Error Building Custom MPI Image Following Documentation

[luancaarvalho]

Hello team,

I followed the documentation to run the PI example, but when I tried to build my own images based on the docs at https://github.com/kubeflow/mpi-operator/tree/master/examples/v2beta1/pi, I encountered the following error:

docker build -t mpi-pi .
[+] Building 0.0s (2/2) FINISHED                                                                         docker:default
 => [internal] load build definition from Dockerfile                                                               0.0s
 => => transferring dockerfile: 257B                                                                               0.0s
 => [internal] load .dockerignore                                                                                  0.0s
 => => transferring context: 2B                                                                                    0.0s
Dockerfile:8
--------------------
   6 |     RUN mpic++ /src/pi.cc -o /pi
   7 |
   8 | >>> FROM mpioperator/openmpi:${BASE_LABEL}
   9 |
  10 |     COPY --from=builder /pi /home/mpiuser/pi
--------------------
ERROR: failed to solve: failed to parse stage name "mpioperator/openmpi:": invalid reference format

Question

Does it make sense to have the ARG BASE_LABEL in this image? It seems that the ${BASE_LABEL} variable is not being resolved, causing the build to fail.

Additional Information

• I'm using the latest version of Docker
• I'm following the example Dockerfile provided in the documentation

Any assistance or clarification would be greatly appreciated. Thank you!

[tenzen-y]

Actually, we need to build the base, openmpi-builder, and openmpi images in advance like here:

mpi-operator/Makefile

Lines 107 to 116 in ef9d049

	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(PLATFORMS) --build-arg port=${BASE_IMAGE_SSH_PORT} -t ${REGISTRY}/base:${RELEASE_VERSION} build/base
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(PLATFORMS) --build-arg BASE_LABEL=${RELEASE_VERSION} -t ${REGISTRY}/openmpi:${RELEASE_VERSION} build/base -f build/base/openmpi.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(PLATFORMS) -t ${REGISTRY}/openmpi-builder:${RELEASE_VERSION} build/base -f build/base/openmpi-builder.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(PLATFORMS) --build-arg BASE_LABEL=${RELEASE_VERSION} -t ${REGISTRY}/mpi-pi:${RELEASE_VERSION}-openmpi examples/v2beta1/pi
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(INTEL_PLATFORMS) --build-arg BASE_LABEL=${RELEASE_VERSION} -t ${REGISTRY}/intel:${RELEASE_VERSION} build/base -f build/base/intel.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(INTEL_PLATFORMS) -t ${REGISTRY}/intel-builder:${RELEASE_VERSION} build/base -f build/base/intel-builder.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(INTEL_PLATFORMS) --build-arg BASE_LABEL=${RELEASE_VERSION} -t ${REGISTRY}/mpi-pi:${RELEASE_VERSION}-intel examples/v2beta1/pi -f examples/v2beta1/pi/intel.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(MPICH_PLATFORMS) --build-arg BASE_LABEL=${RELEASE_VERSION} -t ${REGISTRY}/mpich:${RELEASE_VERSION} build/base -f build/base/mpich.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(MPICH_PLATFORMS) -t ${REGISTRY}/mpich-builder:${RELEASE_VERSION} build/base -f build/base/mpich-builder.Dockerfile
	${IMG_BUILDER} build $(BUILD_ARGS) --platform $(MPICH_PLATFORMS) --build-arg BASE_LABEL=${RELEASE_VERSION} -t ${REGISTRY}/mpi-pi:${RELEASE_VERSION}-mpich examples/v2beta1/pi -f examples/v2beta1/pi/mpich.Dockerfile