOCK: Loop guard blocking port GigabitEthernet1/1/1 on VLAN0044.

001066: Apr 30 20:06:40.846: %LINEPROTO-5-UPDOWN: Line protocol on Interface Vlan44, changed state to down

001067: Apr 30 20:06:40.863: %SPANTREE-2-LOOPGUARD_UNBLOCK: Loop guard unblocking port GigabitEthernet1/1/1 on VLAN0044.

001068: Apr 30 20:06:40.913: %LINEPROTO-5-UPDOWN: Line protocol on Interface Vlan44, changed state to up

Physical connectivity:

Cat6 cable solely for powering (not connected to switch - connected to Lightpointe 48vdc power injector)
multimode fiber cable connected to Lightpointe FSO head.
FSO head has a cat 5e data port that passes power and data to a ubiquiti AirOS 5 powered nanostation (they brand this Airbridge).

Something internally in the FSO linkhead shifts the data flow to the ubiquity nanostation (and back) automatically. Regardless of this activity, all data traverses that multimode fiber. We do not handle any type of routing as the Lightpointe unit does it itself. It is supposed to look like an ethernet handoff.

We use VLAN44 (192.168.4.1) as a transport network and the switches are in layer 3 mode.

Leo Laohoo · ‎04-30-2015

However if I were to want to disable Loop Guard, can I do it and what kind of ill effects would that have - if any?

So what happens if you put both links into an EtherChannel?

keithsauer507 · ‎04-30-2015

It's only one fiber link to the Lightpointe point to point bridge. The lightpoint gear on the roof handles the failover and just presents a single Layer 2 link.

What benefits would ether channel do with one link on each side?

I noticed the uplink ports to the point to point gear are NOT in port fast. Spanning tree loop guard is default so it is on those ports.

Leo Laohoo · ‎04-30-2015

One of the biggest benefits of EtherChannel (aka NIC teaming) is when one link fails, the other one takes over immediately.

keithsauer507 · ‎04-30-2015

Oh I agree with you there, but there's only one link / interface to the link head on the roof. The link head itself has its own interface to the RF backup. The Lightpointe company designed it so their head unit takes care of that fail over. It's really presenting me with a L2 handoff similar to what you would get from an ISP metro-e circuit.

keithsauer507 · ‎05-07-2015

I happened to notice our core switch was not spanning tree primary. A switch on our third floor happened to be root spanning tree switch because it had a lower mac address.

I since corrected this. Between that and widening the buffer where the laser to RF linkheads between our building failover and failback, I have not seen this issue since.

I also took out some unused cluster commands on our core switch which cleared the logs about some update size mismatch. Now I have clean log files where its much easier to spot a real issue.

keithsauer507 · ‎05-01-2015

Ok this is from the vendor:

On a Cisco 3750x2 L3 switch, would there be any special port config statements to mitigate this? Should not be needed, but you can turn off Spanning Tree since the redundancy is being handled by the laser / RF combination on the roof.

So if I put spanning-tree portfast on the uplink port at the HQ and at the Remote side, that would effectively turn off spanning tree and prevent the port from going into loopguard? Am I correct? If the hardware on the roof handles 100% of the laser or RF mode of operation, then I don't need the Cisco switch to intervene. Thoughts?