[j-nsp] MX5 BGP Problems (BGP and/or IPFX)

[Giuliano Medalha]

People,

We have read the last messages of this list related to problems with MX80
routers.

We have twi open cases with J-TAC for 2 different routers.

- One of the MX5 routers has the same problem reported on this list: is
takes too slow to start forwarding packets.
  (only 3 full routing table and five partial routing sessions). It tooks
something like 20 to 30 minutos for BGP to announce the downstream prefixes
to upstream routers
  Without IPFX (inline j-flow) configured.

- Another box has some more sessions (1 full routing from iBGP session and
136 BGP sessions (1.5M routes RIB) to provide full routing to downstream
and receiveing 1 or 2 prefixes).
  When the box restart the memory goes to 84%.  It took 30 minutos to start
forwarding all packtes from all sessions.  7 Gbps total traffic.
  The problem is when some routing policy need to be modified ... the first
commit  level up the memory to 89%  ... the second one to 91 ... so on ...
until 94%
  The box starts to flap some sessions and suddenly it looses all sessions
...
  Any other commit on this box must be a dead shot.  We have open a case
and explain to JUNIPER what happening. They have asked us
  When IPFX is not working the situation is  better .... BGP sessions more
stable ... but 94% of the memory after 2 or 3 commits after reboot ...same
situation without BGP session flaps.

I saw that you create a case with a PR number PR 836197 ...  my SE told me
today that it must be addressed to  the next  13.1 release.

My simple question is ... maybe they can solve the RPD process problems ..
.related to BGP convergence too slow.

But what about to have and IPFX problem  ?   Maybe we have another problem
when IPFX is turned ON (in-line j-flow) ... togheter with BGP.

This two problems could have some relationship ?

Thanks a lot,

Giuliano