[j-nsp] JUNOS 9.6R3.8 on MX

[Richard A Steenbergen]

On Mon, Mar 15, 2010 at 02:21:57PM -0400, Eric Van Tol wrote:
> Hi all,
> Any experiences with 9.6R3.8 on MX?  I have 9.5R4.3 installed on some
> newly acquired MX boxes, per the advice of this list.  However, I
> greatly require the ability to do bridge-domains within
> logical-systems and it appears this is supported starting with 9.6. 
> We plan on utilizing our MX boxes primarily for core routing and
> distribution, as well as MPLS P-routers - nothing too magical.  Is
> 9.6R3.8 just a recipe for disaster?

We've been playing with 9.6R3 on a couple routers with low risk of
customer impact, but 9.5R4 is still our weapon of choice since it is a
known element and has been relatively stable. The weirdest thing I've
seen so far under 9.6R3 was a semi-persistent BGP flap when talking to a
SRC2 IOS box, due to a BGP protocol error. Of course SRC2 is pretty old
and nasty, and has plenty of its own bugs to deal with, but this box
does have a 1.5y uptime and never experienced this issue prior to
upgrading the Juniper it was talking to via IBGP (upgraded from 9.5R3,
plus several versions of junos prior to that).

Mar 1 02:22:17.735 re1.router rpd[1281]: %DAEMON-4:
bgp_read_v4_message:8655: NOTIFICATION sent to x.x.x.x (Internal AS
xxxx): code 1 (Message Header Error) subcode 2 (bad length) value 65535, 
Reason: peer x.x.x.x (Internal AS xxxx): invalid message header 
length 65535 count 4011 errno 0 
Mar 1 02:22:17.738 re1.router rpd[1281]: 
%DAEMON-4-RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer x.x.x.x
(Internal AS xxxx) changed state from Established to Idle (event 
TransportError) 
Mar 1 02:22:28.128 re1.router rpd[1281]: %DAEMON-4: bgp_pp_recv: 
peer x.x.x.x (Internal AS xxxx) sent unexpected extra data, probably insane 
Mar 1 02:22:28.130 re1.router rpd[1281]: %DAEMON-4:
bgp_read_message: peer x.x.x.x (Internal AS xxxx): Update arrived,
expected KeepAlive or Notification 
Mar 1 02:22:28.131 re1.router rpd[1281]: %DAEMON-4: 
bgp_read_message:1979: NOTIFICATION sent to x.x.x.x (Internal AS 
xxxx): code 5 (Finite State Machine Error) 

It flapped like this several times after the initial upgrade, but
eventually settled down and has been stable since. JTAC wasn't
immediately helpful in determining if it was a known issue or not, and
we have far to many bigger more important bugs to deal with as it is, so
this got stashed on the back burner.

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)