You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by "shicheng31604@gmail.com" <sh...@gmail.com> on 2019/07/12 02:58:45 UTC

Date PartitIon and Jdbc Conneciton Host

Hi:
    Ignite does not use a master-slave architecture, so when we use sql, we can link to any node. If we write data, then the data is partitioned according to certain rules, or just write to the node written on JDBC URl?

    I have done a test .When I inserted   so many records that the node I connected could not bear , then an exception was thrown on the connected node because of OOM . And other nodes were all fine . This might show that there was no partiton?
 



shicheng31604@gmail.com

Re: Date PartitIon and Jdbc Conneciton Host

Posted by Anton Kurbanov <an...@gmail.com>.
Hi,

You may refer to following doc pages to understand how it works currently:
https://apacheignite.readme.io/docs/cache-modes
https://apacheignite.readme.io/docs/affinity-collocation

Affinity function uses cache entry key to determine to which node write
will be enlisted and it is done by Rendezvous Affinity Function which is
default and shipped with Ignite, you can find the links below the message.

What kind of exception do you see, are you able to provide log? Is it java
heap out of memory exception or IgniteOutOfMemoryException? What was the
data distribution at the moment of issue happening?

Javadoc reference:
https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/cache/affinity/AffinityFunction.html#partition-java.lang.Object-
Implementation reference:
https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/cache/affinity/rendezvous/RendezvousAffinityFunction.java#L491

пт, 12 июл. 2019 г. в 05:59, shicheng31604@gmail.com <
shicheng31604@gmail.com>:

> Hi:
>     Ignite does not use a master-slave architecture, so when we use sql,
> we can link to any node. If we write data, then the data is partitioned
> according to certain rules, or just write to the node written on JDBC URl?
>
>     I have done a test .When I inserted   so many records that the node I
> connected could not bear , then an exception was thrown on the connected
> node because of OOM . And other nodes were all fine . This might show that
> there was no partiton?
>
>
> ------------------------------
> shicheng31604@gmail.com
>