I've been confounded by a problem I ran into when I tried to implement LAG between an EX4600 2-member virtual chassis and a Cisco ASA 5525-X (LACP active). I am looking for an explanation of the problem.
The EX4600 side of the LAG is using Copper SFP transceivers distributed evenly between the physical members of the virtual chassis. The LAG members are configured on corresponding ports and modules on each member switch [ge-0/2/4,ge-0/2/5,ge-1/2/4,ge-1/2/5]. Originally, I used Juniper-compatible transceivers but have since purchased a genuine unit [QFX-SFP-1GE-T] for troubleshooting. Both of the EX4600 member switches were purchased at the same time and each has two EX4600-EM-8F expansion modules. The JunOS version is the same on both switches [14.1X53-D27.3]
When I tried to implement this design, I discovered that the ports on the primary VC member (FPC 0) would not come up. Troubleshooting the problem, I found the issue is unrelated to the LAG configuration. Using Windows servers and unmanaged switches as test components, I was able to locate a single port on the primary VC member (actually on a different expansion module) which would come up with a copper SFP inserted (genuine or compatible). All of the ports I have tried on the secondary VC member (FPC 1) also work. This is true even if auto-negotiation is not explicity enabled on the port. Setting auto-negotiation appears to have no effect. Connected devices always have good link lights but the Juniper side never negotiates the line protocol on the ports which don't work.
I am wondering what could explain this inconsistent behavior. Also, Juniper's documentation for the EX4600-EM-8F expansion module (ominously) states: "Caution: Copper SFP transceivers (1000BASE-T) are restricted to the top four ports or the bottom four ports; fiber SFP transceivers (1000BASE-X) can be used in any of the eight ports. Attempting to stack copper SFP transceivers causes internal damage to the module." I would like to know what kind of internal damage would result from "stacking" transceivers. Is this an electrical problem or does is refer to physical damage to the transceiver receptacle?
I have attached some diagnostic output from the EX4600 in question including the working test port (ge-0/1/0) and non-working test port (ge-0/1/1). Note that the same transceiver is used in each example. The "show chassis hardware" command shows the genuine transceiver installed in FPC 0 PIC 1 Xcvr 1 while the compatible transceivers (which are working in the LAG) are in FPC 1 PIC 2 Xcvr 4-5.
Thank you for your consideration.