I have an openstack deployment using kolla ansible. I am unable to configure neutron for floating IPs, Where do I start? I have a router which forwards say, 40 different public IPs onto a single high speed port using openflow. This is connected to my openstack controller, which houses neutron too.
Where do I go from here? How do I make the IPs availvable to openstack so it can forward the packets to required instances in Nova?
I'm trying to perform rate limiting using Queues on Openflow. I've been using 3 files. One to create the topology (top.py), one to create the policies (pol.py) and one to create queues (queue.py).
The topology consists of 3 nodes (h1, h2, h3) and 3 switches (s1, s2, s3), with each switch being able to communicate with every node. The default rate limit is set to 1GBps, but I'm trying to limit it to 1MBps using queues.
Top.py contains the default policy between nodes set to 1GBps, which is working fine between h1 and h2
S1Staticflow1 ={'switch':"00:00:00:00:00:00:00:01","name":"S1h1toh2","cookie":"0", "priority":"1","in_port":"1","eth_type":"0x800","ipv4_src":"10.0.0.1", "ipv4_dst":"10.0.0.2","active":"true","actions":"output=2"}
S1Staticflow2 ={'switch':"00:00:00:00:00:00:00:01","name":"S1h2toh1","cookie":"0", "priority":"1","in_port":"2","eth_type":"0x800","ipv4_src":"10.0.0.2", "ipv4_dst":"10.0.0.1","active":"true","actions":"output=1"}
S2Staticflow1 ={'switch':"00:00:00:00:00:00:00:02","name":"S2h2toh1","cookie":"0", "priority":"1","in_port":"1","eth_type":"0x800","ipv4_src":"10.0.0.2", "ipv4_dst":"10.0.0.1","active":"true","actions":"output=2"}
S2Staticflow2 ={'switch':"00:00:00:00:00:00:00:02","name":"S2h1toh2","cookie":"0", "priority":"1","in_port":"2","eth_type":"0x800","ipv4_src":"10.0.0.1","ipv4_dst":"10.0.0.2","active":"true","actions":"output=1"}
However, when I try to use the queue, using set_queue, it doesn't seem to be working.
S1H2 = {'switch': "00:00:00:00:00:00:00:01", "name": "S1H2", "cookie": "0", "priority": "256", "in_port": "1", "eth_type": "0x800", "ipv4_src": "10.0.0.1", "ipv4_dst": "10.0.0.2", "active": "true", "actions": "output=2, set_queue=1"}
S1H1 = {'switch':"00:00:00:00:00:00:00:01","name":"S1H1","cookie":"0", "priority":"256","in_port":"2","eth_type":"0x800","ipv4_src":"10.0.0.2", "ipv4_dst":"10.0.0.1","active":"true","actions":"output=1, set_queue=1"}
S2H1 = {'switch':"00:00:00:00:00:00:00:02","name":"S2H1","cookie":"0", "priority":"256","in_port":"1","eth_type":"0x800","ipv4_src":"10.0.0.2", "ipv4_dst":"10.0.0.1","active":"true","actions":"output=2,set_queue=1"}
S2H2 = {'switch':"00:00:00:00:00:00:00:02","name":"S2H2","cookie":"0", "priority":"256","in_port":"2","eth_type":"0x800","ipv4_src":"10.0.0.1", "ipv4_dst":"10.0.0.2","active":"true","actions":"output=1,set_queue=1"}
Here's my queue implementation:
queuecmd = "sudo ovs-vsctl %s -- --id=@defaultqos create qos type=linux-htb other-config:max-rate=1000000000 queues=0=@q0,1=@q1,2=@q2 -- --id=@q0 create queue other-config:max-rate=1000000000 other-config:min-rate=1000000000 -- --id=@q1 create queue other-config:max-rate=1000000 other-config:min-rate=1000000 -- --id=@q2 create queue other-config:max-rate=512000 other-config:min-rate=512000"
I'm testing using iperf3, but the rates seem the same with or without the queues. How do I fix this?
This question might come off as slightly hypothetical, but I couldn't figure how else to ask it.
The OpenFlow protocol has a feature which lets its rules get deleted when no traffic matches them. This feature is called the "idle timeout".
In a paper, it's stated that typical OpenFlow link discovery is not scalable, because the Controller has to interact with many packets. So I started thinking about a way to continuously poll links without having the controller continuously looking at what was happening. I came up with the following idea:
Detect all new ports (like, every 5 minutes or something), then install rules which idletimeout in 2 seconds, and send exactly one packet on each link. The idle-timeouting rules will make the packet loop on the link.
Apart for the potential for abuse, and the potential to suck up a megabyte per second of traffic with a single packet, the desired behavior should be clear:
If the link is taken down, or there is saturation on the link, then the packet will be lost and the rules will idle-timeout after two seconds. This can be detected with the Controller, and seems more elegant than to continuously send packets to make sure "everything is alright".
My issue is that this would still require the Controller to poll the installed rules on the relevant table, which still feels like an annoying amount of work.
By design, it seems OpenFlow is close to be able to do such messaging, and it seems that by the way OpenFlow is implemented, timeouts can be critically important (although not all of them).
I Installed mininet in Ubuntu system, By default openflow also installing with mininet. But how can I know the installed version of openflow?