Details
-
New Feature
-
Status: Closed
-
Major
-
Resolution: Duplicate
-
0.20.4
-
None
-
None
Description
Refer to 'ideas to improve throughput of the base writing' discussion on hbase user mailing list.
We should use heuristics to adjust various parameters based on load characteristics. Factors include, but not limited to, number of CPU (cores) on each node, amount of physical memory on each node, Heap space allocated to master/region servers.
HBase parameters shown below would be dynamic whose current values can be seen on master.jsp / regionserver.jsp
For JVM parameters such as -Xmx, master.jsp should display recommended value to user.
-------------
Thanks. I will make this change:
<property>
<name>hbase.hregion.memstore.block.multiplier</name>
<value>8</value>
</property>
<property>
<name>hbase.regionserver.msginterval</name>
<value>10000</value>
</property>
<property>
<name>hbase.hstore.compactionThreshold</name>
<value>6</value>
</property>
<property>
<name>hbase.hstore.blockingStoreFiles</name>
<value>18</value>
</property>
and see how it goes.
Jimmy.
With the above change, the write goes faster if it works, but now we have problem that the
insertion doesn't work at all, here is the result:
2010-06-09 22:29:04,330 WARN org.apache.hadoop.hdfs.DFSClient: DFS Read: java.io
.IOException: Could not find target position 0
Attachments
Issue Links
- relates to
-
HBASE-5349 Automagically tweak global memstore and block cache sizes based on workload
- Closed