You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "Sai Boorlagadda (JIRA)" <ji...@apache.org> on 2019/02/27 17:44:00 UTC
[jira] [Updated] (GEODE-6089) Rebalance flow - based on bucket
count and not bucket size
[ https://issues.apache.org/jira/browse/GEODE-6089?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sai Boorlagadda updated GEODE-6089:
-----------------------------------
Fix Version/s: (was: 1.9.0)
> Rebalance flow - based on bucket count and not bucket size
> ----------------------------------------------------------
>
> Key: GEODE-6089
> URL: https://issues.apache.org/jira/browse/GEODE-6089
> Project: Geode
> Issue Type: Improvement
> Reporter: yossi reginiano
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h
> Remaining Estimate: 0h
>
> In out system we have some regions which holds “transient data” – meaning that this data should be picked up by some process and once done should be deleted.
> For those exact regions the current Rebalance mechanism which is based on total buckets size is not sufficient , and causing buckets to be spread not evenly.
> For example we have 2 JVM’s which has “transient regions” and we are starting a third one – then we will run Rebalance , but because most buckets are empty (were handled and deleted by the process) , we can end up with a situation in which the third JVM will get responsibility only on a very minor amount of buckets. Now this unwanted behavior eventually causes the 2 first JVM’s to be responsible to much more buckets then the third one.
> In the solution I propose – I add a new region attribute property that upon defined in a specific (“transient”) region , will execute the rebalance based on bucket count, if the new region attribute property will not be defined - then the regular Rebalance will be used.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)