Run task of same dag on different servers - docker-compose

Question

I need to run the following dag -
dag = DAG('dummy_for_testing',
default_args=default_args,schedule_interval=None)
t1 = BashOperator(
task_id='print_date',
bash_command='date',
dag=dag)
t2 = BashOperator(
task_id='print_host',
bash_command='hostname',
queue='druid_queue',
dag=dag)
t3 = BashOperator(
task_id='print_directory',
bash_command='pwd',
dag=dag)
t3.set_upstream(t2)
t2.set_upstream(t1)
where t1 and t3 runs on server A and t2 runs on server B(queue=druid_queue).
I am currently using puckel/docker-airflow to set up airflow. The yml files for the server looks like :
Server1
version: '2.1'
services:
redis:
image: 'redis:3.2.7'
ports:
- "10.0.11.4:6999:6379"
command: redis-server
postgres:
image: postgres:9.6
container_name: postgres-airflow
ports:
- "10.0.11.4:5434:5432"
environment:
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
webserver:
image: puckel/docker-airflow:1.10.2
container_name: airflow
restart: always
depends_on:
- postgres
- redis
environment:
- LOAD_EX=n
- FERNET_KEY=<>
- EXECUTOR=Celery
- user_logs_config_loc=dags/user_logs/configurations/
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
volumes:
- /data/druid-data/airflow/dags:/usr/local/airflow/dags
- /var/run/docker.sock:/var/run/docker.sock
ports:
- "10.0.11.4:8085:8080"
command: webserver
healthcheck:
test: ["CMD-SHELL", "[ -f /usr/local/airflow/airflow-webserver.pid ]"]
interval: 30s
flower:
image: puckel/docker-airflow:1.10.2
restart: always
depends_on:
- redis
environment:
- EXECUTOR=Celery
ports:
- "5555:5555"
command: flower
scheduler:
image: puckel/docker-airflow:1.10.2
restart: always
depends_on:
- webserver
volumes:
- /data/druid-data/airflow/dags:/usr/local/airflow/dags
environment:
- LOAD_EX=n
- FERNET_KEY=<>
- EXECUTOR=Celery
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
- user_logs_config_loc=dags/user_logs/configurations/
command: scheduler
worker:
image: puckel/docker-airflow:1.10.2
restart: always
depends_on:
- scheduler
volumes:
- /data/druid-data/airflow/dags:/usr/local/airflow/dags
environment:
- FERNET_KEY=<>
- EXECUTOR=Celery
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
- user_logs_config_loc=dags/user_logs/configurations/
command: worker
Server2
version: '2.1'
services:
redis:
image: 'redis:3.2.7'
ports:
- "10.0.11.5:6999:6379"
command: redis-server
postgres:
image: postgres:9.6
container_name: postgres-airflow
ports:
- "10.0.11.5:5434:5432"
environment:
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
webserver:
image: puckel/docker-airflow:latest
container_name: airflow
restart: always
depends_on:
- postgres
- redis
environment:
- LOAD_EX=n
- FERNET_KEY=<>
- EXECUTOR=Celery
- user_logs_config_loc=dags/user_logs/configurations/
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
volumes:
- /data/qa/druid-data/airflow/dags:/usr/local/airflow/dags
- /var/run/docker.sock:/var/run/docker.sock
ports:
- "10.0.11.5:8085:8080"
command: webserver
healthcheck:
test: ["CMD-SHELL", "[ -f /usr/local/airflow/airflow-webserver.pid ]"]
interval: 30s
flower:
image: puckel/docker-airflow:1.10.2
restart: always
depends_on:
- redis
environment:
- EXECUTOR=Celery
ports:
- "5555:5555"
command: flower
scheduler:
image: puckel/docker-airflow:1.10.2
restart: always
depends_on:
- webserver
volumes:
- ./dags:/usr/local/airflow/dags
- /data/qa/druid-data/airflow/dags:/usr/local/airflow/dags
environment:
- LOAD_EX=n
- FERNET_KEY=<>
- EXECUTOR=Celery
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
command: scheduler
worker:
image: puckel/docker-airflow:1.10.2
restart: always
depends_on:
- scheduler
volumes:
- ./dags:/usr/local/airflow/dags
- /data/qa/druid-data/airflow/dags:/usr/local/airflow/dags
environment:
- FERNET_KEY=<>
- EXECUTOR=Celery
- POSTGRES_USER=airflow
- POSTGRES_PASSWORD=airflow
- POSTGRES_DB=airflow
command: worker -q druid_queue
The variable in server 1 looks like
broker_url=redis://redis:6379/1
result_backend=db+postgresql://airflow:airflow#postgres:5432/airflow
The variable in server 2 looks like
broker_url=redis://10.0.11.4:6999/1
result_backend=db+postgresql://airflow:airflow#10.0.11.4:5434/airflow
Is there something wrong with my configurations.
when I run the dag from the webserver of server A the dag is stuck:
The logs captured in the scheduler of Server A container:
[2019-04-12 14:42:35,184] {{jobs.py:1215}} INFO - Setting the follow tasks to queued state:
<TaskInstance: dummy_for_testing.print_date 2019-04-12 14:42:33.552786+00:00 [scheduled]>
[2019-04-12 14:42:35,194] {{jobs.py:1299}} INFO - Setting the following 1 tasks to queued state:
<TaskInstance: dummy_for_testing.print_date 2019-04-12 14:42:33.552786+00:00 [queued]>
[2019-04-12 14:42:35,194] {{jobs.py:1341}} INFO - Sending ('dummy_for_testing', 'print_date', datetime.datetime(2019, 4, 12, 14, 42, 33, 552786, tzinfo=<TimezoneInfo [UTC, GMT, +00:00:00, STD]>), 1) to executor with priority 3 and queue default
[2019-04-12 14:42:35,194] {{base_executor.py:56}} INFO - Adding to queue: airflow run dummy_for_testing print_date 2019-04-12T14:42:33.552786+00:00 --local -sd /usr/local/airflow/dags/dag_test.py
[2019-04-12 14:42:35,199] {{celery_executor.py:83}} INFO - [celery] queuing ('dummy_for_testing', 'print_date', datetime.datetime(2019, 4, 12, 14, 42, 33, 552786, tzinfo=<TimezoneInfo [UTC, GMT, +00:00:00, STD]>), 1) through celery, queue=default
[2019-04-12 14:42:37,152] {{jobs.py:1559}} INFO - Harvesting DAG parsing results
[2019-04-12 14:42:39,154] {{jobs.py:1559}} INFO - Harvesting DAG parsing results
[2019-04-12 14:42:40,610] {{sqlalchemy.py:79}} WARNING - DB connection invalidated. Reconnecting...
[2019-04-12 14:42:41,156] {{jobs.py:1559}} INFO - Harvesting DAG parsing results
[2019-04-12 14:42:41,179] {{jobs.py:1106}} INFO - 1 tasks up for execution:
<TaskInstance: dummy_for_testing.print_host 2019-04-12 14:42:33.552786+00:00 [scheduled]>
[2019-04-12 14:42:41,182] {{jobs.py:1141}} INFO - Figuring out tasks to run in Pool(name=None) with 128 open slots and 1 task instances in queue
[2019-04-12 14:42:41,184] {{jobs.py:1177}} INFO - DAG dummy_for_testing has 12/16 running and queued tasks
[2019-04-12 14:42:41,184] {{jobs.py:1215}} INFO - Setting the follow tasks to queued state:
<TaskInstance: dummy_for_testing.print_host 2019-04-12 14:42:33.552786+00:00 [scheduled]>
[2019-04-12 14:42:41,193] {{jobs.py:1299}} INFO - Setting the following 1 tasks to queued state:
<TaskInstance: dummy_for_testing.print_host 2019-04-12 14:42:33.552786+00:00 [queued]>
[2019-04-12 14:42:41,193] {{jobs.py:1341}} INFO - Sending ('dummy_for_testing', 'print_host', datetime.datetime(2019, 4, 12, 14, 42, 33, 552786, tzinfo=<TimezoneInfo [UTC, GMT, +00:00:00, STD]>), 1) to executor with priority 2 and queue druid_queue
[2019-04-12 14:42:41,194] {{base_executor.py:56}} INFO - Adding to queue: airflow run dummy_for_testing print_host 2019-04-12T14:42:33.552786+00:00 --local -sd /usr/local/airflow/dags/dag_test.py
[2019-04-12 14:42:41,198] {{celery_executor.py:83}} INFO - [celery] queuing ('dummy_for_testing', 'print_host', datetime.datetime(2019, 4, 12, 14, 42, 33, 552786, tzinfo=<TimezoneInfo [UTC, GMT, +00:00:00, STD]>), 1) through celery, queue=druid_queue
Server A config:
Server B config:
Sserver A celery Broker

Answer 1

It appears that you are running the same docker-compose stack on two different servers, but Server B has the worker started with command worker -q druid_queue. Typically you wan to be running airflow with only one scheduler, one database / results backend, and one message broker (redis) across all servers instead of running every service on every server.
Your compose file on the first server exposes redis at 10.0.1.4:6999 and below you noted that the broker_url on the second server is redis://10.0.11.4:6999/1. If the networking is setup properly then it might be as simple as updating the broker_url to redis://10.0.1.4:6999/1 (note: 11 -> 1)