Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
None
-
None
-
None
Description
There is no DNS in some real cluster network.
When we create a HiveClient in a Gateway machine, and submit a sql for HiveonSpark, then
1. the HiveClient will spark-submit a RemoteDriver class to YARN cluster, with args --remote-host which value is the Gateway hostname
2. the RemoteDriver which run on one slave of the YARN cluster, and it will connect to Gateway use the value of arg(--remote-host)
3. if the slave do not resolve the Gateway hostname(no DNS),it will connect failed
So, we can use ip of the Gateway for the --remote-host , and the slave can use ip to connect Hive Client on the Gateway