Problematic multicast packet handling
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
DragonFlow |
New
|
Wishlist
|
Omer Anson |
Bug Description
Hello
It turned out we have a problem with multicast packet handling.
When VM sends a multicast packet, default OVN rules send this packet to all other VMs
in the same network including the original VM.
Lets take the following rule in table 17:
{'cookie': u'0x0,', 'table': u'17,', 'match': u'metadata=
After adding new VM (port_id=51, tunnel_id=0xbb) we got the following rule:
{'cookie': u'0x0,', 'table': u'17,', 'match': u'metadata=
"set_field:
Potential solution:
-------
As I understand we want to keep number of OVN rules at minimal.
I can suggest to drop a new copy of the packet before sending it to the target VM (same as original VM).
We can simply drop multicast packet when "reg6=reg7" and "dl_dst=
Best regards,
Yuli
Changed in dragonflow: | |
importance: | Undecided → Medium |
Changed in dragonflow: | |
assignee: | nobody → Omer Anson (omer-anson) |
This will be treated in the implementation of the multicast spec. /blueprints. launchpad. net/dragonflow/ +spec/igmp- app
https:/