Merge pull request #169 from asmacdo/e2e-slurm

Introducing e2e slurm tests

.github/workflows/e2e-slurm.yml

@@ -0,0 +1,30 @@
+---
+name: Slurm
+
+on: [push]
+jobs:
+    e2e-slurm:
+      name: Test e2e with SLURM
+      runs-on: ubuntu-latest
+      steps:
+        - name: checkout our repo
+          uses: actions/checkout@v4
+        - name: Install apptainer
+          uses: eWaterCycle/setup-apptainer@v2
+          with:
+            apptainer-version: 1.1.2
+        - name: Install Conda
+          uses: conda-incubator/setup-miniconda@v3
+          with:
+            activate-environment: babs
+            auto-update-conda: true
+            python-version: 3.9
+        - name: Conda info
+          shell: bash -el {0}
+          run: conda info
+        - name: Install Babs
+          shell: bash -el {0}
+          run: ./tests/e2e-slurm/install-babs.sh
+        - name: Execute e2e with SLURM
+          shell: bash -el {0}
+          run: ./tests/e2e-slurm/main.sh

.github/workflows/shellcheck.yml

@@ -0,0 +1,24 @@
+name: Shellcheck scripts
+
+on: [push, pull_request]
+
+jobs:
+  test:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Set up system
+      shell: bash
+      run: |
+        sudo apt-get update -qq
+        sudo apt-get install shellcheck
+    - uses: actions/checkout@v4
+    - name: Run shellcheck
+      run: |
+        shellcheck \
+        tests/e2e-slurm/container/babs-user-script.sh \
+        tests/e2e-slurm/container/ensure-env.sh \
+        tests/e2e-slurm/container/walkthrough-tests.sh \
+        tests/e2e-slurm/install-babs.sh \
+        tests/e2e-slurm/main.sh

.gitignore

@@ -11,4 +11,4 @@ build/
 
 # Distribution / packaging
 dist/
-babs/VERSION
+babs/VERSION

Makefile

@@ -0,0 +1,15 @@
+install:
+	./tests/e2e-slurm/install-babs.sh
+
+setup-user:
+	./tests/e2e-slurm/setup-user.sh
+
+e2e: clean
+	./tests/e2e-slurm/main.sh
+
+clean:
+	@ podman stop slurm 2>/dev/null || true
+	@ podman rm slurm 2>/dev/null || true
+	@[ -e .testdata/babs_test_project/toybidsapp-container ] && \
+		datalad remove -d .testdata/babs_test_project/toybidsapp-container --reckless kill || :
+	rm -rf .testdata

babs/utils.py

@@ -1685,7 +1685,13 @@ def submit_one_test_job(analysis_path, type_system, flag_print_message=True):
         # e.g., on MIT OpenMind: no 1st line from MSI; only 2nd line.
     else:
         raise Exception("type system can be slurm or sge")
-    job_id = int(job_id_str)
+
+    # This is necessary SLURM commands can fail but have return code 0
+    try:
+        job_id = int(job_id_str)
+    except ValueError as e:
+        raise ValueError(f"Cannot convert {job_id_str!r} into an int: {e}. "
+            f"That output is a result of running command {cmd} which produced output {msg}.")
 
     # log filename:
     log_filename = job_name + ".*" + job_id_str

setup.cfg

@@ -24,17 +24,18 @@ classifiers =
 [options]
 python_requires = >=3.7
 install_requires =
+    backoff
+    datalad >= 0.17.2
+    datalad_container >= 1.1.6
+    filelock >= 3.8.0
     nibabel >=2.2.1
     numpy
     pandas
-    tqdm
     pyyaml >= 6.0
-    #ruamel.yaml >= 0.17.21
-    datalad >= 0.17.2
-    datalad_container >= 1.1.6
-    regex
-    filelock >= 3.8.0
     qstat >= 0.0.5
+    regex
+    #ruamel.yaml >= 0.17.21
+    tqdm
 packages = find:
 include_package_data = True
 

tests/e2e-slurm/container/babs-user-script.sh

@@ -0,0 +1,117 @@
+#!/bin/bash -i
+
+SUBPROJECT_NAME=test_project
+
+set -eu
+
+echo "=============================================================="
+echo "We are now running as user $(whoami)"
+echo "DEBUG: MINICONDA_PATH=${MINICONDA_PATH}"
+echo "DEBUG: TESTDATA=${TESTDATA}"
+
+# without MINICONDA_PATH set, shellcheck cannot follow
+# shellcheck disable=SC1091
+source  "$MINICONDA_PATH/etc/profile.d/conda.sh"
+conda activate babs
+
+# record the miniconda path so it can added to the test env (slurm jobs do not preserve env)
+cat > /home/"$USER"/miniconda.env << EOF
+. "$MINICONDA_PATH/etc/profile.d/conda.sh"
+EOF
+
+
+
+git config --global user.name "e2e testuser"
+git config --global user.email "[email protected]"
+echo "Git user: $(git config user.name)"
+echo "Git email: $(git config user.email)"
+
+# TODO switch back to osf project
+# Populate input data (Divergent from tuturial, bc https://github.com/datalad/datalad-osf/issues/191
+pushd "${TESTDATA}"
+echo "Installing Input Data"
+datalad install ///dbic/QA
+
+# Singularity image created by root, then chowned to this user, and datalad must be run as this user
+datalad create -D "toy BIDS App" toybidsapp-container
+pushd toybidsapp-container
+datalad containers-add \
+    --url "${PWD}/../toybidsapp-0.0.7.sif" \
+    toybidsapp-0-0-7
+popd
+rm -f toybidsapp-0.0.7.sif
+
+
+# TODO File Issue: --where_project must be abspath file issue for relative path
+babs-init \
+    --where_project "${PWD}" \
+    --project_name $SUBPROJECT_NAME \
+    --input BIDS "${PWD}"/QA \
+    --container_ds "${PWD}"/toybidsapp-container \
+    --container_name toybidsapp-0-0-7 \
+    --container_config_yaml_file "${PWD}"/config_toybidsapp.yaml \
+    --type_session multi-ses \
+    --type_system slurm
+
+echo "PASSED: babs-init"
+echo "Check setup, without job"
+babs-check-setup --project_root "${PWD}"/test_project/
+echo "PASSED: Check setup, without job"
+
+babs-check-setup --project_root "${PWD}"/test_project/ --job-test
+echo "Job submitted: Check setup, with job"
+
+babs-status --project_root "${PWD}"/test_project/
+
+# Wait for all running jobs to finish
+while [[ -n $(squeue -u "$USER" -t RUNNING,PENDING --noheader) ]]; do
+    echo "squeue -u \"$USER\" -t RUNNING,PENDING"
+    squeue -u "$USER" -t RUNNING,PENDING
+    echo "Waiting for running jobs to finish..."
+    sleep 5 # Wait for 60 seconds before checking again
+done
+
+echo "No running jobs."
+
+# TODO make sure this works
+# Check for failed jobs TODO state filter doesn't seem to be working as expected
+# if sacct -u $USER --state=FAILED --noheader | grep -q "FAILED"; then
+if sacct -u "$USER" --noheader | grep -q "FAILED"; then
+    sacct -u "$USER"
+    echo "There are failed jobs."
+    exit 1 # Exit with failure status
+else
+    sacct -u "$USER"
+    echo "PASSED: No failed jobs."
+fi
+
+babs-submit --project-root "${PWD}/test_project/"
+
+# # Wait for all running jobs to finish
+while [[ -n $(squeue -u "$USER" -t RUNNING,PENDING --noheader) ]]; do
+    echo "squeue -u \"$USER\" -t RUNNING,PENDING"
+    squeue -u "$USER" -t RUNNING,PENDING
+    echo "Waiting for running jobs to finish..."
+    sleep 5 # Wait for 60 seconds before checking again
+done
+
+echo "========================================================================="
+echo "babs-status:"
+babs-status --project_root "${PWD}"/test_project/
+echo "========================================================================="
+
+# Check for failed jobs TODO see above
+# if sacct -u $USER --state=FAILED --noheader | grep -q "FAILED"; then
+if sacct -u "$USER" --noheader | grep -q "FAILED"; then
+    sacct -u "$USER"
+    echo "========================================================================="
+    echo "There are failed jobs."
+    exit 1 # Exit with failure status
+else
+    sacct -u "$USER"
+    echo "========================================================================="
+    echo "PASSED: No failed jobs."
+fi
+
+babs-merge --project_root "${PWD}"/test_project/
+echo "PASSED: e2e walkthrough successful!"

tests/e2e-slurm/container/config_toybidsapp.yaml

@@ -0,0 +1,21 @@
+# Arguments in `singularity run`:
+singularity_run:
+    --no-zipped: ""
+    --dummy: "2"
+    -v: ""
+
+# Output foldername(s) to be zipped, and the BIDS App version to be included in the zip filename(s):
+zip_foldernames:
+    toybidsapp: "0-0-7"
+
+# How much cluster resources it needs:
+cluster_resources:
+    interpreting_shell: /bin/bash
+    hard_memory_limit: 2G
+
+script_preamble: |
+    . ~/miniconda.env
+    conda activate babs
+
+# Where to run the jobs:
+job_compute_space: "/tmp"

tests/e2e-slurm/container/ensure-env.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+#
+if [ -z "${MINICONDA_PATH:-}" ]; then
+    if hash conda; then
+        # We don't need the return value, we already catch the error
+        # shellcheck disable=SC2155
+        export MINICONDA_PATH=$(/bin/which conda | xargs dirname | xargs dirname)
+    else
+        echo "ERROR: must have MINICONDA_PATH set or have 'conda' available"
+        exit 1
+    fi
+fi

tests/e2e-slurm/container/walkthrough-tests.sh

@@ -0,0 +1,50 @@
+#!/bin/bash -i
+# Here we perform all actions that must be done as root inside the container and then
+# execute the walkthrough as BABS_USER
+set -eu
+
+export TESTDATA=/opt/testdata
+BABS_USER=testuser
+
+# Install singularity inside the container
+yum update -y && yum install -y epel-release &&  yum update -y &&  yum install -y singularity-runtime apptainer
+
+# Wait for slurm to be up
+max_retries=10
+delay=10  # seconds
+
+echo "Try connecting to slurm with sacct until it succeeds"
+set +e # We need to check the error code and allow failures until slurm has started up
+export PATH=${PWD}/tests/e2e-slurm/bin/:${PATH}
+for ((i=1; i<=max_retries; i++)); do
+	# Check if the command was successful
+	if sacct > /dev/null; then
+		echo "Slurm is up and running!"
+		break
+	else
+		echo "Waiting for Slurm to start... retry $i/$max_retries"
+		sleep $delay
+	fi
+	# exit if max retries reached
+	if [ "$i" -eq "$max_retries" ]; then
+		echo "Failed to start Slurm after $max_retries attempts."
+	exit 1
+    fi
+done
+set -e
+
+# Currently we are root inside the container. Now we create a user to own the testdata
+useradd "$BABS_USER"
+# cp rather than use bind directly so it can be owned by the container user and not cause issues outside
+mkdir "${TESTDATA}"
+cp /opt/outer/* "${TESTDATA}"
+
+
+# We build the singularity container now while we are root, and use it later as testuser
+pushd "${TESTDATA}"
+singularity build  \
+    toybidsapp-0.0.7.sif \
+    docker://pennlinc/toy_bids_app:0.0.7
+
+chown -R "$BABS_USER:$BABS_USER" "${TESTDATA}"
+su "${BABS_USER}" "${TESTDATA}/babs-user-script.sh"

tests/e2e-slurm/install-babs.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+set -eu
+
+. tests/e2e-slurm/container/ensure-env.sh
+
+conda install -c conda-forge datalad git git-annex -y
+
+# Optional dependencies, required for e2e-slurm
+pip install datalad_container
+pip install datalad-osf
+
+pip install .

tests/e2e-slurm/main.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+#
+set -eux
+
+# Expects: Conda env to be activated
+# Expects: Babs to be installed
+#
+# WIP-NOT-WORKING
+# Reminder :Z for selinux
+
+# TODO switch back to upstream after build
+# Currently using asmacdo, OpenSSL bump upstream, but no new docker build
+# https://github.com/giovtorres/docker-centos7-slurm/pull/49
+REGISTRY=docker.io
+HUBUSER=asmacdo
+# HUBUSER=giovtorres
+REPO=centos7-slurm
+# REPO=docker-centos7-slurm
+TAG=23.11.07 # TODO
+
+FQDN_IMAGE=${REGISTRY}/${HUBUSER}/${REPO}:${TAG}
+THIS_DIR="$(readlink -f "$0" | xargs dirname )"
+
+# Sets MINICONDA_PATH
+. tests/e2e-slurm/container/ensure-env.sh
+
+if [ "$MINICONDA_PATH/envs/$CONDA_DEFAULT_ENV/bin/babs-init" != "$(which babs-init)" ]; then
+    echo "Error: This script expects to be run inside a conda env with 'babs-init'!" >&2
+    echo "       We have not found it in conda env '$CONDA_DEFAULT_ENV' under '$MINICONDA_PATH'" >&2
+    exit 1
+fi
+
+echo "Success, we are in the conda env with babs-init!"
+
+# PWD shared so babs can be optionally be installed with develop install
+podman run -it --rm \
+	--name slurm \
+	--hostname slurmctl  \
+	-e "MINICONDA_PATH=${MINICONDA_PATH}" \
+	--privileged \
+	-v "${PWD}:${PWD}:ro,Z" \
+	-v "${MINICONDA_PATH}:${MINICONDA_PATH}:Z" \
+	-v "${THIS_DIR}/container:/opt/outer:ro,Z" \
+	"${FQDN_IMAGE}" \
+	/bin/bash -c ". /opt/outer/walkthrough-tests.sh"