I would like to increase the ephemeral_storage of a Metaflow ECS task so that it can load larger Docker images (Pytorch GPU Docker images are 10Gb+ compressed). Currently, the instance runs out of disk space and shows the following error:
CannotPullContainerError: no space left on device
The ephemeral_storage is currently not defined in the metadata-service submodule so it is set to its default value of 21Gb - would it be possible to have a variable so that we can define this parameter?
jkuypers93
commented
6 months ago
Hello,
I would like to increase the ephemeral_storage of a Metaflow ECS task so that it can load larger Docker images (Pytorch GPU Docker images are 10Gb+ compressed). Currently, the instance runs out of disk space and shows the following error:
CannotPullContainerError: no space left on deviceThe
ephemeral_storageis currently not defined in the metadata-service submodule so it is set to its default value of 21Gb - would it be possible to have a variable so that we can define this parameter?I'd be happy to work on this and share a PR.
Many thanks!