Rare project developers

[mc36 <>]

v3 iteration after abandoning the "server stacker" and the cli relaying, 
because that way,
the routing instances sitting on different line cards (members of a fabric) 
would look like
different routing entities to the external world, with different lsa/lsp 
db... so i danced
back to the original idea, and seemingly this is it... as the first step, i 
reworked the
p4lang server to be able to accommodate multiple switch configurations, and 
this involves
the capability of multiple connections for each exported asic:
https://github.com/rare-freertr/freeRtr/commit/3d982e8575cbbd765eeb65de00ec7f1c46695535
with the current idea, one will have a single controlling freerouter with all 
the dataplanes
connecting to it's single p4lang server... it'll handle the all the routing 
protocols and
fabric discovery and send out the per asic(/dpdk) instructions to perform the 
programming...

the next steps will be:
- additional knob "dataplanes <num> <remoteip>" to enable all the potential 
in the above change....
- additional knob "backplane-port <swid> <portid>" to select the backplane 
ports...
- a very basic discovery mechanisn over the configured backplane ports 
feeding the spf class...
- and finally, the awareness of this kind of nexthop over a backplane port 
and proper propagation of it to the dataplanes...
if all the above is completed, we'll have a "server p4lang" implementation 
that could be used in a
single-switch, or multi-asic/multi-chassis/mulit-switch, stacking or even dc 
scenarios,
and it'll behave equally nice regardless of the setup....

ps: the new server p4lang is already running on my homenet for 8 hours now 
without a glitch...:)

On 5/12/22 00:32, mc36 wrote:
[..]