This project is to offer a solution for a multi-HBase client that will allow for the following benefits:
The first graph is the break down of which cluster was being used. The red areas are when I bounced the primary HBase cluster and it failed to deliver responses.
Note that the average response time is pretty much unaffected.
Also note there is a hit when the cluster got down but the max delay is at 700 milliseconds.
More testing will happen in the coming weeks.
There is a design doc in the root folder named MultiHBaseClientDesignDoc.docx. That document goes over the approach and the configuration options.
java -cp cxf-rt-transports-http-2.7.5.jar:hbase.multicluster-0.0.1-SNAPSHOT.jar org.apache.hadoop.hbase.test.MultiThreadedMultiClusterWithCmApiTest tedmalaska-hbase-mcc-2-1.ent.cloudera.com admin admin PrimaryCluster HBase tedmalaska-hbase-mcc-2-1.ent.cloudera.com admin admin FailoverCluster HBase2 mcc c 100 100 2 stats.csv