06-29-2009 01:42 AM - edited 06-29-2009 02:09 AM
I'm testing a WXC architecture for which the first one which is in a trunk does no do anything, whereas the one at the other end (inline) is working well.
Let's first get in the technical details:
Vlan 8 and 9 are passing through WXC A, and WXC B and C are at remote sites, B destination for the vlan 8, C for the vlan 9 (both of them are using different routers and leased lines, and arrives at remote site without tagging).
The WXC has an IP address on vlan 8.
The WXC B is not up yet, but A and C are.
Lan behind WXC are correctly advertised, I used the " configure reduction-subnet set wan-reduction-subnet on " on the device on the trunk.
Tunnels are up.
On the device C, I can see compression to the network behind A, but stats on A's monitor are empty.
One thing I'm not sure about: what is used the "native vlan" for ?
Any idea ?
Update: in the device A, in Monitor->Compression, there's no stats at all. If a check the "Show passthrough data", there is only stats for the "Data In + Passthrough" line.
06-30-2009 11:10 AM
You probably read this from the Operator's Guide, but just in case:
Native VLAN ID. Enter the default VLAN ID (1 through 4095) used for
untagged frames in the VLAN environment where the WX or WXC device is
One quick thought - Have you checked the remote routes on WXC-A and make sure it has
subnets advertised by WXC-C?
Could you attach a network diagram too? (more interested in WXC-A Lan side, what's connected
on the Local / Remote interface etc..)
And how is WXC-A configured? (in terms of VLAN)
06-30-2009 01:09 PM - edited 06-30-2009 01:22 PM
In fact, both of my links are tagged.
Does it works only if one if the links is untagged ?
06-30-2009 02:46 PM
Here it is:
The trunks are in brown, and both link that should be optimized are tagged (there's no native vlan without tag). The fact that the WXC are in HA has no impact on the pb, let's forget the slave one.
On the local site both links comes from the same FW, and on the remote side, they are going to different routers, going to different location.
07-02-2009 12:52 PM
a. Let's assume this is pure L2 VLAN, and that indeed you have the 2 trunks as shown.
How is the "decision" made to use either link? SPT?
b. Let's assume that traffic goes over the link connected to the Master. It will then hit
the CPE router for VLAN9 (or VLAN90 on the drawing). And then? Is there does it "leak"
to VLAN8 so that WXC-A optimizes it?
Traffic might just leave CPE router un-optimized, go over the WAN, and come back optimized
because return traffic is tunneled?
Sorry if I ask simple questions and actually don't really solve your problem
07-02-2009 01:54 PM
If it is simple question, I hope that my anwser will not be complicated
For question A, I first tried with spanning tree, but now switched to EAPS (an Extreme Networks proprietary solution that converges in les than 1 sec).
B: Hum, I'm not sure to really understand what you wrote.
To what I've understood about how the "trunk mode" works in WXC is that the WXC should intercept data from the vlan 90 and re-inject it in the tunnel (so from th IP address in vlan 80). Yes the tunnel will go back to the local router, and it will repass trough the WXC in vlan 90 in the way to the remote site.
Is it correct ?
07-10-2009 06:55 PM
I noticed this tread seams to have stopped and with out what I see as a resolution.
Please let us know if you need additional help to resolve the problem you had.
Technical Marketing Manager - Access and Acceleration Business Unit
07-28-2009 02:57 AM
After spending some time with the JTAC, and other Juniper guys, we were unable to found a solution and we arrived to the conclusion that this architecture does not work :/
We have to start over by adding an new router, and a layer 3 network.