[
https://issues.apache.org/jira/browse/HBASE-8817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13713357#comment-13713357
]
Thomas Pan commented on HBASE-8817:
-----------------------------------
Memstore flush
Memstore flush can be triggered under three conditions listed below. The
minimum flush unit is per region, not at individual memstore level.
1. When a memstore reaches the value specified by
hbase.hregion.memstore.flush.size, memstores that belong to its region will be
flushed out to disk.
2. When overall memstore usage reaches the value specified by
hbase.regionserver.global.memstore.upperLimit, memstores from various regions
will be flushed out to disk to reduce overall memstore usage in a Region
Server. The flush order is based on the descending order of region memstore
usage. There will be enough regions with their memstore flushed out until the
overall memstore usage drops to or slightly below
hbase.regionserver.global.memstore.lowerLimit.
3. When number of HLog per region server reaches the value specified in
hbase.regionserver.max.logs, memstores from various regions will be flushed out
to disk to reduce HLog count. The flush order is based on time with regions
having oldest memstores got flushed first until HLog count drops below
hbase.regionserver.max.logs.
> Enhance The Apache HBase Reference Guide
> ----------------------------------------
>
> Key: HBASE-8817
> URL: https://issues.apache.org/jira/browse/HBASE-8817
> Project: HBase
> Issue Type: Bug
> Components: documentation
> Reporter: Thomas Pan
>
> I am reading the guide here: http://hbase.apache.org/book.html#regions.arch
> I have noticed the following:
> - There is lack of coverage on scan. Note that Lars has a very good
> presentation from last year's HBaseCon:
> http://www.slideshare.net/cloudera/3-learning-h-base-internals-lars-hofhansl-salesforce-final
> - Under section 9.7.6.1, there is no mentioning of flush per region if a
> memstore reached the size threshold (and a possible flush of all memstores if
> over all memory usage of all the memstores reached memstore.upperlimit). Note
> that people need to combine the information provided in section 9 with
> section 2.3.1.1 to fully understand the situation.
> - Compaction tuning strategies. Both Nicolas from Facebook and Sergey from
> Hortonworks have presented various compaction algorithms. Sometime, once all
> the necessary code is in master, we need to have the coverage.
> Let me know how I can help on the matter.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
