Steve Buckingham created AIRFLOW-4349:
-----------------------------------------
Summary: Athena Operator Marked Successful on Failure
Key: AIRFLOW-4349
URL: https://issues.apache.org/jira/browse/AIRFLOW-4349
Project: Apache Airflow
Issue Type: Bug
Components: operators
Affects Versions: 1.10.3
Environment: Airflow 1.10.3 running on K8s with K8s Executor in AWS.
Reporter: Steve Buckingham
Steps to recreate:
Execute an Athena operator that returns a failed stage (in this case some
missing S3 pemissions). Athena returns a failed state but the failed state code
is not operating as expected.
{code:java}
[2019-04-18 00:06:35,222] {{logging_mixin.py:95}} INFO - [2019-04-18
00:06:35,222] {{aws_athena_hook.py:129}} INFO - Trial 1: Query is still in an
intermediate state - RUNNING [2019-04-18 00:07:05,272] {{logging_mixin.py:95}}
INFO - [2019-04-18 00:07:05,272] {{connectionpool.py:203}} INFO - Starting new
HTTP connection (1): XXX.XXX.XXX.XXX [2019-04-18 00:07:05,282]
{{logging_mixin.py:95}} INFO - [2019-04-18 00:07:05,282]
{{connectionpool.py:203}} INFO - Starting new HTTP connection (1):
XXX.XXX.XXX.XXX [2019-04-18 00:07:05,286] {{logging_mixin.py:95}} INFO -
[2019-04-18 00:07:05,286] {{connectionpool.py:238}} INFO - Resetting dropped
connection: athena.us-east-1.amazonaws.com [2019-04-18 00:07:05,377]
{{logging_mixin.py:95}} INFO - [2019-04-18 00:07:05,377]
{{aws_athena_hook.py:132}} INFO - Trial 2: Query execution completed. Final
state is FAILED
...
...
...
[2019-04-18 00:07:08,560] {{logging_mixin.py:95}} INFO - [2019-04-18
00:07:08,559] {{jobs.py:2562}} INFO - Task exited with return code 0
```
{code}
Looking at the code if the query is a failure it should then return this from
the Operator:
{code:java}
'Final state of Athena job is {}, query_execution_id is {}.'
{code}
But that is not in the logs. The task is then mark as successfully completed in
Airflow and the DAG continues.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
