https://github.com/yunionio/sdnagent

Cloudpods SDN Agent for Network Virtualization
https://github.com/yunionio/sdnagent

cloudpods openflow ovn ovs sdn

Last synced: about 2 months ago
JSON representation

Cloudpods SDN Agent for Network Virtualization

• Host: GitHub
• URL: https://github.com/yunionio/sdnagent
• Owner: yunionio
• License: apache-2.0
• Created: 2018-09-17T04:37:53.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2024-04-24T03:31:24.000Z (about 1 year ago)
• Last Synced: 2024-04-24T07:34:13.179Z (about 1 year ago)
• Topics: cloudpods, openflow, ovn, ovs, sdn
• Language: Go
• Homepage:
• Size: 25.4 MB
• Stars: 13
• Watchers: 8
• Forks: 16
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# todo

6. failsafe trigger

7. more usable cmdline

	add-flow br1 cookie=0x99,priority=99,,actions=

8. lock yunioncloud/pkg/log in Gopkg.toml

10. ping check on startup

21. encode who in cookie

22. intranet, external net

23. config file

24. vlan and ct zone allocation

26. match field, order by Name()

27. ovsdb port external_id

29. hostconfig with ct zone management, collision with ovn-controller?

31. check availability of conntrack

25. cgo libopenvswitch

33. maybe, robustness, add logic to detect ct() , ct_state arguments order

34. TODO redirect broadcast ip traffic to sec_IN

36. do we need to kill existing connection when new secrule applies

	- delete zone conntrack entries

37. conntrack entry timeout setting

# Test

Prepare dummy desc directory

- br0 in namespaces as physical hosts

- veth in namespace as virtual hosts

virtual hosts with single nic on the same host or different hosts

 - 2 on the same hosts

 - 2 on different hosts

virtual host with 2 nics enslaved to the same br0

 - 1 with 2 nics on different networks

 - 1 with the above as gateway in one of the network

 - 1 with the above as gateway in the other network

32. test ftp rel

20. regrestion test

38. nat for testing purposes

traffic control test

 - server start

 - server stop

 - server change bandwidth

 - zero bandwidth: no limit: api forbids this

 - qdisc statistics no reset on normal running

 - qdisc automatic restore after manual delete, change

		sudo tc qdisc delete dev vnet2-197 root

		sudo tc qdisc replace dev vnet2-197 root handle 1: tbf rate 100Kbit burst 100b latency 10ms

 - initial noqueue

 - initial with ingress

 - initial pfifo_fast

 - minimize erruption on restart

# plan: stateless flavour

- PRO: More efficient

- PRO: More straightforward, less error-prone

- CON: Bob can DoS Alice with invalid TCP traffic

`in: any`

	dl_dst=,ip[,nw_src=] 

`out: any`

	dl_src=,ip[,nw_dst=] 

`in: tcp`

	dl_dst=,tcp,tcpflags=+syn-ack[,tp_dst=][,nw_src=] 

	dl_dst=,tcp[,tp_dst=][,nw_src=] accept

	dl_src=,tcp[,tp_src=][,nw_dst=] accept

`out: tcp`

	dl_src=,tcp,tcpflags=+syn-ack[,tp_dst=][,nw_dst=] 

	dl_dst=,tcp[,tp_src=][,nw_src=] accept

	dl_src=,tcp[,tp_dst=][,nw_dst=] accept

`in: udp`

	dl_dst=.udp[,tp_dst=][,nw_src=] 

	dl_src=.udp[,tp_src=][,nw_dst=] 

`out: udp`

	dl_src=.udp[,tp_dst=][,nw_dst=] 

	dl_dst=.udp[,tp_src=][,nw_src=] 

`in: icmp`

	dl_dst=,icmp[,nw_src=] 

`out: icmp`

	dl_src=,icmp[,nw_dst=]