OK, I just investigated switch connection patch in [1] and the failure I see is
openflow breaks after switch in HA mode disconnects from candidate instance.
To reproduce:
1) connect switch to all instances:
sudo ovs-vsctl set-controller s1 "tcp:192.168.0.101:6633"
"tcp:192.168.0.102:6633" "tcp:192.168.0.103:6633"
2) use sudo ovs-vsctl list Controller to determine master instance (e.g. 101
above)
3) disconnect a candidate (103 above):
sudo ovs-vsctl set-controller s1 "tcp:192.168.0.101:6633"
"tcp:192.168.0.102:6633"
After that I see the following in karaf log of the candidate:
2017-02-10 03:01:19,830 | INFO | entLoopGroup-7-1 |
SystemNotificationsListenerImpl | 227 - org.opendaylight.openflowplugin.impl -
0.4.0.SNAPSHOT | ConnectionEvent: Connection closed by device,
Device:/192.168.0.24:59753, NodeId:openflow:1
2017-02-10 03:01:19,832 | INFO | entLoopGroup-7-1 | ContextChainHolderImpl
| 227 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Device
openflow:1 disconnected.
2017-02-10 03:01:19,832 | INFO | entLoopGroup-7-1 | ContextChainHolderImpl
| 227 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Context
chain for device openflow:1 successfully stopped
2017-02-10 03:01:20,838 | INFO | pool-31-thread-1 | ContextChainHolderImpl
| 227 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Removing
device: openflow:1 from DS
As a result of candidate clearing the DS, the master gets all kind of
exceptions and switch cannot connect anymore to controller. Workaround: restart
cluster.
Conclusion: the patch creates regression.
BR/Luis
[1] https://git.opendaylight.org/gerrit/#/c/51500
_______________________________________________
openflowplugin-dev mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
