Choosing the running node for HASingletons

[dimitris]

As I understand it, all HASingletons currently run on the same (master) node. It would be interesting to make the decision of which node to pick policy based (e.g. dynamic pick the node with the most free memory, or static, run on nodeX, if available).

If a user has many HASingletons, that would at least allow him to distribute the load and not have them all running on the same node.

[brian.stansberry]

Done :-) http://jira.jboss.com/jira/browse/JBAS-2499

The policy impl we have is a simple modulus based one (make node #X in the view become master), but the policy has access to the HAPartition, so a policy instance could communicate with its peers to negotiate something more complex.

[dimitris]

Excellent! Any chance of backporting this?

[brian.stansberry]

IIRC there shouldn't be any incompatibility or API-breakage problems. I'll ask Alex Fu if he can backport it; I won't have any cycles for a while.

I assume you are talking for 4.0.6.CR1.

[dimitris]

Yes, 4.0.6.CR1

[brian.stansberry]

Alex can do it; JIRA is http://jira.jboss.com/jira/browse/JBAS-3636 .

[alexfu.novell]

Dimitris,

I'm confused. You said 4.0.6.CR1 in this thread but you just modified http://jira.jboss.com/jira/browse/JBAS-3636 to say it needs to be done in Branch_4_2?

Should I back out the changes from 4_0 and redo it in 4_2?

Thanks,
-Alex

[dimitris]

Are you subscribed to the jboss-development list?

There won't be any 4.0.6 release.

Branch_4_0 is copied onto Branch_4_2 so your changes are now there.

Branch_4_0 was then rollbacked to what was when version 4.0.5.GA was produced, so we need to re-apply your changes to Branch_4_0.

Confusing? :)

[dimitris]

Actually, I only want you to reply JBAS-3805 in Branch_4_0.

JBAS-3636 should only be in Branch_4_2, so we are ok

[alexfu.novell]

Thanks.

I didn't realize that you created Branch_4_2 yesterday so I thought it was a parallel branch there for a long time. Now I'm not confused :)