[j-nsp] Update on 10.4R9 stability for MX?

[Richard A Steenbergen]

On Mon, May 21, 2012 at 03:43:33PM +0200, Mark Tinka wrote:
> On Thursday, May 10, 2012 01:59:55 AM Richard A Steenbergen 
> 
> > There is a serious issue with MPLS RSVP auto-bandwidth in
> > 10.4R9, which can cause the reservation calculations to
> > be off by quite a bit. The least broken code we've found
> > so far is 10.4S9, I'm surprised they haven't done an R10
> > yet.
> 
> As far back as I can remember, Richard, you've pretty much 
> had this particular issue with almost every major train 
> Juniper have released :-).

I wish it was a single issue, at least then I would have one definite 
thing to bitch about. It's more like a dozen different issues affecting 
the same subsystem, in a dozen different versions of code, over the last 
couple of years. You could say the same thing about SNMP, it's been 
broken about as many times over about as many different revisions 
(including recent ones :/).

10.4S9 does seem to have solved this latest issue, which was causing 
wildly inaccurate reservations. It's almost comical when RSVP tries to 
reserve a several petabit LSP (not much to do with recent JUNOS except 
laugh at this point, right? :P), but it's a lot less funny when nearly 
empty high-priority LSPs get mis-calculated to several gigabits and 
start causing incorrect preemptions...

-- 
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)