通过使用hbase项目主分支中的hbase spark connector，您可以完成您正试图完成的任务。由于某些原因，连接器似乎没有包含在任何正式的hbase构建中，但是您可以自己构建它，而且它工作正常。只需构建jar并将其包含在pom.xml中。
一旦构建，连接器将允许您在worker类中传递hbase连接对象，因此您不必担心序列化连接或构建singleton/etc。
例如：

JavaSparkContext jSPContext ...; //Create Java Spark Context
HBaseConfiguration hbConf = HBaseConfiguration.create();
hbConf.set("hbase.zookeeper.quorum", zkQuorum);
hbConf.set("hbase.zookeeper.property.clientPort", PORT_NUM);
// this is your key link to HBase from Spark -- use it every time you need to access HBase inside the Spark parallelism:
JavaHBaseContext hBaseContext = new JavaHBaseContext(jSPContext, hbConf);   

// Create an RDD and parallelize it with HBase access:
JavaRDD<String> myRDD = ... //create your RDD
hBaseContext.foreachPartition(myRDD,  new SparkHBaseWorker());
// You can also do other usual Spark tasks, like mapPartitions, forEach, etc.

// The Spark worker class for foreachPartition use-case on RDD of type String would look something like this:
class SparkHBaseWorker implements VoidFunction<Tuple2<Iterator<String>, Connection>>
{
    private static final long serialVersionUID = 1L;

    public WorkerIngest()
    {
    }

// Put all your HBase logic into this function:
    @Override
    public void call(Tuple2<Iterator<String>, Connection> t) throws Exception
    {           
        // This is your HBase connection object:
        Connection conn = t._2();
        // Now you can do direct access to HBase from this Spark worker node:
        Table hbTable = conn.getTable(TableName.valueOf(MY_TABLE));
        // now do something with the table/etc.
    }
}