[j-nsp] 10.3 on MX960 with MPC only?

[Chris Evans]

I have real concerns with juniper.  We are primarily a Cisco shop and are
using juniper devices here and there.  I have to honestly say, anymore Cisco
code is way more stable than Junos. I'm always finding major bugs in junos,
yet any Cisco bugs we are finding are usually cornercase ones.

The latest junos bug I've found is that with the mx platform igmp snooping
doesn't work at all in 10.0 and other train releases. In 10.3 all mulicast
is broken when using irb interfaces.  Seriously is no one testing?

I hope they get their act together soon....
On Aug 30, 2010 7:57 PM, "Richard A Steenbergen" <ras at e-gerbil.net> wrote:

> On Mon, Aug 30, 2010 at 08:01:50AM -0700, Derick Winkworth wrote:
> > In fact, yes! 10.3 is primarily...
> The story I've heard is that they put the brakes on almost all new
> feature development for 10.3 and I think 10.4, to give them a chance to
> get their systest act together. I'm not sure if halting feature
> development (and switching to a goofy quarterly date-based release
> system :P) is really the right way to fix things, but at least they're
> trying to do something to address the very serious issue of the
> disasterously buggy code they've been churning out lately, so at least
> that's something. :)
>
>
> > I think they are really, really wanting to have another golden release
> > similar to 8.5. Perhap...
> Don't get your hopes up, I don't think there has been any significant
> progress made on the route stall issue. As far as I can tell it appears
> to be a long-standing issue in rpd that Juniper just doesn't have the
> ability or will to fix (at least not at the rate they're losing rpd
> developers).
>
> The root of the issue seems to be that exchanging large numbers of
> routes over BGP will starve something within rpd, to the point that new
> routes cannot be installed to the FIB, and in some cases the box won't
> even start advertising any routes outbound,  while it is still receiving
> new BGP routes. For MX's our best solution was to roll out a layer of
> dedicated route reflector boxes to replace our current IBGP core mesh,
> thus hiding most of the inactive paths from the real routers (e.g.
> bringing it down from 3-4 million paths over ~30 sessions to 330k * 2
> paths from 2 sessions). EX is still pretty bad though, you can stall it
> for 20 minutes with less than a million paths, which is a real problem
> when you have a customer aggregation router which won't advertise
> customer prefixes for 20 minutes following rpd restart.
>
> I'm really wondering how things are coming with SMP-capable JUNOS. I
> tried booting a jinstall64 olive, and while the image isindeed using an
> amd64 kernel it STILL doesn't have SMP enabled. Considering they briefly
> slipped in a line about upcoming Core2 Duo/Quad RE's into some recent
> release notes, I really hope they're gonna have this working soon. :)
>
> --
> 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)
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net...
>