Has anyone deploying open source Dagster on kubern...
# deployment-kubernetes
m
Has anyone deploying open source Dagster on kubernetes had to deal with increasing shared memory size (shm) of the job container? I am using GKE with the helm chart and figuring out where to pass through values to increase the
shm
size.
a
Here you go:
Copy code
extra_tags["dagster-k8s/config"]["pod_spec_config"]["volumes"] = [
            {
                "name": "shm-volume",
                "empty_dir": {"medium": "Memory", "size_limit": "10Gi"},
            }
        ]
        extra_tags["dagster-k8s/config"]["container_config"] = {
            "resources": {
                "limits": {"<http://nvidia.com/gpu|nvidia.com/gpu>": "1"},
            },
            "volume_mounts": [{"mountPath": "/dev/shm", "name": "shm-volume"}],
        }
🥳 1
m
@Andrea Giardini you saved me so much time! Thank you
🎉 1
s
Thanks a lot from 1 month later, it is useful for me as well!
🎉 1
a
Happy to hear. @daniel do we have a place to put these kinds of snippets? An example section in the docs or something like that maybe?
d
My default for useful snippets that don't necessarily fit naturally into the docs is to make a Github discussion and answer it
These are also searchable at discuss.dagster.io although the UI is a bit clunky
a
That’s a good suggestion. I will do that