contiv / ofnet Goto Github PK

Vagrant mount the path successfully but when I do vagrant rsync-auto, it returns the following error:

There are no paths to watch! This is either because you have no synced folders using rsync, or any rsync synced folders you have have specified rsync_auto to be false.

When ovs switch attached to controller (ofnetagent), once controller failed, new flow entry will not send to ovs, and arp request will not go up to controller. How can we do for that?

If there is a mode change - we will need to remove the uplink - uplink related go routines continue to run without this.

Some use cases might require ARP flooding. There needs to be an option to enable such use cases.

When I use vxlan bridge mode, docker run seems stuck when ovs get restart.

Steps to reproduce:

1. make demo
2. netctl global set --fwd-mode bridge
3. netctl network create -s 10.0.0.0/24 -g 10.0.0.1 -e vxlan -pkt-tag 100 net1
4. docker run -itd --name c1 --net net1 debian /bin/bash
5. restart ovs-switchd
6. docker run -itd --name c2 --net net1 debian /bin/bash

step #6 will stuck

currently, ofnet only uses Veth_Pair to link containers to bridge, Is_use_veth_pair is set true by default and there is no other option in fact. We think ovs patch has a better performance over linux veth pair type and want to add it as an option.
Besides, when delete the veth pair, we should also delete the corresponding ovs interface, ovs-vsctl del-port XXX vvport17...

    Port "vvport17"
        tag: 101
        Interface "vvport17"
            error: "could not open network device vvport17 (No such device)"
    Port "eth2"
        Interface "eth2"
    Port "vvport21"
        tag: 101
        Interface "vvport21"
            error: "could not open network device vvport21 (No such device)"
    Port "vvport14"
        tag: 101
        Interface "vvport14"
            error: "could not open network device vvport14 (No such device)"
    Port "vvport15"
        tag: 101
        Interface "vvport15"
            error: "could not open network device vvport15 (No such device)"
    Port "vvport38"
        tag: 101
        Interface "vvport38"

TestPolicyAddDelete is failing with Go 1.7. This needs more investigation to figure out what's going on.

--- FAIL: TestPolicyAddDelete (2.29s)
ofnetPolicy_test.go:161: Could not find the flow priority=110,tcp,metadata=0x640190/0x7ffffffe,nw_src=10.10.10.0/24,nw_dst=10.1.1.0/24,tp_src=200,tp_dst=100 on ovs ovsbr60

Fwd mode change needs to be a blocking call - config should not be possible

Hi,

Itl looks like the controller does not support the ovs switches that has support for > OF 1.3.
I have the fix for it. But, I can't push to the branch.

Please let me know so that I can add the support.

Thanks
Chakri

If add to ofnetmaster fails (e.g. due to RPC failure), the endpoint needs to be removed from datapath before returning failure.

Hi,

I found that current flow matchfield and actions only support parts of standard Openflow messages. Do you have a plan to support Nicira extensions in flow entries?

I also found that controller only supports listen on TCP connections. Will it support Unix Domain Socket in the future?

If the switches have the same datapath_id, it can lead to issues. In OVS, if two bridges are attached to the same physical interface, their datapath_id will be the same.

Just a tracking issue so we don't forget to do this. After this is done, we'll need to update Godeps in netplugin again.