[OPNFLWPLUG-698] Robot test for RETRY during FAILOVER scenarios

Mon, 27 Sep 2021 09:01:49 +0000

This issue continues #5919.

Run scale flow/group configuration to inventory config using RESTCONF.
During ODL & Switch commnunication make RETRY/RECONCILE happen.

There are these alternative scenarios:

Scenario 1: without cluster

run ODL single node, OVS (connected to ODL) and TCP/OF proxy
restconf inventory config put flows/groups
TCP/OF proxy return ERROR to ODL, but FRM starts RETRY
TCP/OF proxy will not return ERROR to ODL after RETRY
assertion: check flows and groups directly dump (they must be same as configure using RESTCONF)

Scenario 2: 3-node cluster RPC FlatBatch FAILOVER only

run ODL cluster, OVS (connected to ODL) and TCP/OF proxy
use restconf get to find out leader of the switch
restconf inventory config put flows/groups
disconnect switch from the leader of the switch
assertion: check flows and groups directly dump (they must be same as configure using RESTCONF and expect that FRM made retry to the newly registered RPC)

Scenario 3a: 3-node cluster FRM FAILOVER only with 1x RESTCONF

run ODL cluster, OVS (connected to ODL) and TCP/OF proxy
use jolokia http get to findout which ODL node is DS master
restconf inventory config put flows/groups
islolate ODL node which is DS master during configuring of the Switch (TPC/OF proxy will slow down the network communication and isolate ODL DS master)
assertion: check flows and groups directly dump (they must be same as configure using RESTCONF and expect that new master of the DS started RECONCILIATION)

Scenario 3b: 3-node cluster FRM FAILOVER only with 2x RESTCONF

run ODL cluster, OVS (connected to ODL) and TCP/OF proxy
use jolokia http get to findout which ODL node is DS master
restconf on DS master inventory config put flows/groups
islolate ODL node from another nodes (but the connectivity to the switch remains) which is DS master during configuring of the Switch (TPC/OF proxy will slow down the network communication and isolate ODL DS master).
restconf on new DS master inventory config put flows/groups (use different test data)
assertion: check flows and groups directly dump (they must be same as configure using RESTCONF and expect that new master of the DS started RECONCILIATION). By comparing flow/groups according to the second restconf check wether old leader has not overwritten the data.

Scenario 4: both 3-node cluster FRM FAILOVER only and RPC FlatBatch FAILOVER only

run ODL cluster, OVS (connected to ODL) and TCP/OF proxy
use jolokia http get to findout which ODL node is DS master
use restconf get to find out leader of the switch
restconf inventory config put flows/groups
islolate ODL node which is DS master during configuring of the Switch (TPC/OF proxy will slow down the network communication and isolate ODL DS master)
disconnect switch from the leader of the switch
assertion: check flows and groups directly dump (they must be same as configure using RESTCONF and expect that FRM RECONCILIATION happend)

Please, use TCP/OF proxy to:

Please use existing vz JSON files with flows depending on groups.

If possible reuse existing vz keywords for checking flows/group in ovs.

OpenDaylight JIRA