SparkFetcher with Namenode HA

[hoesler]

Am I right, that the configuration option for Namenode HA added with #55 to the SparkFSFetcher class (with has been removed in 28f4025) is not available for the new SparkFetcher class introduced with #162?

In my case, the SparkLogClient is unable to resolve the hostname for webhdfs (is null) when spark.eventLog.dir=hdfs:///spark-history, which is the default setting for HDP.

[akshayrai]

@hoesler, yes you are right. Namenode HA support seems to have been lost will migrating to the new fetcher. We need to add HA support for the new Spark Fetcher.