[HIVE-21866] LLAP status service driver may get stuck with wrong Yarn app ID - ASF JIRA

Log work

Agile Board

Rank to Top

Rank to Bottom

Bulk Copy Attachments

Bulk Move Attachments

Voters

Watch issue

Watchers

Create sub-task

Convert to sub-task

Move

Link

Clone

Labels

Update Comment Author

Replace String in Comment

Update Comment Visibility

Delete Comments

XML

Word

Printable

JSON

Details

Type: Bug
Status: Closed
Priority: Major
Resolution: Fixed
Affects Version/s: None
Fix Version/s: 4.0.0-alpha-1
Component/s: None
Labels:
- pull-request-available

Description

LLAPStatusDriver might get stuck polling status from Yarn if the following happen in this order:

there was a running LLAP Yarn app previously which is now finished / killed
Yarn was restarted
LLAPStatusDriver is invoked before any new LLAP app gets kicked off
LLAPStatusDriver receives the old app ID, which is then cached in the Yarn serviceClient object (no evicition)
In the meantime if any new LLAP app gets kicked off, LLAPStatusDriver will not see it, as it constantly retries fetching info about the wrong, old app ID (this is because we don't create new serviceClient objects)

ERROR status.LlapStatusServiceDriver: FAILED: 20: Failed to get Yarn AppReport
org.apache.hadoop.hive.llap.cli.status.LlapStatusCliException: 20: Failed to get Yarn AppReport
	at org.apache.hadoop.hive.llap.cli.status.LlapStatusServiceDriver.getAppReport(LlapStatusServiceDriver.java:292) [hive-llap-server-3.1.0.7.0.0.0-112.jar:3.1.0.7.0.0.0-134]
	at org.apache.hadoop.hive.llap.cli.status.LlapStatusServiceDriver.run(LlapStatusServiceDriver.java:209) [hive-llap-server-3.1.0.7.0.0.0-112.jar:3.1.0.7.0.0.0-134]
	at org.apache.hadoop.hive.llap.cli.status.LlapStatusServiceDriver.main(LlapStatusServiceDriver.java:537) [hive-llap-server-3.1.0.7.0.0.0-112.jar:3.1.0.7.0.0.0-134]....