Description
Only in branch-1 and branch-1.2.
Fails look like this:
https://builds.apache.org/view/H-L/view/HBase/job/HBase-1.3/jdk=latest1.8,label=Hadoop/397/
TEST-org.apache.hadoop.hbase.regionserver.TestSplitTransactionOnCluster.xml.<init>
If I look in the xml, I see this:
<testcase name="testFailedSplit" classname="org.apache.hadoop.hbase.regionserver.TestSplitTransactionOnCluster" time="8.275"> <flakyFailure type="java.lang.AssertionError:">java.lang.AssertionError: null at org.junit.Assert.fail(Assert.java:86) at org.junit.Assert.assertTrue(Assert.java:41) at org.junit.Assert.assertTrue(Assert.java:52) at org.apache.hadoop.hbase.regionserver.TestSplitTransactionOnCluster.testFailedSplit(TestSplitTransactionOnCluster.java:1339) <system-err><![CDATA[
... the xml is cut off.
testFailedSplit seems to be the culprit.
If I look in the -output.txt I see:
.... 2015-11-25 09:00:37,816 DEBUG [asf905.gq1.ygridcore.net,48894,1448441976062_ChoreService_1] balancer.BaseLoadBalancer$Cluster(838): Lowest locality region index is 0 and its region server contains 3 regions 2015-11-25 09:00:37,816 DEBUG [asf905.gq1.ygridcore.net,48894,1448441976062_ChoreService_1] balancer.BaseLoadBalancer$Cluster(813): Lowest locality region server with non zero regions is asf905.gq1.ygridcore.net with locality 0.0 2015-11-25 09:00:37,816 DEBUG [asf905.gq1.ygridcore.net,48894,1448441976062_ChoreService_1] balancer.BaseLoadBalancer$Cluster(838): Lowest locality region index is 0 and its region server contains 3 regions 2015-11-25 09:00:37,817 DEBUG [asf905.gq1.ygridcore.net,48894,1448441976062_ChoreService_1] balancer.BaseLoadBalancer$Cluster(813): Lowest locality region server with non zero regions is asf905.gq1.ygridcore.net with locality 0.0 ...
spewing...
This test was added here:
kalashnikov:hbase.git.commit stack$ git log -S testFailedSplit ./hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestSplitTransactionOnCluster.java commit 871444cb0a733b82af843952253b4545a407979a Author: Andrew Purtell <apurtell@apache.org> Date: Mon Dec 15 17:31:33 2014 -0800 HBASE-12686 Failures in split before PONR not clearing the daughter regions from regions in transition during rollback (Vandana Ayyalasomayajula)
The balancer is not coming back true (line #1339 assert is null according to above)
... 1337 regions = TESTING_UTIL.getHBaseAdmin().getTableRegions(tableName); 1338 assertTrue(regions.size() == 1); 1339 assertTrue(admin.balancer()); ...
Line #1339 was not in original test. It was added later:
commit 46f993b19fa11d1a8880d08045be43e38017b46a
Author: Virag Kothari <virag@yahoo-inc.com>
Date: Wed Jan 7 10:58:32 2015 -0800
HBASE-12694 testTableExistsIfTheSpecifiedTableRegionIsSplitParent in TestSplitTransactionOnCluster class leaves regions in transition (Vandana Ayyalasomayajula)
We are having trouble achieving a balance.... Let me see.