dhume
09/30/2020, 7:18 PMconfig = {
"broker": f"sqs://{aws_access_key}:{aws_secret_key}@",
"backend": f"dynamodb://{aws_access_key}:{aws_secret_key}@us-east-1/{DYNAMO_TABLE}",
"task_default_queue": SQS_QUEUE,
"task_queues": {"test": {"queue": SQS_QUEUE}},
}
app = Celery("dagster", **config)
Using this as a celery worker outside of dagster it’s working fine. I’m just having trouble getting it to run within dagster. Reading through the docus I’m getting lost where to configure where between celery.yaml
, pipeline_run.yaml
and the execution config.
I can get the worker started doing
dagster-celery worker start --name worker_1 --app app
Launching the pipeline (the celery example) I set the broker and backend in the execution config. The tasks are sent to the broker but not to the backend. I’m not sure what I’m missing.
Also am I forcing extra steps? It seems cumbersome to have to write out the full broker and backend in the execution configmax
09/30/2020, 7:20 PMdhume
09/30/2020, 7:21 PMstorage:
filesystem:
execution:
celery:
config:
broker: <sqs://xxx:xxx@>
backend: <dynamodb://xxx:xxx@us-east-1/dagster-celery>
With xxx being my actual keysmax
09/30/2020, 7:28 PMdhume
09/30/2020, 7:31 PMbackend
. What I set in my celery app are broker
, backend
, task_queues
and task_default_queue
max
09/30/2020, 7:31 PM-y
and pointing it to a yaml file containing that config--app
flagdhume
09/30/2020, 7:48 PMcelery.yaml
I think I’m running into some problems with the default queue being used. But I’m also just confused on the role of the celery.yaml
if all the broker and backend and everything are defined within the celery in the app.py
that is being used by the dagster celery worker cli command and then the connection is made via the execution configmax
09/30/2020, 7:49 PMdhume
09/30/2020, 7:57 PMdagster-celery
but then I saw that SQS was creating a new queue and the messages were just remaining therecat
09/30/2020, 8:01 PMdhume
09/30/2020, 8:02 PM--queue
flag when launching the workermax
10/01/2020, 10:56 PMdhume
10/02/2020, 11:52 AMmax
10/02/2020, 1:37 PMdhume
10/02/2020, 1:39 PMmax
10/02/2020, 1:45 PMdhume
10/02/2020, 1:47 PM