[ 
https://issues.apache.org/jira/browse/HBASE-30398?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

huginn updated HBASE-30398:
---------------------------
    Description: 
HBase currently allows only one off-peak compaction at a time within a 
RegionServer. HStore uses a single AtomicBoolean as the admission gate, so 
stores contend for one permit even when the RegionServer has sufficient 
resources.

Support configuring the number of concurrent off-peak compactions per 
RegionServer.

The proposed implementation:
* replaces the single boolean gate with a semaphore-backed tracker;
* adds hbase.hstore.offpeak.compaction.concurrency, defaulting to 1;
* updates the permit count when configuration changes;
* adds unit coverage for configured concurrency and runtime updates.

  was:
HBase currently allows only one off-peak compaction at a time within a 
RegionServer. HStore uses a single AtomicBoolean as the admission gate, so 
stores contend for one permit even when the RegionServer has sufficient 
resources.

Support configuring the number of concurrent off-peak compactions per 
RegionServer.

The proposed implementation:
* replaces the single boolean gate with a semaphore-backed tracker;
* adds hbase.hstore.offpeak.compaction.concurrency, defaulting to 1;
* updates the permit count when configuration changes;
* adds unit coverage for configured concurrency and runtime updates.

A patch is available at:
https://github.com/apache/hbase/pull/8669


> Support configuring the concurrency of off-peak compactions
> -----------------------------------------------------------
>
>                 Key: HBASE-30398
>                 URL: https://issues.apache.org/jira/browse/HBASE-30398
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: huginn
>            Priority: Major
>
> HBase currently allows only one off-peak compaction at a time within a 
> RegionServer. HStore uses a single AtomicBoolean as the admission gate, so 
> stores contend for one permit even when the RegionServer has sufficient 
> resources.
> Support configuring the number of concurrent off-peak compactions per 
> RegionServer.
> The proposed implementation:
> * replaces the single boolean gate with a semaphore-backed tracker;
> * adds hbase.hstore.offpeak.compaction.concurrency, defaulting to 1;
> * updates the permit count when configuration changes;
> * adds unit coverage for configured concurrency and runtime updates.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to