Problems with Multilink PPP between Cisco and Junipers?

Question

We'd a one of two T-1s in a Multilink PPP bundle go down. Annoying. But this also appears to have somehow "broken" the Multilink itself. Very Annoying.

I'm on the Cisco side: a 3662 running 12.2(23a). I've removed the "bad" serial line from the Multilink, as did a fellow on the Juniper side, but the link didn't return to a working state.

When the broken serial line is fixed, we'll be able to do a little more testing. It *might* only be a coincidence that both the Multilink breaking and the serial line going down occurred at the exact same time. But I've my doubts.

Anyone have any thoughts or ideas?

I've no minimum number of member links set on my side, and the Juniper side had a minimum of one (and we did confirm that the remaining T-1 was working). The Multilink reported itself as up on both sides, but we couldn't ping each other's IPs. I couldn't even ping my own!

Both IPs could be pinged from elsewhere, however.

The configuration for the multilink on my side is pretty basic:

interface multilink 2 description Upstream 2 no ip address ip address a.b.c.d 255.255.255.252 ip nat outside ppp multilink multilink-group 2 ip access-group 101 in ip access-group out-to-world out no ip directed-broadcast no ip unreachables no cdp enable

with a couple of serial interfaces "joining" via:

encapsulation ppp ppp multilink

Any thoughts or suggestions welcome...

Andrew

aunraza · Accepted Answer

Did you try pinging from the inside? Or sourcing the ping on the routerfrom the internal interface?

Hansang Bae · Answer

[snip]Let us know how it turns out.  Might be a Cisco/Juniper thing.

Andrew Gideon · Answer

The T-1 was brought back into service. The Multilink was reestablished on both sides, and all was well. I could ping both my side of the link and the remote site, and traffic moved properly.

I'm leaning slightly towards the Cisco/Juniper "thing" idea, except I'd be hard pressed to describe the details of the thing. I know that I (on the Cisco side) had at one point completely removed the Multilink interface and rebuilt it with the lone working circuit. I believe that this was done on the Juniper side too.

Yet that didn't provide a working Multilink. How could that not, unless one cannot build a Multilink out of a single circuit between Cisco and Juniper?

Another puzzle to me within the larger puzzle is why I couldn't ping my side of the Multilink. How could that not be possible?

- Andrew

Andrew Gideon · Answer

I don't recall doing this. I suspect that it would have worked, but gone "the long way round". That is, the pings to those IPs would have left my "inside" via another egress path and then worked their way back.

I did source the ping to the far side of the Multilink from the near side's IP to no effect. I didn't try sourcing the pings to the near side from anything in particular; I'm not sure what source would have been "interesting".

- Andrew

Rob · Answer

Yes, there is a known issue between Cisco and Juniper doing MLPPP. At least that is what Global Crossing told me when they wouldn't let me do it until Juniper fixed their code. That was back in December. Now, I don't know if this fix was specific for the ISP Global Crossing or public. You'll have to ask Juniper.

BTW, it works fine now. I have several Cisco routers (mostly 2600XM series) doing MLPPP with either two or four T1's to Global Crossing, and they only run Juniper. There were some specific commands that needed to be done, like turn off MLPPP fragmentation on both sides, and an MTU bug which affects only OSPF, but that's it.

-Bob

Problems with Multilink PPP between Cisco and Junipers?

Join the Discussion

Didn't find your answer?