You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Alexey Goncharuk (JIRA)" <ji...@apache.org> on 2017/06/23 09:18:00 UTC

[jira] [Created] (IGNITE-5578) Discovery events coalescing

Alexey Goncharuk created IGNITE-5578:
----------------------------------------

             Summary: Discovery events coalescing
                 Key: IGNITE-5578
                 URL: https://issues.apache.org/jira/browse/IGNITE-5578
             Project: Ignite
          Issue Type: Improvement
          Components: cache, general
    Affects Versions: 1.0
            Reporter: Alexey Goncharuk
             Fix For: 2.1


There is an issue that has been in Ignite long ago and with the growing community and growing cluster sizes, it becomes more tangible.

When a bunch of nodes leave or join cluster, we generate a separate discovery event for each node, and each discovery event generates a partition map exchange. 
The first idea that came to my mind was to coalesce the partition map exchanges, but this is extremely difficult to implement on unstable topology.
Instead, we can introduce NODES_JOINED / NODES_FAILED events and batch multiple events on discovery level when possible. In this case, very few extra partition map exchanges are possible.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)