https://dagster.io/ logo
Powered by
#deployment-kubernetes
Title
# deployment-kubernetes
a

ArturoCaracciola

03/30/2022, 8:43 AM
I found this issue https://github.com/dagster-io/dagster/issues/6988 but does "requiring you to set the namespace in each celery_k8s_job_executor in the job_namespace field" mean I have to edit the python or can I edit the config in the launchpad ?
j

johann

03/30/2022, 2:20 PM
That issue will only impact you if you’re using the 
CeleryK8sRunLauncher
in your Helm values. By default it will set you up with 
K8sRunLauncher
, which are you using? https://github.com/dagster-io/dagster/blob/master/helm/dagster/values.yaml#L386
a

ArturoCaracciola

03/31/2022, 6:38 PM
Indeed I'm using CeleryK8sRunLauncher. From a quick look at the architecture it did seem much more capable to handle high loads. But I may be wrong. Is it ready to use ? Should I stick with K8sRunLauncher ?
j

johann

03/31/2022, 7:20 PM
Both should be prepared for high load, the reasons to use 
CeleryK8sRunLauncher
are just if you like the features that having a celery queue brings, the most prominent is being able to limit concurrency by sending different ops to different queues
If using the CeleryK8sRunLauncher, then you just need to configure job_namespace in run config. That can either be provided in python (though note it needs to be provided directly, and can’t use configured), or in Dagit via the launchpad
We’ll also get that fixed soon so you don’t have to provide it each time.
a

ArturoCaracciola

04/01/2022, 7:37 AM
Alright, thanks a lot for the help. I tried to set job_namespace in the launchpad but I didn't find how and where to set it. I will stay with K8sRunLauncher for now.
j

johann

04/01/2022, 4:45 PM
The yaml block in the launchpad would look something like
Copy code
execution:
  config:
    job_namespace: "dagster"
But if you don’t need Celery, the K8sRunLauncher is certainly simpler
Powered by