On doing a LACP timeout scenario, the ports associated with the bond interface would wait for 90 sec under the LAG group before moving to non-lag group. In this case, the LAG group id would be removed as there are no LAG’s. But on bringing up the bond interface again, the LAG group that gets created is assigned the same LAG group id. It looks like the LAG group id is retained and not cleared on the datastore and hence the problem.
OVS log:
-------------
Initial LAG creation,
mininet> sh ovs-ofctl -O Openflow13 dump-groups s1
OFPST_GROUP_DESC reply (OF1.3) (xid=0x2):
group_id=56923,type=select,bucket=weight:0,actions=output:1,bucket=weight:0,actions=output:2
group_id=15350,type=all,bucket=weight:0,actions=drop
mininet>
mininet>
mininet>
mininet>
Bond interface down scenario for performing the timeout case,
mininet> py net.get('h1').cmd('ifconfig bond0 down')
Waits for 90 sec
OFPST_GROUP_DESC reply (OF1.3) (xid=0x2):
group_id=56923,type=select,bucket=weight:0,actions=output:1,bucket=weight:0,actions=output:2
group_id=15350,type=all,bucket=weight:0,actions=drop
mininet>
mininet>
mininet>
Post 90 secs,
mininet> sh ovs-ofctl -O Openflow13 dump-groups s1
OFPST_GROUP_DESC reply (OF1.3) (xid=0x2):
group_id=15350,type=all,bucket=weight:0,actions=output:1,bucket=weight:0,actions=output:2
mininet>
mininet>
mininet> py net.get('h1').cmd('ifconfig bond0 up')
mininet>
After bond interface bringup,
mininet> sh ovs-ofctl -O Openflow13 dump-groups s1
OFPST_GROUP_DESC reply (OF1.3) (xid=0x2):
group_id=56923,type=select,bucket=weight:0,actions=output:1,bucket=weight:0,actions=output:2
group_id=15350,type=all,bucket=weight:0,actions=drop
The same group id gets created prior to the bond interface down and after the bond up scenario as well. Looks like the LAG group id is still retained and not cleared from the datastore.