Lag configuration question

[Stephen_McGuire]

I have an S4 (main campus) with a lag on one end and I have three stacked B5's (remote campus) on the other. They are connected via 1 mile long fiber.

Complaint from end user's has been poor internet speeds at the remote campus. They are on a VLAN and come back via the fiber to go out through our internet provider.

I just started this job and am new to the extreme EOS though I just finished the bootcamp and passed the switching and routing exam.

I started to just look at the VLAN performance and noted it was terrible and I'm on main campus side. I'm trying to document the existing switching design etc. So I began to dig into how things are connected L1/L2.

I did notice though that I have two ports on the S4 setup as a LAG to the stacked B5's but on the stacked B5's there are NO LAGs configured. Globally LACP is enabled on both ends but the stack hasn't been configured with a lag.

So first question is, how does this even work? I thought that you had to have both ends configured properly for the LAG to work. I know they can form automatically due to the global setting but I can't find anything on the B5's to indicate that has occurred.

Should I go ahead and setup the lag properly on each end with aadminkey's which one side appears to have already?

Port status shows the ports on the remote end to both be UP, all of the lag's appear as down this is the same as the main campus.

So it looks to me that the lag's are not configured and up properly and that neither end is setup.

Sorry for the long story but wanted to get some feed back on the situation. 🙂

Thoughts and suggestions are appreciated!
-Stephen

[Stephen_McGuire]

Ok, so here's what I found when I was able to get to the location:
B5 has port ge.1.47 connected to the the S4 via fiber to ge.3.112
E1 has port ge.3.1 connected to the S4 via fiber to ge.2.112

There is a link between the E1 ge.2.2 and the B5 ge.1.46 which explains the neighbor info.

Currently the spanning tree info shows the following:
SID Port State Role Cost Priority
-----------------------------------------------------------
0 ge.2.2 forwarding designated 1 128

SID Port State Role Cost Priority
--- ---------- ---------------- ----------- -------- --------
0 ge.1.46 Discarding Alternate 20000 128

So if I understand this correctly, there is no loop but the two fiber runs are not setup in a LAG configuration.

So my question is this, should I place both fiber runs on the B5, and lag them? I have proposed that and included adding a 3rd fiber run to make a single 3GB lag between the B5 and S4 then have the E1 off the B5 stack (there's 3 switches).

Thoughts on that? It's been approved but I could change it. Fiber is already existing and the only financial cost is getting single mode gbic's and a patch cable (under $75 from amazon).

Would like to hear anyone's opinion's on that change in configuration, thanks.
-Stephen

[Erik_Auerswald]

As to why the configuration with a LAG on one end only can "work": EOS automatically falls back to using individual ports if no LACP PDUs are received or there are too few links to form the LAG (e.g. one link w/o singleportlag). Spanning tree will block redundant links and prevent a loop.

The LAG and individual port configurations can differ, so "work" should not be read as "work correctly".

[Jeremy_Gibbs]

So, start with on the B5

show lacp lag.0.*

Does it show any ports in a LAG?
Do you have single port LAG turned on (Enabled)?

Also, run show neighbors to confirm the uplink is what you think it is from both ends... if it is, check the ports for errors...

On the B5 show port counters ge.2.4

on the S4

show port counters errors nonzero

Are you seeing any errors on the ports or the supposed lag port?

Also, on the S4, for that particular problem VLAN. run the command

show run int vlan.0.20

Paste the output here.

[Stephen_McGuire]

I looked at that output several times and completely missed that. LOL

[Erik_Auerswald]

OK, so the mystery device is an old Enterasys E1 switch.

I would assume the E1 does not understand CiscoDP or LLDP and thus floods it to the other ports (it's L2 multicast).

The E1 switch seems to be between the B5 stack and the S4, with other switches connected to it as well.