[j-nsp] half-duplex fxp0 on MX despite hard-coding speed/duplex

[David Ball]

  This is likely the 3rd time I've come across this on MXes (240s and
480s), and have never seen it on the T-series.  Wondered if anyone
else had run across it.

  With redundant REs in my MXes, I created management interfaces as
follows, then use 'apply-groups [re0 re1]' immediately afterwards,
like so:

# show groups
re0 {
    system {
        host-name myrouter;
    }
    interfaces {
        fxp0 {
            speed 100m;
            link-mode full-duplex;
            unit 0 {
                family inet {
                    address 192.168.101.46/24 {
                        master-only;
                    }
                    address 192.168.101.44/24;
                }
            }
        }
    }
}
<snip same thing for re1, changing only the non-master IP and host-name>

apply-groups [ re0 re1 ];

   My management interface goes to a 100M port on a cisco switch which
is hard coded for 100-full.  For whatever reason, and so far only seen
on the MXes, the management link comes up as 100-half.  Tried
different cables and ports on the switch to no avail.  The only way to
fix it this time was to leave both sides to autoneg, and link came up
as 100-full.  Anyone run across this before?  Again, it doesn't happen
for my t-series (nor any other non-juniper device being managed
through those Cisco switches).  Our standard deployment is to
hard-code speed and duplex (some will argue this), so I'd rather keep
it uniform.......

David