| 376 | == Running multiple !OpenFlow devices using one controller. 10/30 == |
| 377 | Two IP8800's have been connected to eachother through a trunk connection allowing VLAN 100 and 28. Two instances of the controller were launched on SB9, along with virtual switch instances on the IP8800s. Here's the configs: |
| 378 | |
| 379 | === SB9 === |
| 380 | controller instances started with assive connections: |
| 381 | * vtcp:6633 |
| 382 | * vtcp:6634 |
| 383 | === SW-SB09 === |
| 384 | on interface VLAN 100 (IP addr. 172.16.100.10/24): |
| 385 | * Virtual switch on vlan 28 on ports 25-36 |
| 386 | * OFP interface on 0/37, listening to port 6633 |
| 387 | === CS switch === |
| 388 | on interface VLAN 100 (IP addr. 172.16.100.30/24): |
| 389 | * virtual switch on vlan 888 on ports 0/1-0/4 |
| 390 | * vlan 28 on ports 25-36 |
| 391 | * OFP interface on port 0/48, listening on port 6634 |
| 392 | * CentOS NetFPGA host connected to port 0/1 |
| 393 | |
| 394 | VLAN 888 was used to test if hosts can connect through the virtual switch on the new OF switch. VLAN 888 was downlinked to the winlab network to handle production data - except, configuration changes to te CS switch were not saved, allowing ORBIT network traffic and backbone to be visible to the WINLAB side through the switch with a trunk conection to ORBIT. While not crippling, this did allow for DHCP messages from ORBIT to travel terrifyingly convoluted paths to still cause complications on the WINLAB network. |
| 395 | |
| 396 | moral of the story: double check configuration changes. |
| 397 | |
| 398 | |