[c-nsp] OIR on 7600s: Pretty much evil?

[David Sinn]

I think the issue is more complicated then just does it work or not.  It is dependent on how you have your 6500/7600 deployed.

Some form of bus-stalls will occur with any OIR.  They may be minor, they may not and that comes down to how long it takes for the shared bus to re-stabilize because you just changed it's physical properties.  It is a shared bus after-all...

Those of us that have some CFC or even non-fabric cards in the mix, bus-stalls will impact the one way you have to get packets (or just headers) off the LC's and to the PFC.  Bad things can ensue.  Move traffic before hand...  :)

Those of us that have all DFC, well the bus really isn't doing much to begin with so a stall is nominally a non-visable event.

I do have to question Ben's comments about a stall killing interrupts.  At least on the Sup720 the LC-bus is isolated from the backplane bus and so I can't see why the LC-bus would stall too (and experience shows that it's a non-event).  If we are talking about Sup32's or Sup2's without fabrics, the LC-bus is bridged to the backplane bus, so stalls will be visible.

David


On Nov 11, 2010, at 6:55 AM, Geoffrey Pendery wrote:

> I'll second Gert - I've personally performed close to 100 OIRs on a
> variety of 6500 chassis, and never had it cause a problem.
> 
> There was a previous thread almost exactly like this, BTW - if you
> feel like searching the archive.  It was half-filled with "OIR always
> fails, I call it Online Insert and Reboot!" and half-filled with "I've
> never had a problem, ever.  Works like a charm."  A couple people
> explained the bus stall behavior:
> As you're sliding the blade in, it stalls the bus when it first makes
> contact, then releases the stall once it's all the way in.
> 
> My take-away from that thread was that it's a self-fulfilling
> prophecy:  The techs who approach it confidently, expecting no
> problems, slide the blade in quickly and experience none.  The techs
> who are worried and skeptical, slide the blade in slowly and
> cautiously - and their caution leads to a reboot.
> 
> That said, the others in this thread are also correct - if your RP is
> doing stuff on tight timers, especially BFD, even a very short bus
> stall can still be service impacting.  And of course, it's better to
> plan a maintenance window expecting problems and be pleasantly
> surprised than to assume it's no big deal and get hit with an outage.
> 
> 
> -Geoff
> 
> 
> On Thu, Nov 11, 2010 at 2:00 AM, Gert Doering <gert at greenie.muc.de> wrote:
>> Hi,
>> 
>> On Wed, Nov 10, 2010 at 03:01:21PM -0700, John Neiberger wrote:
>>> I'm just curious to hear your thoughts on OIR on this platform. Is
>>> this something that you prefer to avoid? Do you have any OIR-related
>>> horror stories you'd like to share?
>> 
>> On 6500/7600 (and 7200), we *never* had any issues.
>> 
>> On 7500, the "R" in "OIR" translates to "Reboot".
>> 
>> gert
>> --
>> USENET is *not* the non-clickable part of WWW!
>>                                                           //www.muc.de/~gert/
>> Gert Doering - Munich, Germany                             gert at greenie.muc.de
>> fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
>> 
>> _______________________________________________
>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>> 
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/