[j-nsp] inline-jflow on MX MPC (Trio) - experiences?

[Saku Ytti]

On (2012-11-20 16:54 +0100), Sebastian Wiesinger wrote:

Just started with IPFIX export on two nodes this monday.

> 2) In Douglas Hanks Juniper MX Series book it is noted that the
>    sampling rate for inline jflow must always be 1 (other rates are
>    not valid). Still it seems to work with rate 1000 for example (and
>    this is also used as example on the Juniper website).

Noticed this also, we're sending 'rate 500' and it is working, as Arbor is
showing correct traffic levels and it is multiplying by 500.

> 3) The test collector is reporting missed flows. I'm not sure if that
>    is a problem with the collector or if I'm really missing flows.
>    Anyone else had this problem?

Configured IPFIX on two boxes (same peer, same hardware, same software,
same configuration, only difference is, working has unidirectional traffic,
non-working bidirectional). One box stop exporting after 1-2h. If I look at
'show jnh 0 sample-inline statistics' I only see 'Flow insert Policer
Drops' incrementing. If I reload PFE, it'll again work for some time.

Dunno yet if it is configuration problem or PR or what, bangalore JTAC is
on it, so I'm sure it'll be solved in no time at all .

-- 
  ++ytti