Build images with BentoML base image

[mat-berna]

Hi,

I'm trying to build a container image from BentoML base image (https://docs.bentoml.org/en/latest/quickstart.html#containerize-model-api-server) and push it on Docker Hub.
I'm using Buildah technology inside a Tekton task manifest:

name: 3-build-buildah
workingdir: /workspace
image: quay.io/buildah/stable:latest
script: |
#!/usr/bin/env bash
sed -i -e 's/=\ "overlay"/=\ "vfs"/g' /etc/containers/storage.conf
echo "------------------------------ Find bentoml bundle"
...
echo "------------------------------ Start building bentoml bundle"
cd $LAST_IMAGE_PATH
ls -lh
buildah bud --layers --file $LAST_IMAGE_PATH/Dockerfile --tag $CR_REPO:$LAST_IMAGE --tag $CR_REPO:latest .
echo "------------------------------ Start pushing"
buildah push --tls-verify false $CR_REPO:$LAST_IMAGE
volumeMounts:
name: build-task-volume
mountPath: /workspace
name: build-container-registry-volume
mountPath: "/root/.docker"
securityContext:
privileged: true
runAsUser: 0

When I execute buildah bud command I have the following error:

STEP 4: COPY environment.yml requirements.txt setup.sh* bentoml-init.sh python_version* /bento/
error building at STEP
"COPY environment.yml requirements.txt setup.sh* bentoml-init.sh python_version* /bento/":
error adding sources
[
/tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216/environment.yml
/tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216/requirements.txt
/tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216/setup.sh*
/tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216/bentoml-init.sh
/tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216/python_version*
]
error checking on source /tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216/setup.sh*
under "/tekton/home/bentoml/repository/IrisClassifier/20210314110109_58D216": copier: stat: "/setup.sh*": no such file or directory

I tried with similar technologies that allows to build a container image inside another container, e.g. Kaniko and DIND (Docker in Docker).
With these technologies I haven't errors.

Can you help me?
Here you can find the source code to build container image with Dockerfile: irisclassifier-20210319171817-c36331.tar.gz

Thanks.

[rhatdan]

I could only get this to happen if setup.sh did not exists. Does it not exists in your tests? Does Docker and Kaniko allow specification with * that do not exist?

[mat-berna]

Exactly.
The files setup/* don't exist, however Docker and Kaniko don't give errors.
The Dockerfile is auto generated from BentoML and maybe in some contexts these file are necessary.