airflow-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From davy...@apache.org
Subject incubator-airflow git commit: [AIRFLOW-1074] Don't count queued tasks for concurrency limits
Date Wed, 12 Apr 2017 18:56:42 GMT
Repository: incubator-airflow
Updated Branches:
  refs/heads/master 6b1c327ee -> 8f9f5084b


[AIRFLOW-1074] Don't count queued tasks for concurrency limits

There may be orphaned tasks queued but not in a
running dag run that
will not cleared. We should not count these as
they will interfere.

I hate to do this, but I changed my mind on
counting queued tasks.

1. Queued tasks that are actually queued generally
get set to running pretty quickly.
2. Because of the worker-side check, we won't
actually pass concurrency.

I don't think the queued thing is a big deal
because of this, I'm more worried about orphaned
tasks that are in QUEUED state but not in a
running dag_run (so they wont get reset)
interfering with concurrency.

There may be orphaned tasks queued but not in a
running dag run that
will not cleared. We should not count these as
they will interfere.

Closes #2221 from saguziel/aguziel-concurrency-2


Project: http://git-wip-us.apache.org/repos/asf/incubator-airflow/repo
Commit: http://git-wip-us.apache.org/repos/asf/incubator-airflow/commit/8f9f5084
Tree: http://git-wip-us.apache.org/repos/asf/incubator-airflow/tree/8f9f5084
Diff: http://git-wip-us.apache.org/repos/asf/incubator-airflow/diff/8f9f5084

Branch: refs/heads/master
Commit: 8f9f5084bfdc2aa4017fee12e22d2e94672765ba
Parents: 6b1c327
Author: Alex Guziel <alex.guziel@airbnb.com>
Authored: Wed Apr 12 11:56:03 2017 -0700
Committer: Dan Davydov <dan.davydov@airbnb.com>
Committed: Wed Apr 12 11:56:06 2017 -0700

----------------------------------------------------------------------
 airflow/jobs.py | 3 ++-
 tests/jobs.py   | 6 +++---
 2 files changed, 5 insertions(+), 4 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/incubator-airflow/blob/8f9f5084/airflow/jobs.py
----------------------------------------------------------------------
diff --git a/airflow/jobs.py b/airflow/jobs.py
index f031f6e..18cd82e 100644
--- a/airflow/jobs.py
+++ b/airflow/jobs.py
@@ -1064,11 +1064,12 @@ class SchedulerJob(BaseJob):
                 dag_id = task_instance.dag_id
 
                 if dag_id not in dag_id_to_possibly_running_task_count:
+                    # TODO(saguziel): also check against QUEUED state, see AIRFLOW-1104
                     dag_id_to_possibly_running_task_count[dag_id] = \
                         DAG.get_num_task_instances(
                             dag_id,
                             simple_dag_bag.get_dag(dag_id).task_ids,
-                            states=[State.RUNNING, State.QUEUED],
+                            states=[State.RUNNING],
                             session=session)
 
                 current_task_concurrency = dag_id_to_possibly_running_task_count[dag_id]

http://git-wip-us.apache.org/repos/asf/incubator-airflow/blob/8f9f5084/tests/jobs.py
----------------------------------------------------------------------
diff --git a/tests/jobs.py b/tests/jobs.py
index e3caa5d..e99778a 100644
--- a/tests/jobs.py
+++ b/tests/jobs.py
@@ -504,14 +504,14 @@ class SchedulerJobTest(unittest.TestCase):
         ti1.refresh_from_db()
         ti2.refresh_from_db()
         ti1.state = State.RUNNING
-        ti2.state = State.QUEUED
+        ti2.state = State.RUNNING
         session.merge(ti1)
         session.merge(ti2)
         session.commit()
 
         self.assertEqual(State.RUNNING, dr1.state)
         self.assertEqual(2, DAG.get_num_task_instances(dag_id, dag.task_ids,
-            states=[State.RUNNING, State.QUEUED], session=session))
+            states=[State.RUNNING], session=session))
 
         # create second dag run
         dr2 = scheduler.create_dag_run(dag)
@@ -538,7 +538,7 @@ class SchedulerJobTest(unittest.TestCase):
         self.assertEqual(3, DAG.get_num_task_instances(dag_id, dag.task_ids,
             states=[State.RUNNING, State.QUEUED], session=session))
         self.assertEqual(State.RUNNING, ti1.state)
-        self.assertEqual(State.QUEUED, ti2.state)
+        self.assertEqual(State.RUNNING, ti2.state)
         six.assertCountEqual(self, [State.QUEUED, State.SCHEDULED], [ti3.state, ti4.state])
 
         session.close()


Mime
View raw message