Rare project developers

[mc36 <>]

but guys, once finally,
i know a better idea to have the current reading for a given flow, aka 
mpls-path-tracing:
first of all, sorry for the cisco terminology:

interface tunnel-te1
 tunnel destination <remote-pe-lo0>
 tunnel mode mpls traffeng
 ipv4 unnumbered lo0
 no shut

at this point you'll have an unidirectional lsp to the remote pe
each p in the chain allocated a label for it when it propagated the resv 
message backwards the tunnel headend
in the cisco boxes, we have a separate byte/packet counter for each label in 
the lfib
and it's a fresh label when you stopped here so always starts at 0...

then ipv4 route interesting-traffic/24 tunnel-te1

then you have to a lunch

then no ipv4 route interesting-traffic/24 tunnel-te1

and the counters must match in every p unless you have a packet loss 
somewhere...

btw the stuff is also available in the rare dataplanes..

for the low latency, well, if the receiver can do the dedup, then 2 constant 
rsvp tunnels and that's it...




On 8/25/22 11:27, mc36 wrote:
> On 8/25/22 11:25, mc36 wrote:
> > On 8/25/22 11:23, mc36 wrote:
> > > with influx, which imho used in poznan already, so there is a cisco streaming 
> > > telemetry importer for influx,
> > so there is an _official_ cisco 9k mdt importer... the only one in the influx 
> > package... but frederic knows this better...
>
>
> so i'm writing these because we did it together, i mean fl put the effort to 
> show me that there was a bug in the exporter...
> but i swear when i introduced the exporter, it worked fine with the current 
> latest image... :)