[
https://issues.apache.org/jira/browse/HBASE-18650?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16137943#comment-16137943
]
cuizhaohua commented on HBASE-18650:
------------------------------------
Before I am enable replication, I put about ten recored in cluster 1, but only
one is different when I run "hbase
org.apache.hadoop.hbase.mapreduce.replication.VerifyReplication clusterID
repliTest" after I am setting master-master model .
I am not sure ,but I guess the code below in VerifyReplication.java is the
reason , when scan source cluster with timestamp and check sink cluster with
" scan.setStartRow(value.getRow());
scan.setStopRow(tableSplit.getEndRow()); "
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
HConnectionManager.execute(new HConnectable<Void>(conf) {
@Override
public Void connect(HConnection conn) throws IOException {
String zkClusterKey = conf.get(NAME + ".peerQuorumAddress");
Configuration peerConf = HBaseConfiguration.create(conf);
ZKUtil.applyClusterKeyToConf(peerConf, zkClusterKey);
TableName tableName = TableName.valueOf(conf.get(NAME +
".tableName"));
replicatedTable = new HTable(peerConf, tableName);
scan.setStartRow(value.getRow());
scan.setStopRow(tableSplit.getEndRow());
replicatedScanner = replicatedTable.getScanner(scan);
return null;
}
});
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> HBase replication Master-Master model issue
> ---------------------------------------------
>
> Key: HBASE-18650
> URL: https://issues.apache.org/jira/browse/HBASE-18650
> Project: HBase
> Issue Type: Bug
> Components: Replication
> Affects Versions: 1.1.6
> Reporter: cuizhaohua
> Priority: Minor
>
> When I use HBase replication with Master-Master model, install like below:
> 1> with cluster 1 I create table "repliTest", and put some data on it .
> 2> after about 6 month, I have cluster 2 ,and install master-master model
> with two cluster.
> 3> I put cluster 1 some data and It replication to cluster 2, it's correct.
> 4> I put cluster 2 some data and it replication to cluster 1, It's correct
> also.
> 5> The issue is : when I run command
> "hbase org.apache.hadoop.hbase.mapreduce.replication.VerifyReplication
> clusterID repliTest " on each cluster.
> the result is different :
> on cluster 1 the result is
>
> org.apache.hadoop.hbase.mapreduce.replication.VerifyReplication$Verifier$Counters
> BADROWS=33
> GOODROWS=20
> ONLY_IN_PEER_TABLE_ROWS=4
> ONLY_IN_SOURCE_TABLE_ROWS=29
> on cluster 2 the result is:
>
> org.apache.hadoop.hbase.mapreduce.replication.VerifyReplication$Verifier$Counters
> BADROWS=32
> GOODROWS=20
> ONLY_IN_PEER_TABLE_ROWS=28
> ONLY_IN_SOURCE_TABLE_ROWS=4
> It means one record is lost on cluster 2 verifierreplication . I check the
> table and get the recode is one recode I have put it 6 month ago.
> I delete this record like this:
> hbase(main):017:0> delete 'repliTest','3','score:english'
> and after that, when I run hbase verifyreplication is also correct on
> every cluster.
> environment :
> HBase 1.16
> hdfs 2.7.1
> ps: I am not good at english ,sorry about that.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
