Agent Cluster fail-over followed by fail-back should work for network isolation

Current Situation

Server1 with

• JOC Cockpit being active
• Secondary Controller being standby
• Primary Director Agent being active

Server2 with

• JOC Cockpit being standby
• Primary Controller being active
• Secondary Director Agent being standby

Steps to reproduce the problem:

Isolate Server1 from network.

• Fail-over takes place for JOC Cockpit and Controller
• Agent Cluster remains active. The Primary Director Agent remains coupled and active
• Loss of Secondary Controller must be confirmed

Add server1 to the network.

• Primary JOC Cockpit becomes operational/standby
• Secondary Controller remains unknown

Isolate Server2 from network.

• Fail-over takes place.
• Secondary Controller becomes active on Server1
• Primary Director Agent on Server1 remains active
• Loss of Director Agent must be confirmed. If this is done, Agent cluster status is "PassiveLost".

Add server2 to the network.

• Primary Controller becomes standby on Server1
• Secondary Controller remains active on Server1
• Primary Director Agent on Server1 remains active
• Loss of Director Agent must be confirmed. If this is done, Agent cluster status is "NodeLossToBeConfirmed".
• It is not possible to make the Agent cluster work.

Desired Behavior

Cluster fail-over and fail-back should work.

Maintainer Notes

• Changes are effective since release 2.6.4.
• Further testing for JS-2092 confirms absence of problems.