Using 3750 and EEM 3.2 to prevent loops with redundant VPLS solution

[d.calton]

Cisco's DCI Design and Deployment Guide Release 2.0 describes an implementation where SP(s) provide a redundant VPLS solution with n-PE, where PWs terminate on a VSS pair at each cust premise.  To prevent L2 loop, Cisco recommends implementing EEM scripts that track the port channel connected to primary nPE, and react by bringing up secondary link when primary fails, as well as recovery script when port channel returns to service.  My read on all of this is that this should be possible using EEM version 3.2 on a 3750 platform as well.  Does anyone have an opinion on this as to whether it can really be done, and would be an acceptable solution?

[rsimoni]

Hi,

I suppose you are referring to the so-called 'Cisco IOS EEM Semaphoring in N-PE' feature described on this white paper

http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/white_paper_c11_493718.html#wp9000295

Are you see the eem script is meant to keep the PW on your backup N-PE down while the PW on the primary N-PE is active and hence avoid a dual-active condition leading to L2 loops.

As such it is configured on the N-PEs only and not on the bridges belonging to the local STP domain.

Since you cannot have AtoM nor VPLS on the 3750 there is no point in configuring eem there.

About VSS the same paper points out that 'The two VSS devices appear as one single device to the directed Label Distribution Protocol (LDP) peer and use dual-homing support similar to MEC. This eliminates the need for Spanning Tree Protocol and Cisco IOS EEM mechanisms to track availability or redundant devices as described earlier.'. Meaning that you can avoid the eem scripts if you use VSS as N-PEs.

Does this answer to your question?

Riccardo

[d.calton]

No, I'm afraid not.  My reference was actually the pdf http://www.cisco.com/en/US/docs/solutions/Enterprise/Data_Center/DCI/DCI2_External.pdf.  This describes the need for customer equipment (not nPE) to arbitrate between the two SP-provided VPLS networks, and to avoid a L2 loop because of that.  Provides two solutions, one with Nx7Ks, which is painfully complex due to the separate control planes of the two units, and on with VSS, which is fairly straight-forward with the unified control plane.  Thanks anyway!

[rsimoni]

Hi,

Ok you are referring to that DCI solution then.

I am more SP oriented and therefore more familair with SP related features (the one I mentioned is for 7600 used as PEs) even though I recently started being involved on DC area activities too.

Anyway the concept is similar with the difference that on DCI we allow active/active scenarios by moving the semaphore logic one step below > aggregation.

Even though I am not an expert on DCI design what I can tell you already is that for A-VPLS the requirement is that you have at Aggregation layer VSS or vPC capable devices. This remove the 3750 from the picture and therefore the question on eem.

However from a therotical point of view I would say that if your aggregation layer doing SVI routing consists in 3750 devices I don't see why you could not configure the eem script on it. Maybe there are some drawbacks I am not aware of (and I cannot possibly think of).

I will leave to somebody else more familiar with DC technologies to comment further.

Also try to repost on Data Center section, maybe you get more luck there.

Riccardo