Hi All, I'm trying to achieve the following: I would like to send gratuitous arp replies on a OpenvSwitch (tagged) virtual machine (VM) interface (vnet5) from the host itself (Ubuntu 14.04, OVS 2.0.2) (to update upstream switch(es) their mac-address-tables after a live-migration of the VM). Therefore I made a little python script that utilizes Scapy, relevant part:
sendp(Ether(src=args.macaddress,dst="ff:ff:ff:ff:ff:ff") / ARP(op=2,
hwsrc=args.macaddress, psrc=args.ipaddress, hwdst="ff:ff:ff:ff:ff:ff",
pdst="255.255.255.255"), iface=args.vnet, count=4, inter=0.2)
The gratuitous arp replies do come in one the Virtual Machine itself, but never
make it out of the bridge (uplink), bond (bond1) and physical interface(s)
(eth2, eth5):
Bridge uplink
Port uplink
Interface uplink
type: internal
Port "bond1"
trunks: [998, 999]
Interface "eth5"
Interface "eth2"
Port "vnet5"
tag: 999
Interface "vnet5"
There are no specific flows active on the "in_port" of vnet5. I have tried an
openflow rule to allow a packet out if its ingress port (ovs-ofctl add-flow
uplink in_port=10,actions=in_port) but to no avail. So basically I can inject
packets into the vnet, but they will not make it out of the vnet. Does
OpenvSwitch filter this somewhere, or is it the linux kernel discarding these
packets?
Thanks for any hints on how to debug this issue.
Kind regards,
Stefan Kooman
P.s. Sending the gratuitous arp replies from the virtual machine itself works
like expected.
--
| BIT BV http://www.bit.nl/ Kamer van Koophandel 09090351
| GPG: 0xD14839C6 +31 318 648 688 / [email protected]
pgpV12yOsnj2R.pgp
Description: PGP signature
_______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
