[IMPALA-9227] Test coverage for query retries when there is a network partition

• Type: Sub-task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Backend
• Epic Color: ghx-label-3
• Assignee: Wenzhe Zhou
• Reporter: Sahil Takiar
• Created: 10/Dec/19 18:48
• Updated: 08/Sep/20 20:56

Description

The initial version of transparent query retries just adds coverage for retrying a query if an impalad crashes. Now that the Impala has an RPC fault injection framework (IMPALA-8138) based on debug actions, integration tests can introduce network partitions between two impalad processes.

Node blacklisting should cause the Impala Coordinator to blacklist the nodes with the network partitions (IMPALA-9137), and then transparent query retries should cause the query to be successfully retried.

Issue Links

• is related to: IMPALA-9137 Blacklist node if a DataStreamService RPC to the node fails
• is related to: IMPALA-8138 Re-introduce rpc debugging options