Installation Guide ODL Based Router 16A - onfsdn/atrium-docs GitHub Wiki
This virtual machine for the distribution has an implementation for a BGP peering router bundled with OpenDaylight (based off Lithium). This distribution also has the code for flow objectives driver for the OVS 2-Table reference pipleine and the NoviFlow Open Flow switch. To start using the Open Daylight Distribution of Atrium Release 2015/A, download the distribution VM (Atrium_ODL_2016_A.ova) from here: size ~ 3GB
login: admin
password: bgprouter
NOTE: This distribution VM is NOT meant for development. Its sole purpose is to have a working system up and running to use as a starting point for test/deployment as painlessly as possible.
Once you have the VM up and running, the following steps will help you to bring up the system.
You have two choices:
A) You can bring up the Atrium Router completely in software
B) Or you could bring up the Atrium Router using supported switch hardware
You can bring up the Atrium Router completely in software, completely self-contained in this VM. In addition, you will get a complete test infrastructure (other routers to peer with, hosts to ping from, etc.) that you can test with using the router-test.py script. Note that when using this setup, we emulate hardware-pipelines using an OVS software switch (2 Table Pipeline).
Following are the steps required to bring up the test topology:
- Start the ODL controller from the using the distribution VM. ODL based Atrium
codebase can be found in path
/home/admin/atrium-odl. Launch the controller.
cd atrium-odl/
admin@atrium:~/atrium-odl$ ./distribution-karaf/target/assembly/bin/karaf
- Install the feature
odl-atrium-all. This feature will bringup the required components to make stack working and install BGP Application and DIDM (Device Identification and Device Management).
opendaylight-user@root>feature:install odl-atrium-all
Check the logs using command log:tail from OpenDayLight console
2016-02-14 01:30:28,022 | INFO | config-pusher | Bgprouter | 296 - org.opendaylight.atrium.bgprouter-impl - 1.0.0.SNAPSHOT | BGP Router started
Check if all the key features required are running properly
opendaylight-user@root>feature:list | grep atrium
atrium-thirdparty | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :Atrium : thirdparty
atrium-utilservice | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: atrium-utilservice
atrium-bgpservice | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: bgpservice
atrium-routingconfig | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: routingconfig
odl-didm-all | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: didm
atrium-hostservice | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: hostservice
atrium-routingservice | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: routingservice
atrium-atriumcli | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: atrium-cli
odl-atrium-all | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: bgprouter
opendaylight-user@root>feature:list | grep didm
odl-didm-identification-api | 0.2.0-SNAPSHOT | x | odl-didm-0.2.0-SNAPSHOT | OpenDaylight :: didm identification :: api
odl-didm-identification | 0.2.0-SNAPSHOT | x | odl-didm-0.2.0-SNAPSHOT | OpenDaylight :: didm identification
odl-didm-drivers-api | 0.2.0-SNAPSHOT | x | odl-didm-0.2.0-SNAPSHOT | OpenDaylight :: didm drivers :: api
odl-didm-all | 1.0-SNAPSHOT | x | odl-atrium-1.0-SNAPSHOT | OpenDaylight :: didm
odl-didm-mininet | 0.2.0-SNAPSHOT | x | odl-didm-0.2.0-SNAPSHOT | OpenDaylight :: DIDM Mininet reference driver
opendaylight-user@root>feature:list | grep openflowplugin
odl-openflowplugin-all | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: All
odl-openflowplugin-southbound | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: SouthBound
odl-openflowplugin-flow-services | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: Flow Services
odl-openflowplugin-nsf-services | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: OpenflowPlugin :: NSF :: Services
odl-openflowplugin-nsf-model | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: OpenflowPlugin :: NSF :: Model
odl-openflowplugin-flow-services-rest | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: Flow Services :
odl-openflowplugin-flow-services-ui | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: Flow Services :
odl-openflowplugin-drop-test | 0.2.0-SNAPSHOT | | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: Drop Test
odl-openflowplugin-app-table-miss-enforcer | 0.2.0-SNAPSHOT | | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: Application - t
odl-openflowplugin-app-config-pusher | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: app - default c
odl-openflowplugin-app-lldp-speaker | 0.2.0-SNAPSHOT | x | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: app lldp-speake
odl-openflowplugin-app-bulk-o-matic | 0.2.0-SNAPSHOT | | openflowplugin-0.2.0-SNAPSHOT | OpenDaylight :: Openflow Plugin :: app bulk flow o
- Now, start the mininet topology using the "router-test.py" script. It will create topology as described in the above topology diagram.
admin@atrium:~$ sudo ./router-test.py &
admin@atrium:~$ ps aux | grep mininet
root 3896 0.0 0.0 21144 2140 pts/2 Ss+ 01:35 0:00 bash --norc -is mininet:c0
root 3903 0.0 0.0 21144 2140 pts/4 Ss+ 01:35 0:00 bash --norc -is mininet:bgp1
root 3909 0.0 0.0 21144 2140 pts/5 Ss+ 01:35 0:00 bash --norc -is mininet:host1
root 3911 0.0 0.0 21144 2140 pts/6 Ss+ 01:35 0:00 bash --norc -is mininet:host2
root 3913 0.0 0.0 21144 2140 pts/7 Ss+ 01:35 0:00 bash --norc -is mininet:peer1
root 3915 0.0 0.0 21144 2140 pts/8 Ss+ 01:35 0:00 bash --norc -is mininet:peer2
root 3919 0.0 0.0 21144 2140 pts/9 Ss+ 01:35 0:00 bash --norc -is mininet:root1
root 3924 0.0 0.0 21144 2144 pts/10 Ss+ 01:35 0:00 bash --norc -is mininet:router
root 3927 0.0 0.0 21144 2140 pts/11 Ss+ 01:35 0:00 bash --norc -is mininet:s1
admin 4198 0.0 0.0 11708 668 pts/1 S+ 01:36 0:00 grep --color=auto mininet
- The BGP peering application uses two configuration files (addresses.json and sdnip.json) to install the correct flow rule for any topology. "addresses.json" holds the port level details for the OF switch and "sdnip.json" holds the information required for the router peering.
Configuration file can be found at following path:
"/home/admin/atrium-odl/distribution-karaf/target/assembly/configuration/initial"
The configuration files can be modified at runtime using the "restconf".
- Controller will install the flows to control plane and data plane OVS switch to enable the BGP peering. Now the BGP peers will exchange the routes with each other. The BGP application on the controller learns the routes from the control plane Quagga using i-BGP connection and pushes the flows to data plane switch for packet forwarding using the OVS 2-Table pipeline driver.
Verifying the flow installation:
Flows in control plane switch:
$ sudo ovs-ofctl dump-flows s1 -O OpenFlow13
OFPST_FLOW reply (OF1.3) (xid=0x2):
cookie=0x0, duration=1901.409s, table=0, n_packets=1203, n_bytes=94173, tcp,tp_dst=179 actions=CONTROLLER:65535
cookie=0x0, duration=1901.409s, table=0, n_packets=1228, n_bytes=95786, tcp,tp_src=179 actions=CONTROLLER:65535
cookie=0x0, duration=1901.408s, table=0, n_packets=2, n_bytes=88, arp actions=CONTROLLER:65535
cookie=0x0, duration=1901.409s, table=0, n_packets=0, n_bytes=0, icmp actions=CONTROLLER:65535
Flows in data plane switch: Table 0
admin@atrium:~$ sudo ovs-ofctl dump-flows router -O OpenFlow13 table=0
OFPST_FLOW reply (OF1.3) (xid=0x2):
cookie=0x0, duration=1983.152s, table=0, n_packets=0, n_bytes=0, ip,in_port=1,dl_vlan=100,dl_dst=00:00:00:00:00:01 actions=pop_vlan,goto_table:1
cookie=0x0, duration=1983.151s, table=0, n_packets=0, n_bytes=0, ip,in_port=2,dl_vlan=200,dl_dst=00:00:00:00:00:02 actions=pop_vlan,goto_table:1
cookie=0x0, duration=1983.151s, table=0, n_packets=1228, n_bytes=96392, priority=65535,ip,in_port=1,dl_vlan=100,dl_dst=00:00:00:00:00:01,nw_dst=192.168.10.101 actions=CONTROLLER:65535
cookie=0x0, duration=1983.151s, table=0, n_packets=1236, n_bytes=96963, priority=65535,ip,in_port=2,dl_vlan=200,dl_dst=00:00:00:00:00:02,nw_dst=192.168.20.101 actions=CONTROLLER:65535
cookie=0x0, duration=1983.151s, table=0, n_packets=2, n_bytes=92, arp actions=CONTROLLER:65535
Flows in data plane switch: Table 1
$ sudo ovs-ofctl dump-flows router -O OpenFlow13 table=1
OFPST_FLOW reply (OF1.3) (xid=0x2):
cookie=0x0, duration=1972.617s, table=1, n_packets=0, n_bytes=0, priority=180,ip,nw_dst=1.0.0.0/16 actions=write_actions(group:1)
cookie=0x0, duration=1971.616s, table=1, n_packets=0, n_bytes=0, priority=180,ip,nw_dst=2.0.0.0/16 actions=write_actions(group:2)
Flows in data plane switch: Group Table
$ sudo ovs-ofctl dump-groups router -O OpenFlow13
OFPST_GROUP_DESC reply (OF1.3) (xid=0x2):
group_id=1,type=indirect,bucket=weight:0,actions=set_field:00:00:00:00:20:01->eth_dst,set_field:00:00:00:00:00:02->eth_src,push_vlan:0x8100,set_field:4296->vlan_vid,output:2
group_id=2,type=indirect,bucket=weight:0,actions=set_field:00:00:00:00:10:01->eth_dst,set_field:00:00:00:00:00:01->eth_src,push_vlan:0x8100,set_field:4196->vlan_vid,output:1
In Opendaylight console you should see the learnt routes.
opendaylight-user@root>atrium:fib
Type : UPDATE Next Hop Ip : 192.168.10.1 Prefix : 1.0.0.0/16 Next Hop Mac : MacAddress [_value=00:00:00:00:10:01]
Type : UPDATE Next Hop Ip : 192.168.20.1 Prefix : 2.0.0.0/16 Next Hop Mac : MacAddress [_value=00:00:00:00:20:01]
- Now we are ready to test ping from one host to another.
Using the mininet utilities login to host1(1.0.0.1) and ping to the host2(2.0.0.1).
$ ./mininet/util/m host1
to enter host1 to ping 2.0.0.1 (host2)
Veriy that ping is successful !
NoviFlow is currently the only hardware Atrium driver that is supported. The Noviflow hardware platform uses the OVS-2TP (mininet) driver to install the learnt routes from control plane quagga.
To configure the Atrium with the hardware openflow switch, make the required changes in configuration file (sdnip.json and addresses.json) according to topology.
Driver for the hardware platform is getting selected based on the the manufacturer name
and switch model identifier reported by the OpenFlowPlugin module. So after registering
the open flow switch to the driver all the requests from application is routed to the appropriate
driver using the routed-rpc mechanism.
- The distribution needs to be optimized to get better performance
- Some time the controller goes into deadlock condition, in such cases we need to rebuild
distribuition-karaf artifact (using
mvn clean install -DskipTests=True) - Flows are not getting deleted in the OVS 2-TP driver (reported)