[
https://issues.apache.org/jira/browse/AIRFLOW-92?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15278028#comment-15278028
]
Bence Nagy commented on AIRFLOW-92:
-----------------------------------
While I agree that a design change is needed here, I don't think this tells us
the full story:
{quote}
in TI.are_dependencies_met a session.commit is called to make sure that in
evaluate_trigger_rule a state change will be saved
{quote}
{{evaluate_trigger_rule}} has a kwarg called {{flag_upstream_failed}} which is
set to False in {{get_active_runs}}. This means that in the context where
committing can cause this error, the commit is actually totally unnecessary
since state changes are disabled anyway, and a database update can never happen.
> Tasks not being retried at all due to a 'obj not bound to a Session' exception
> ------------------------------------------------------------------------------
>
> Key: AIRFLOW-92
> URL: https://issues.apache.org/jira/browse/AIRFLOW-92
> Project: Apache Airflow
> Issue Type: Bug
> Components: scheduler
> Affects Versions: Airflow 1.7.0
> Environment: EC2 t2.medium instance,
> Docker `version 1.11.1, build 5604cbe`,
> Host is `Linux ip-172-31-44-140 3.13.0-85-generic #129-Ubuntu SMP Thu Mar 17
> 20:50:15 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux`,
> Docker containers are built upon the `python:3.5` image,
> LocalExecutor is used with two scheduler containers running
> Reporter: Bence Nagy
> Priority: Critical
>
> I have some tasks that are stuck in {{up_for_retry}} state, below is an
> extract from the database. (here it is in a [Google Drive
> spreadsheet|https://docs.google.com/spreadsheets/d/14dtb3zYa583V1SaLcpOq6hDM4ThCeN7JhHjftRwKxbI/edit?usp=sharing]
> with better formatting)
> {code}
> task_id dag_id execution_date start_date end_date
> duration state try_number hostname unixname
> job_id pool queue priority_weight operator queued_dttm id
> dag_id state job_type start_date end_date
> latest_heartbeat executor_class hostname unixname id
> dag_id execution_date state run_id external_trigger conf
> end_date start_date
> task_a dag_a1 2016-05-09 08:00:00.000000 2016-05-09
> 12:00:12.382775 2016-05-09 12:01:12.473914 60.091139
> up_for_retry 1 d5593c115c22 root 46266 default 4
> ExternalTaskSensor 46266 success LocalTaskJob
> 2016-05-09 12:00:08.195711 2016-05-09 12:01:13.261937 2016-05-09
> 12:00:08.195732 LocalExecutor d5593c115c22 root 17799 dag_a1
> 2016-05-09 08:00:00.000000 failed scheduled__2016-05-09T08:00:00 false
> 2016-05-09 12:00:04.406875
> task_a dag_a2 2016-05-09 10:00:00.000000 2016-05-09
> 12:00:13.102094 2016-05-09 12:01:13.185960 60.083866
> up_for_retry 1 d5593c115c22 root 46270 default 4
> ExternalTaskSensor 46270 success LocalTaskJob
> 2016-05-09 12:00:08.896527 2016-05-09 12:01:13.960936 2016-05-09
> 12:00:08.896550 LocalExecutor d5593c115c22 root 17800 dag_a2
> 2016-05-09 10:00:00.000000 failed scheduled__2016-05-09T10:00:00 false
> 2016-05-09 12:00:04.531888
> task_b dag_b 2016-04-07 18:00:00.000000 2016-05-09
> 12:53:59.990395 2016-05-09 12:54:00.393259 0.402864
> up_for_retry 1 0a8613c2b5d2 root 46366 default 1
> PostgresOperator 46366 success LocalTaskJob
> 2016-05-09 12:53:58.881987 2016-05-09 12:54:03.891450 2016-05-09
> 12:53:58.882006 LocalExecutor 0a8613c2b5d2 root 17836 dag_b
> 2016-04-07 18:00:00.000000 running scheduled__2016-04-07T18:00:00 false
> 2016-05-09 12:51:59.713718
> task_c dag_b 2016-04-07 16:00:00.000000 2016-05-09
> 12:53:49.822634 2016-05-09 12:54:49.924291 60.101657
> up_for_retry 1 0a8613c2b5d2 root 46359 default 2
> ExternalTaskSensor 46359 success LocalTaskJob
> 2016-05-09 12:53:44.739355 2016-05-09 12:54:54.810579 2016-05-09
> 12:53:44.739575 LocalExecutor 0a8613c2b5d2 root 17831 dag_b
> 2016-04-07 16:00:00.000000 running scheduled__2016-04-07T16:00:00 false
> 2016-05-09 12:51:55.078050
> {code}
> I'm getting the following exception which seems to be halting the scheduler
> just before it could queue the tasks for retrying:
> {code}
> [2016-05-10 09:42:33,562] {jobs.py:706} ERROR - Instance <DagRun at
> 0x7f48a6b87550> is not bound to a Session; attribute refresh operation cannot
> proceed
> Traceback (most recent call last):
> File "/usr/local/lib/python3.5/site-packages/airflow/jobs.py", line 703, in
> _do_dags
> self.process_dag(dag, tis_out)
> File "/usr/local/lib/python3.5/site-packages/airflow/jobs.py", line 507, in
> process_dag
> active_runs = dag.get_active_runs()
> File "/usr/local/lib/python3.5/site-packages/airflow/models.py", line 2731,
> in get_active_runs
> active_dates.append(run.execution_date)
> File "/usr/local/lib/python3.5/site-packages/sqlalchemy/orm/attributes.py",
> line 237, in __get__
> return self.impl.get(instance_state(instance), dict_)
> File "/usr/local/lib/python3.5/site-packages/sqlalchemy/orm/attributes.py",
> line 578, in get
> value = state._load_expired(state, passive)
> File "/usr/local/lib/python3.5/site-packages/sqlalchemy/orm/state.py", line
> 474, in _load_expired
> self.manager.deferred_scalar_loader(self, toload)
> File "/usr/local/lib/python3.5/site-packages/sqlalchemy/orm/loading.py",
> line 610, in load_scalar_attributes
> (state_str(state)))
> sqlalchemy.orm.exc.DetachedInstanceError: Instance <DagRun at 0x7f48a6b87550>
> is not bound to a Session; attribute refresh operation cannot proceed
> {code}
> I've managed to fix this by removing all {{ti.are_dependencies_met()}} calls
> which have a commit at the end; after doing this there's no exceptions and
> the tasks are getting retried correctly.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
