[IMPALA-5345] Under stress, some TransmitData() RPCs are not responded to

• Type: Sub-task
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: Impala 2.10.0
• Fix Version/s: Impala 2.10.0
• Epic Color: ghx-label-3
• Assignee: Henry Robinson
• Reporter: Henry Robinson
• Created: 20/May/17 20:43
• Updated: 10/Jun/17 05:06
• Resolved: 10/Jun/17 05:06

Description

Under stress conditions on two separate clusters (one secure, one not), I've seen some TransmitData() RPCs stay unresponded to forever, blocking the query's completion. The RPCs are seen by the recipient, but are not in the pending sender list.

Need to test further to see if this is related to the fix for IMPALA-5093 or if a response is dropped on some path if an row batch is 'retried' from the pending sender list.