Rich
Thanks for your testing.
The conclusion seems to be that, before bringing two or more switches
into a domain,
1) make sure that all switches have the same VLAN configurations (do it
manually)
2) make sure that the revision numbers are the same in all switches
3) Now you can link all switches together
If you don't follow these steps, the results are unpredictable.
Once a domain is stabilized, i.e. all switches in the domain have the
same config, then you can go to any server mode switch, make the desired
changes and have all switches automatically move to the new config.
There is a white paper on CCO,
http://www.cisco.com/warp/public/473/21.html
on Understanding and Configuring VLAN Trunk Protocol (VTP), which males
for good reading. It talks about how to reset the configuration revision
number and all sorts of things.
One interesting passage from the white paper deals with the problems
caused if you don't follow the three steps outlined above. It tells you
how to recover from such a situation.
"SOLUTION:
Quickly reconfigure all of the VLANs on one of the VTP servers."
The word "quickly" made me laugh. That was what I had to do, in 10
minutes before production time started for the network.
Rich Mallory wrote:
>
> Interesting, I didn't realize that the 1900's did not change the rev number
> to 0 when the VTP domain name is changed. Just another case of switch/OS
> version having a different behavior. I'll have to remember that or I'm sure
> it will burn me some day.
>
> Unfortunately, I will only have the 6509's for about 2 more weeks before
> they go into production. So I am busily thinking of any tests to run and
> this was a good one.
>
> Rich
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
> Stefan Dozier
> Sent: Monday, July 09, 2001 8:42 PM
> To: [EMAIL PROTECTED]
> Subject: Re: Cisco 5505 switch puzzle [7:11335]
>
> At 07:37 PM 7/9/01 -0400, you wrote:
> >One thing I never really thought about until I read something the other
> day:
> >
> >If a switch was set to server mode I would have been wary of the
> >configuration revision before connecting it to an existing network.
> >If it was set to client mode I don't think I would have been quite as
> >cautious.
> >No matter whether the switch is client or server, the switch with the
> >highest revision will override the others. You can't change the VLAN
> >configuration, but if the revision is already higher it will override any
> >other switches VLAN config.
>
> Agreed....
>
> >
> >Another thing which seems a little strange - The CP BCMSN book suggests
the
> >following steps before inserting a new switch into an existing domain(cut
> >down a little):
> >
> >1. Issue a Clear Config All
> >2. Power Cycle the switch to clear the VTP NVRAM
> >3. Determine the mode of operation and if Server mode is to be used,
verify
> >that the revision is set to zero.
> >
> >Step one and two seem to be a bit of overkill.
>
> >From memory, the revision was set to zero after each change of VTP mode,
> VTP
> >domain, and a few other things.
>
> Whoa.....I haven't read anywhere that the configuration revision number
> gets reset
> to zero when changing the switch's VTP mode, VTP Domain name, etc.
>
> In testing that theory, unfortunately I'll have to rely for now on my
> trusty 1924 with
> the latest enterprise software installed as a testbed, but with continued
> discussions
> and education afforded by the list, perserverance in the pursuit of the
> knowledge
> required to obtain the ultimate certification, and my ability to keep a job
> that
> right now affords me the opportunity reach a higher level of understanding
> each and
> everyday.....one day I'll be in a environment where 55xx and 65xx are
> available as
> testbeds for scenarios just like this..[can't you just see the envy,
> "Rich"].
>
> Anyway, I'm back from la la land now, and my test doesn't indicate that the
> config rev
> number resets when changing the vtp mode or the vtp doamin name, atleast
> not on the
> 1900 series switches, which I'm well aware are miles apart from the big
> boys, 55xx and 65xx
> series.
>
> After reading how non-volatile it was, I was
> >quite surprised how easy it was to reset it, although this could be an
> image
> >related thing. Anybody any other experiences? Rich?
> >Step 3 seems to go against my previous comment that it doesn't matter what
> >mode the switch is in. Why take more care for Server Mode if Client will
> >have the same disastrous effect?
>
> Agreed, client mode is just as dangerous as server mode if the config rev
> number
> is higher and maybe even moreso for the casual implementer who could
> possibly
> surmize that in client mode the switch couldn't perform changes in the VTP
> domain.
>
> >If we carry on like this we may actually get a topical thread to the same
> >length as a troll thread or a thread discussing whether piracy is a good
or
> >bad thing. :-]
>
> nooooo, say it ain't so!
>
> Stefan
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=11759&t=11335
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
