[j-nsp] Topology failure on EX4200

Roger Wiklund roger.wiklund at gmail.com
Mon Jul 17 06:34:13 EDT 2017


Hi

I'm not sure what happens exactly without show output during the failure.

I was just guessing that you need loop protection because you have a
failure of an intermediate device, so interface is still up but you
stop receiving BPDUs, and thus transition from blocking to forwarding
on your specific mgmt VLAN, and creating a black hole. Again just a
guess I might be wrong.

EX4200 and EX4550 have dedicated high speed VC ports, but you can use
your 10G optical interfaces instead.
Then you will have all active/active and can skip STP all together.

https://www.juniper.net/us/en/local/pdf/implementation-guides/8010018-en.pdf

/Roger


On Sun, Jul 16, 2017 at 5:17 PM, Victor Sudakov <vas at mpeks.tomsk.su> wrote:
> Roger Wiklund wrote:
>> > There is a ring of EX4200 switches, please look at http://noc.sibptus.ru/jun1.png
>> >
>> > If MUX1 fails, the MSTP topology adjusts and the PCs continue to see
>> > one another just fine.
>> >
>> > However, some switches become inaccessible in the management vlan
>> > (vlan3 in this example). For example, you can still ping 192.168.1.3
>> > from 192.168.1.2, but not 192.168.1.4 from 192.168.1.2.
>> >
>> > One important note. If MUX1 fails, the corresponding interfaces on
>> > 192.168.1.2 and 192.168.1.4 don't go down, it is only the traffic
>> > (including BPDUs) that stops flowing through the mux.
>> >
>> > If I shutdown the corresponding interfaces on 192.168.1.2 and
>> > 192.168.1.4 (or use OAM to shutdown the interfaces automatically when
>> > the mux fails), the problem disappears and I can ping any switch from
>> > any switch.
>> >
>> > What's the theory behind this?
>> >
>> > "clear arp" and "clear ethernet-switching table" don't fix the
>> > problem.
>>
>> Have you configured loop protection?
>> https://www.juniper.net/documentation/en_US/junos/topics/example/stp-loop-protection-qfx-series.html
>
> Dear Roger,
>
> I have configured OAM (link-down on link-adjacency-loss) which helped.
>
> But I'm wondering why the situation in question was happening without
> the interfaces going physically down.
>
>>
>> On a design note, why not use Virtual Chassis instead?
>
> Sorry, I don't understand. Virtual Chassis uses a special short cable,
> doesn't it? My switches are not located nearby enough to use that
> cable. In fact, they are connected by multiplexers and can be far
> away from one another.
>
>
> --
> Victor Sudakov,  VAS4-RIPE, VAS47-RIPN
> AS43859


More information about the juniper-nsp mailing list