[ 
https://issues.apache.org/jira/browse/HBASE-27805?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Zheng Wang updated HBASE-27805:
-------------------------------
    Description: 
The default size of chunk is 2m, when we use G1, if heapRegionSize equals 4m, 
these chunks are allocated as humongous objects, exclusively allocating one 
region, then the remaining 2m become memory fragement.

Lots of memory fragement may lead to fullgc even if the percent of used heap 
not high enough.

I have tested to reduce the chunk size to 2047k(2m-1k, a bit lesser than half 
of heapRegionSize), there was no repeat of the above.

BTW, in g1, humongous objects are objects larger or equal the size of half a 
region.

 

  was:
The default size of chunk is 2m, when we use G1, if heapRegionSize equals 4m, 
these chunks are allocated as humongous objects, exclusively allocating one 
region, then the remaining 2m become memory fragement.

Lots of memory fragement may leading to fullgc even if the percent of used heap 
not high enough.

I have tested to reduce the chunk size to 2047k(2m-1k, a bit lesser than half 
of heapRegionSize), there was no repeat of the above.

BTW, in g1, humongous objects are objects larger or equal the size of half a 
region.

 


> The chunk created by mslab may cause memory fragement and lead to fullgc
> ------------------------------------------------------------------------
>
>                 Key: HBASE-27805
>                 URL: https://issues.apache.org/jira/browse/HBASE-27805
>             Project: HBase
>          Issue Type: Improvement
>          Components: regionserver
>            Reporter: Zheng Wang
>            Assignee: Zheng Wang
>            Priority: Major
>         Attachments: chunksize-2047k.png, chunksize-2048k-fullgc.png
>
>
> The default size of chunk is 2m, when we use G1, if heapRegionSize equals 4m, 
> these chunks are allocated as humongous objects, exclusively allocating one 
> region, then the remaining 2m become memory fragement.
> Lots of memory fragement may lead to fullgc even if the percent of used heap 
> not high enough.
> I have tested to reduce the chunk size to 2047k(2m-1k, a bit lesser than half 
> of heapRegionSize), there was no repeat of the above.
> BTW, in g1, humongous objects are objects larger or equal the size of half a 
> region.
>  



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

Reply via email to