Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
1.13.0
-
None
Description
Steps:
1. Execute some query (for example "select * from sys.version")
2. Go to the profiles page: http://node1:8047/profiles
3. Open the details page for the query
4. Expand the duration section
Expected result:
The duration should be adequate and shouldn't be changed after the page reload
Actual result:
The duration is continuously increasing (You should reload page to notice that)
Planning | Queued | Execution | Total |
---|---|---|---|
0.092 sec | 0.012 sec | 59 min 36.487 sec | 59 min 36.591 sec |
Workaround: none
Note:
The issue introduced after the following fix:
https://issues.apache.org/jira/browse/DRILL-5963
ROOT CAUSE
After refactoring made in DRILL-5963 making query start / queue / end time was moved to another abstraction queryStateProcessor. On Foreman close queryStateProcessor was closed after final query profile was stored thus query end time was never updated. To fix this issue we need to close queryStateProcessor before writing query final profile.
Attachments
Issue Links
- is related to
-
DRILL-5963 Canceling a query hung in planning state, leaves the query in ENQUEUED state for ever.
- Closed
- links to