bonding: Adding limited support for ARP monitoring with ovs
Get series with b4
b4 am [email protected]Versions newest first
- v1 Latest1 patch 11 repliesIn review
Patches in v1 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/1 | bonding: Adding limmited support for ARP monitoring with ovs.drivers/ | +7 | 11 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.
Cover letter [PATCH net-next v1 0/1] bonding: Adding limited support for ARP monitoring with ovs
Configurations with the bonding driver and openvswitch are unable to use the bond's "ARP Monitor" feature. If an ovs bridge sits above the bonding driver use of ARP Monitoring results in the bond flapping between slaves. bond_verify_device_path() gathers all vlan tags between the bond and the interface the arp is to be routed by walking the list of adjacent net_device's. When OVS is in the stack, this process breaks since ports on OVS bridge are not linked as they are with other configurations.
This patch adds limited support for the ARP Monitoring feature when OVS is configured above the bond. When no vlan tags are configured or when the tag is added between the bond interface and the OVS bridge arp monitoring will function correctly. The use of tags between the OVS bridge and the routed interfaces are not supported.
For example:
bond0 -> bond0.100 -> ovs-br -> ovs-port (x.x.x.x) is supported. bond0 -> ovs-br -> ovs-port -> ovs-port.100 (x.x.x.x) is not supported.
We recognize that some other advance network configurations (with-out OVS) may encounter the same issue. This is not an attempt to provide a generic solution, it will provide a solution for known use cases with OVS and the bonding driver as used by OpenShift with OVN-Kubernetes.
OVS bonding with BFD was evaluated as a possible solution. There are some limitations to adopting it.
In our environment the hypervisor manages SR-IOV interfaces. OVS bonding requires that all slave interfaces have promiscuous mode enabled. However, for promiscuous mode to function the hypervisor must also enable promiscuous mode on the VF. Unfortunately, the hypervisor allows only a single VF per PF to have promiscuous mode enabled.
This is a real customer problem, and they have expressed a strong desire to continue to use the bonding driver to maintain backward compatibility with their existing setup.