[j-nsp] vMX for ESXi

[Robert Hass]

>
> It would seem that vMX is still in the quite early stages.  When I looked
> at it a few months ago, the documentation was lacking lots of details.  In
> the end, I just didn't feel like it was ready for production in the role
> that I wanted to use it in.  The lack of VMWare support at the time was one
> of the main reasons as well
>

After short testing I can confirm same feelings. Got a lot of error
messages like this:

Message from syslogd at vMX-ESXi-01 at Jan  7 14:12:19  ...
vMX-ESXi-01 fpc0 Frame 1: sp = 0xb9236978, pc = 0x871e148

Message from syslogd at vMX-ESXi-01 at Jan  7 14:12:19  ...
vMX-ESXi-01 fpc0 Frame 2: sp = 0xb92369f8, pc = 0x871b483

Message from syslogd at vMX-ESXi-01 at Jan  7 14:12:19  ...
vMX-ESXi-01 fpc0 Frame 3: sp = 0xb9236a68, pc = 0x871ba4f

vFPC is dumping a lot of core dump all the time with really simple
configuration (just plain IP routing):

-rw-r--r--   1 root  wheel  8112272 Jan  7 14:07
core.J-UKERN.mpc.1452175610.31088.gz
-rw-r--r--   1 root  wheel  8114089 Jan  7 14:08
core.J-UKERN.mpc.1452175703.31388.gz
-rw-r--r--   1 root  wheel  8268070 Jan  7 14:11
core.J-UKERN.mpc.1452175843.31701.gz
-rw-r--r--   1 root  wheel  8265175 Jan  7 14:13
core.J-UKERN.mpc.1452175953.32024.gz
-rw-r--r--   1 root  wheel  8086045 Jan  7 14:14
core.J-UKERN.mpc.1452176046.32322.gz
% date
Thu Jan  7 14:15:08 UTC 2016
% date

BUT at the end - we very like the design where vFPC is separate VM. If you
kill VCP (Virtual RE), vFPC is still forwarding traffic. So design with
redundant vFPC is possible. Finally we have ESXi version - I hope this will
help Juniper to correct bugs and provide stable release in next mid-2016 or
even end of 2016.

Rob