GH-259: Ensure lock state after HZ cluster shutdown

[RobbanHoglund]

I have tested this implementation/fix against a running HZ cluster (client/server mode) running version 5.1.1.

This will ensure that, if a HZ cluster is totally shut down and then restarted the leader state will be reset if the client still possess/regain the lock.

[artembilan]

So, summarizing your change.
It looks like we need only to ensure that on every leader election we call cpSubSystem.getLock() instead of that stored into the property.

The rest of your changes are somehow not relevant...

Thanks

[RobbanHoglund]

The actual fix is here:

if (LeaderInitiator.this.getLock().isLockedByCurrentThread()) {
	if (!this.leader) {
	  // Since we have the lock we need to ensure that the leader flag is set
	 this.leader = true;
	}

Btw, Thanks for reviewing my code. :)

[artembilan]

Merged as e11e65e.

@RobbanHoglund ,

thank you very much for the contribution; looking forward for more!

BTW, we are in agreement to merge this Extensions project with main Spring Integration one: #254.
So, if you wish to contribute more, feel free to migrate HZ module and upgrade it respectively to the latest deps and so!

[RobbanHoglund]

Thanks for the good collaboration. And indeed I would gladly contribute if needed, especially with the Hazelcast parts.

[artembilan]

I would gladly contribute if needed

Right, as I said: we need now to migrate HZ extension project to Spring Integration as a module.
We did recently something similar for SMB one: spring-projects/spring-integration#3791.

So, that's kinda the story we may consider as a contribution at the moment for HZ.

Thanks

[RobbanHoglund]

Hi,
I did an attempt here:
spring-projects/spring-integration#3812