Tushar, A few observations:
1. It appears that the current implementation is not memory scalable. I say this because there are a whole bunch of port specific flows when the L2Switch starts. I'd replace these with a NORMAL + SendToController flow. The reactive flow overrides the NORMAL flows with MAC to MAC flows.That can mostly port over unmodified. 2. The current L2Switch does not support VLAN (I suppose one could configure OVS to support vlans so one would not need to support it natively in the L2Switch). I hacked a few things together to support VLAN's using flow entries and I also utilized some of the reactive flow part of L2Switch so if you decide to run with this project I can point you to relevant pieces of code to look at. A nice feature to have would be to make l2-switch selective i.e. be able to say which switches it will manage. Another nice feature to add would be to integrate with genius so it can coexist in the pipeline along with other apps. Ranga On Thu, Dec 6, 2018 at 6:10 PM Tom Pantelis <tompante...@gmail.com> wrote: > > > On Wed, Dec 5, 2018 at 1:10 PM Luis Gomez <ece...@gmail.com> wrote: > >> Hi Thushar, see my answers below... >> >> On Dec 2, 2018, at 10:20 PM, Thushar Prakash <thushar.m.prak...@gmail.com> >> wrote: >> >> Hi, >> >> I am new to ODL and I couldn't find a users mailing list. I hope this is >> the right one. >> >> I was trying to do an experiment with the ODL controller and mininet. I >> needed reactive flows and since l2switch was not included in the latest >> fluorine release I used Carbon instead, for now. >> >> >> FYI oxygen is last release with l2switch. >> >> This is the topology that I am using for my experiment. >> >> <02-hostsfound.png> >> But the flow table entries for the OVS switch in mininet was (for switch >> OpenFlow:1): >> >> >> NXST_FLOW reply (xid=0x4): >> cookie=0x2b00000000000002, duration=30.63s, table=0, n_packets=0, >> n_bytes=0, idle_age=30, priority=2,in_port=3 >> actions=output:2,output:1 >> cookie=0x2b00000000000001, duration=30.633s, table=0, n_packets=0, >> n_bytes=0, idle_age=30, priority=2,in_port=1 >> actions=output:2,output:3,CONTROLLER:65535 >> cookie=0x2b00000000000000, duration=30.636s, table=0, n_packets=0, >> n_bytes=0, idle_age=30, priority=2,in_port=2 >> actions=output:1,output:3 >> cookie=0x2b00000000000004, duration=36.03s, table=0, n_packets=14, >> n_bytes=1190, idle_age=4, priority=100,dl_type=0x88cc >> actions=CONTROLLER:65535 >> cookie=0x2b00000000000004, duration=36.03s, table=0, n_packets=0, >> n_bytes=0, idle_age=36, priority=0 actions=drop >> >> It seems that the packets are being flooded to all the other ports all >> the time. At first, I thought it was for LLDP packets, but the flow entries >> are not changing even after sending traffic between the hosts and even >> after removing one of the paths from host1 to host2. >> >> >> Flood flows are installed by default to broadcast any incoming traffic to >> all ports and controller (to discover hosts). Once hosts are discovered you >> should see MAC-to-MAC flows on top of the existing flood flows. This >> behavior is very similar to a real switch where known MAC traffic gets >> unicast forwarded ed while BUM traffic is broadcasted. >> >> >> The main requirement for me is that we get a traffic matrix and that the >> connectivity retains even after links are switched off. But here the >> traffic through all the other ports will be the same for a switch and thus >> can't find the way traffic is behaving. >> >> >> If something does not work after link/node goes down, this may be a bug >> or some limitation in the l2switch. >> >> >> >> 1. Am I doing something wrong or is there some configuration in the >> l2switch feature that I need to change? >> >> I do not think you are doing wrong, anyway here is the l2switch config >> setting for reference: >> https://docs.opendaylight.org/en/stable-oxygen/user-guide/l2switch-user-guide.html >> >> >> 1. ODL being a well-known controller is there any other way by which >> we can achieve reactive flows (other than using l2switch). >> >> You can try the VTN project: >> https://docs.opendaylight.org/en/stable-oxygen/user-guide/virtual-tenant-network-(vtn).html >> >> PS: Where do I begin if I wanted to get involved in the development and >> get the l2switch feature for a later release of ODL. >> >> >> Today the l2switch project is abandoned, if you really have cycles to >> revive it I would suggest to submit a proposal in >> https://wiki.opendaylight.org/view/Project_Proposals:Main so once it >> gets approved you can get committers rights to push code. >> > > Thushar, > > To begin: > > - Create a Linux Foundation account to get access to the ODL git if you > don't already have one > - You'll need Java 8 and maven 3.5.0. You can reference > https://wiki.opendaylight.org/view/GettingStarted:Development_Environment_Setup > although it looks a bit out-of-date > - Download the l2switch git project (for Fluorine: git clone -b > stable/fluorine ssh://<your user name>@ > git.opendaylight.org:29418/l2switch.git) > - Bump the odlparent version to 3.1.4 in all the pom files and the > yangtools version to 2.0.13. Fix build errors > - Push patches upstream > > Only someone with committer rights can merge patches. The project page, > https://wiki.opendaylight.org/view/L2_Switch:Main, lists the last known > PTL and committers. I don't of who those are still active in ODL at all. If > none are then, as Luis mentioned, the TSC could promote you to a committer. > > There was also some discussion on > https://stackoverflow.com/questions/53109420/problem-when-compiling-the-latest-version-of-l2switch > . > > >> >> >> _______________________________________________ >> Discuss mailing list >> disc...@lists.opendaylight.org >> https://lists.opendaylight.org/mailman/listinfo/discuss >> >> >> _______________________________________________ >> Discuss mailing list >> disc...@lists.opendaylight.org >> https://lists.opendaylight.org/mailman/listinfo/discuss >> > _______________________________________________ > Discuss mailing list > disc...@lists.opendaylight.org > https://lists.opendaylight.org/mailman/listinfo/discuss > -- M. Ranganathan
_______________________________________________ L2switch-dev mailing list L2switch-dev@lists.opendaylight.org https://lists.opendaylight.org/mailman/listinfo/l2switch-dev
