Hi,

My data is affinity collocated. The issue is my 2nd node is not in a
baseline yet. Persistent cache had not been re-balanced yet but non
persistent cache had been done.
Distributed join is not a good choice for us because of performance. I'm
thinking about those solutions :

1. Don't sent request out of baseline node
2. Don't automatically re-balanced none persistent caches when cluster is
persistent enabled.
3. Add node to baseline right after it join cluster.
Is it possible to do this ?

On Thu, Mar 21, 2019 at 6:15 PM Ilya Kasnacheev <[email protected]>
wrote:

> Hello!
>
> Ignite assumes that your data is affinity collocated. While rebalancing is
> in process, obviously this is not the case.
>
> I think you can have this problem go away by setting
> distributedJoins=true. Have you tried that? Note that there is a
> potentially large performance impact.
>
> Regards,
> --
> Ilya Kasnacheev
>
>
> чт, 21 мар. 2019 г. в 14:08, Tâm Nguyễn Mạnh <[email protected]>:
>
>> Hi Igniters,
>>
>> Today i got this exception during testing cluster scaling.
>> I have 2 caches X and Y in cluster, X is persistent and Y is none
>> persistent.
>>
>> 1. I start 1st node A and activate cluster and populate data for X and Y
>> 2. I start 2nd node B but not yet update topology. Few seconds later Y
>> had been auto rebalanced.
>>
>> In A we have both X and Y data but in B we only have Y. Exception will be
>> thrown when I try to join X and Y in sql query: Caches have distinct sets
>> of data nodes [cache1=X, cache2=Y]
>> A        B
>> ==     ==
>> X        Y
>> Y
>> .
>> I think it because of none persistent data is automaticly rebalanced
>> whenver a node is join into cluster (event not in baseline yet) but not
>> persistent data. Persistent data is only rebalanced whenever baseline is
>> changed. Is it correct? And how do i prevent this exception during scaling ?
>>
>> --
>> Thanks & Best Regards
>>
>> Tam, Nguyen Manh
>>
>>

-- 
Thanks & Best Regards

Tam, Nguyen Manh

Reply via email to