[IMPALA-4268] Rework coordinator buffering to buffer more data

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Impala 2.8.0
• Fix Version/s: Impala 3.4.0
• Component/s: Backend
• Labels: query-lifecycle resource-management
• Target Version: Impala 3.4.0
• Assignee: Unassigned
• Reporter: Henry Robinson
• Created: 10/Oct/16 18:36
• Updated: 26/Sep/19 15:26
• Resolved: 26/Sep/19 15:26

Description

PlanRootSink executes the producer thread (the coordinator fragment execution thread) in a separate thread to the consumer (i.e. the thread handling the fetch RPC), which calls GetNext() to retrieve the rows. The implementation was simplified by handing off a single batch at a time from the producers to consumer.

This decision causes some problems:

• Many context switches for the sender. Adding buffering would allow the sender to append to the buffer and continue progress without a context switch.
• Query execution can't release resources until the client has fetched the final batch, because the coordinator fragment thread is still running and potentially producing backpressure all the way down the plan tree.
• The consumer can't fulfil fetch requests greater than Impala's internal BATCH_SIZE, because it is only given one batch at a time.

The tricky part is managing the mismatch between the size of the row batches processed in Send() and the size of the fetch result asked for by the client without impacting performance too badly. The sender materializes output rows in a QueryResultSet that is owned by the coordinator. That is not, currently, a splittable object - instead it contains the actual RPC response struct that will hit the wire when the RPC completes. As asynchronous sender does not know the batch size, because it can in theory change on every fetch call (although most reasonable clients will not randomly change the fetch size).

Issue Links

• is blocked by: IMPALA-2905 Use FragmentMgr to manage coordinator fragments
• is depended upon by: IMPALA-7351 Add memory estimates for plan nodes and sinks with missing estimates
• is related to: IMPALA-7312 Non-blocking mode for Fetch() RPC
• is related to: IMPALA-558 HS2::FetchResults sets hasMoreRows in many cases where no more rows are to be returned
• relates to: IMPALA-1618 Impala server should always try to fulfill requested fetch size
• relates to: IMPALA-8656 Support for eagerly fetching and spooling all query result rows

Sub-Tasks

• Improve QueryResultSet interface to allow appending a batch of rows at a time: Resolved
• Fix test gap - no test coverage of non-trivial HS2 result sets: Resolved
• Impala server should always try to fulfill requested fetch size: Resolved
• HS2::FetchResults sets hasMoreRows in many cases where no more rows are to be returned: Resolved
• Non-blocking mode for Fetch() RPC: Resolved