05-25-2012 12:16 PM
I have two ex4200 stacks that are on completely different networks for different uses. One has only a single vlan, 17. The other has several, but I only want to use vlan 10. Both of these stacks are connected to a dell m6220 on the back of a n m1000 blade server. The idea is to have a virtual machine be on both networks. The m6220 does not route the traffic between the vlans. The m6220 can ping the stack on vlan 17 perfectly fine, but it can't ping the stack with vlan10. HOWEVER, if I shutdown the port connecting to the vlan 17 stack, the ping will start working to the vlan 10 stack. While this might seem like an m6220 question, it isn't. I hooked another ex4200 to the m6220 instead of the vlan 10 stack. I gave it very basic configuration. The m6220 can ping it and see it along with the other stack no problem. It looks like to me the vlan 10 stack must have some kind of configuration that won't let it send or receive traffic to the m6220 when the other stack is actively communicating with it. I'd post the config, but it's sensitive information. Does this situation sound like anything anybody would know about? Thanks!
05-25-2012 12:18 PM
By the way, since I'm only using one vlan on both stacks, I'm not using trunks. I tried using them thinking something about them that is beyond my knowledge might make it work, but same situation.
05-25-2012 12:41 PM
Here is a little diagram I threw together. I can ping x.x.x.1 all day long from the m6220 when both are connected. I can't ping x.x.x.4 though. If I disconnect x.x.x.1, suddenly pings start flowing through to x.x.x.4. As soon as I reconnected x.x.x.1, traffic stops to x.x.x.4.
05-25-2012 08:25 PM
05-26-2012 11:39 AM
No sir, they are two fully separate networks. There is no routing between the two VLANS either. Other than this single virtual machine, nothing else from either network is to touch the other network.
05-29-2012 04:43 AM
05-31-2012 08:36 AM
Alright, I have another diagram, this has snippets of my config with generic information put in. The virtual machine can talk to the 192.168.2.x network without issue when everything is hooked up. It can not talk past the m6220 on the 172.16.2.x network WHILE the 192.168.2.x network is active. If I remove the link to the 192.168.2.x network, data starts flowing from the juniper stack on the 172.16.2.x network. I hooked wireshark up, and it looks like the juniper stack on the 172.16.2.x network receives arp requests, but doesnt respond while the 192.168.2.x network is connected. Anymore thoughts?
05-31-2012 08:54 AM
Could you please include routing information from all 4 devices? Looking for 192.168.2/23 , 172.16.2/24, and 0/0 routes while your 192.168.2 connection is plugged in and when it's unplugged.
Where did you connect Wireshark?
05-31-2012 09:28 AM
192.168.2.x ex4200 stack:
0.0.0.0/0 *[static/5] 12w1d17:42:31
> to x.x.x.6 via ge-3/0/13.0
[OSPF/150] 6w6d 19:31:01
Metric 69, tag 1722
>to x.x.x.6 via ge-3/0/13.0
C 172.16.2.0/24 [0/1] directly connected, V110
172.16.2.0 ex4200 stack:
> to 172.16.2.1 via vlan.10
That is the routing information before and after removing the link to 192.168.2.x. Wireshare was plugged in to another port on the 172.16.2.x ex4200 stack that was analyzing both the ingress and egress of the port uplinked to the m6220. Just to clarify, I'm not interested in routing between these vlans. Just want the one VM to be dual homed. The m6620 purposely doesn't have an ip for its vlan17 because it doesn't need to have one, and its management ip is in-band and on the same network, so it won't let me give another interface on that subnet anyways. I'll remove the ip for vlan10 on the m6220 when I am all done, it only has one right now so that I can ping. Only layer 2 stuff is happening.
05-31-2012 09:40 AM
Both EX4200 stacks have the same IP subnet configured -- one on VLAN 17, the other on VLAN 10. Typo ??
You did not include routing information from your virtual machine.
With the routing tables as depicted, there's no way for you to ping anything on 192.168.2/23 from the M6220 however things may be connected.
Basically what I'm trying to see is what your connected routes are, and where the default routes go.
05-31-2012 10:18 AM
You are correct, the m6220 can't ping 192.168.2.0 because it isn't layer 3 on that network. I was pinging the vswitch on the blade where the vmlives from the 192.168.2.0 network. To access it, the pings have to traverse the m6220. Here is the routing information from the VM. It is only for when 192.168.2.0 is linked, because when it is unlinked, I can't access the machine. I left out some of the loopback entries.
Network Destination Netmask Gateway Interface Metric
0.0.0.0 0.0.0.0 192.168.2.1 192.168.2.15 10
172.16.2.0 255.255.255.0 172.16.2.32 172.16.2.32 10
192.168.2.0 255.255.254.0 192.168.2.15 192.168.2.15 10
05-31-2012 10:31 AM
What I meant was that both ex4200 stacks have IP addresses on the 192.168.2/23 network, so I was wondering if that was intended, or if it was a typo.
Of the captured ARP packets, did you find the packets to be going to the correct stack (based on the destion IP), and do the source IP and MAC addresses belong to the subnet (and VLAN) that the wiresharked switch is handling?
05-31-2012 12:04 PM
I'm sorry, total typo, that should read 172.16.2.0 on the corresponding stack. I'm writing stuff down then retyping it Also, wireshark was set to listen to the uplink port on the 172.16.2.0 stack, and it sees the ARP requests. Specifically, I was trying to ping 172.16.2.1 (the gateway) from the m6220, and the ARP packets were trying to determine where that was. However, after receiving the query, the stack didn't respond. However, as I'm sure you are familiar with, if I unplug the uplink to the 192.168.2.0 stack, it will begin to respond.
05-31-2012 12:15 PM
This is getting more and more interesting ...
Where is 172.16.2.1? I see the EX4200 stack is .4, so I am guessing the .1 default gateway is somewhere further upstream. What does its routing table look like ?