[
https://issues.apache.org/jira/browse/HBASE-17323?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15755944#comment-15755944
]
Duo Zhang commented on HBASE-17323:
-----------------------------------
I think the queue full is in the setUp method where I put 1000 rows
concurrently. This is what [~carp84] said, with async client we need to pay
more attention on the pressure of RS. One simple solution is to use a separated
sleep mechanism when call queue is full, and a more beautiful solution is to
track the pressure of RS and give a reasonable sleep time when retrying. We
have done some works before, see the classes under o.a.h.h.client.backoff
package, but I haven't found a proper way to use them. But I do not think this
is the reason why this test fails. It just wastes some time...
And the error posted by Ted is intenional. In this test I keep moving
regions(include meta region) across RSes to test if there are data races which
will lead to incorrect result. Let me dig more.
Thanks.
> TestAsyncGetMultiThread fails in master
> ---------------------------------------
>
> Key: HBASE-17323
> URL: https://issues.apache.org/jira/browse/HBASE-17323
> Project: HBase
> Issue Type: Test
> Reporter: Ted Yu
> Attachments: 17323.hack.txt, testAsyncGetMultiThread-output.gz
>
>
> From
> https://builds.apache.org/job/HBase-Trunk_matrix/2137/jdk=JDK%201.8%20(latest),label=Hadoop/testReport/junit/org.apache.hadoop.hbase.client/TestAsyncGetMultiThread/test/
> :
> {code}
> java.util.concurrent.ExecutionException: java.lang.NullPointerException
> at org.apache.hadoop.hbase.util.Bytes.toInt(Bytes.java:1003)
> at org.apache.hadoop.hbase.util.Bytes.toInt(Bytes.java:980)
> at
> org.apache.hadoop.hbase.client.TestAsyncGetMultiThread.run(TestAsyncGetMultiThread.java:108)
> at
> org.apache.hadoop.hbase.client.TestAsyncGetMultiThread.lambda$null$1(TestAsyncGetMultiThread.java:122)
> {code}
> This can be reproduced locally.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
