Hi Chetan,

This is currently not possible. The main reason being the definition of
the discovery API stating that a 'leader is stable' - ie that it retain
leader status unless it crashes. If you now fiddle with the leader
election you can create a situation where a leader gets informed that it
lost leader status. I'm unsure if all TopologyEventListeners can cope with
that.

Having said that, this could be built into discovery of course at [0]: the
leader is defined as the one with the lowest 'leaderElectionId' - and that
one is set basically to the instance startup time. And that could be
modified to contain something configurable as a prefix.

Cheers,
Stefan
--
[0] - 
https://github.com/apache/sling/blob/528d9de30bfd134c0c5081ed09c2de4f8a9b20
bf/bundles/extensions/discovery/oak/src/main/java/org/apache/sling/discover
y/oak/pinger/OakViewChecker.java#L290

On 22/09/16 08:09, "Chetan Mehrotra" <[email protected]> wrote:

>I have a hetrogeneous cluster deployment consisting of different type
>of Sling instance. Of this some of the Sling instance play worker role
>and connect to same repository.
>
>Would it be possible to have a config where such cluster members do
>not participate in leader election and hence do not become candidate
>for running cluster singleton jobs
>
>Chetan Mehrotra


Reply via email to