[HIVE-18467] support whole warehouse dump / load + create/drop database events

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0
• Fix Version/s: 3.0.0
• Labels: pull-request-available
• Assignee: Anishek Agarwal
• Reporter: Anishek Agarwal
• Created: 17/Jan/18 09:05
• Updated: 27/Jul/18 05:13
• Resolved: 07/Feb/18 04:28

Description

A complete hive warehouse might be required to replicate to a DR site for certain use cases and rather than allowing only a database name in the REPL DUMP commands, we should allow dumping of all databases using the "*" option as in

REPL DUMP *

On the repl load side there will not be an option to specify the database name when loading from a location used to dump multiple databases, hence only

REPL LOAD FROM [location] would be supported when dumping via REPL DUMP *

Additionally, incremental dumps will go through all events across databases in a warehouse and hence CREATE / DROP Database events have to be serialized correctly to allow repl load to create them correctly.

Issue Links

• causes: HIVE-18538 Updating last.repl.id on replica warehouse when dbName specified as a pattern
• links to: GitHub Pull Request #300