[ 
https://issues.apache.org/jira/browse/IGNITE-1230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14693309#comment-14693309
 ] 

Anton Vinogradov commented on IGNITE-1230:
------------------------------------------

New implementation (IGNITE-1093) allows to check rebelancing state.
For example you can check finished preloading or no when node stopped in this 
way:
You have to get all syncFutures at new topology (when node stopped) and wait 
for them.

        IgniteInternalFuture f1 = 
((GridCacheAdapter)grid(1).context().cache().internalCache(CACHE_NAME_DHT)).preloader().syncFuture();
        IgniteInternalFuture f2 = 
((GridCacheAdapter)grid(2).context().cache().internalCache(CACHE_NAME_DHT)).preloader().syncFuture();

        stopGrid(0);

        //TODO: refactor to get futures by topology
        while (f1 == 
((GridCacheAdapter)grid(1).context().cache().internalCache(CACHE_NAME_DHT)).preloader().syncFuture()
 ||
            f2 == 
((GridCacheAdapter)grid(2).context().cache().internalCache(CACHE_NAME_DHT)).preloader().syncFuture())
            U.sleep(100);        

        
((GridCacheAdapter)grid(1).context().cache().internalCache(CACHE_NAME_DHT)).preloader().syncFuture().get();
        
((GridCacheAdapter)grid(2).context().cache().internalCache(CACHE_NAME_DHT)).preloader().syncFuture().get();

I'm going to refactor this. This code is just initial implementation of this 
idea ;) and stopGrid(..) should be sync as StartGrid(..) in future.

> Add a boolean flag that will indicate that rebalance in progress or not
> -----------------------------------------------------------------------
>
>                 Key: IGNITE-1230
>                 URL: https://issues.apache.org/jira/browse/IGNITE-1230
>             Project: Ignite
>          Issue Type: Task
>          Components: cache
>    Affects Versions: ignite-1.4
>            Reporter: Alexey Kuznetsov
>            Assignee: Anton Vinogradov
>             Fix For: ignite-1.4
>
>
> It will be useful in some cases to check single boolean flag instead of 
> looping over partitions.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to