Added Slurm Docker images

.github/workflows/ci-tests.yaml

@@ -0,0 +1,81 @@
+name: "CI Tests"
+on:
+  push:
+    branches:
+      - master
+  pull_request:
+    branches:
+      - master
+concurrency:
+  group: build-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+jobs:
+  slurm:
+    name: "Slurm Docker images tests"
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - uses: docker/setup-qemu-action@v3
+      - uses: docker/setup-buildx-action@v3
+      - name: "Get current date"
+        run: echo "NOW=$(date +'%Y-%m-%dT%H:%M:%S')" >> $GITHUB_ENV
+      - name: "Build slurmctld image"
+        uses: docker/build-push-action@v5
+        with:
+          build-args: |
+            COMMIT=${{ github.sha }}
+            CREATION_DATE={{ env.NOW }}
+          context: slurm/slurmctld
+          load: true
+          tags: alphaunito/slurmctld:21.08.5
+      - name: "Build slurmd image"
+        uses: docker/build-push-action@v5
+        with:
+          build-args: |
+            COMMIT=${{ github.sha }}
+            CREATION_DATE={{ env.NOW }}
+          context: slurm/slurmd
+          load: true
+          tags: alphaunito/slurmd:21.08.5
+      - name: "Start Docker Compose"
+        id: start-compose
+        run: |
+          docker compose                      \
+              --file slurm/docker-compose.yml \
+              --project-name slurm            \
+            up                                \
+              --wait                          \
+              --wait-timeout 60               \
+      - name: "Run test with Docker Compose"
+        id: run-tests
+        run: |
+          docker compose                      \
+              --file slurm/docker-compose.yml \
+              --project-name slurm            \
+            exec                              \
+              --user hpcuser                  \
+              slurmctld srun hostname
+      - name: "Show slurmctld logs on failure"
+        if: ${{ always() && (steps.start-compose.outcome == 'failure' || steps.run-tests.outcome == 'failure') }}
+        run: |
+          docker compose                      \
+              --file slurm/docker-compose.yml \
+              --project-name slurm            \
+            logs                              \
+              slurmctld
+      - name: "Show slurmd logs on failure"
+        if: ${{ always() && (steps.start-compose.outcome == 'failure' || steps.run-tests.outcome == 'failure') }}
+        run: |
+          docker compose                      \
+              --file slurm/docker-compose.yml \
+              --project-name slurm            \
+            logs                              \
+              slurmd
+      - name: "Stop Docker Compose"
+        if: ${{ always() }}
+        run: |
+          docker compose                      \
+              --file slurm/docker-compose.yml \
+              --project-name slurm            \
+            down                              \
+              --volumes

.github/workflows/release.yml

@@ -0,0 +1,42 @@
+name: "Release Docker images"
+on:
+  workflow_run:
+    workflows:
+      - "CI Tests"
+    branches:
+      - master
+    types:
+      - completed
+jobs:
+  slurm:
+    name: "Slurm Docker images"
+    runs-on: ubuntu-22.04
+    if: ${{ github.event.workflow_run.conclusion == 'success' }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: docker/setup-qemu-action@v3
+      - uses: docker/setup-buildx-action@v3
+      - name: "Get current date"
+        run: echo "NOW=$(date +'%Y-%m-%dT%H:%M:%S')" >> $GITHUB_ENV
+      - uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: "Build slurmctld image"
+        uses: docker/build-push-action@v5
+        with:
+          build-args: |
+            COMMIT=${{ github.sha }}
+            CREATION_DATE={{ env.NOW }}
+          context: slurm/slurmctld
+          push: true
+          tags: alphaunito/slurmctld:21.08.5
+      - name: "Build slurmd image"
+        uses: docker/build-push-action@v5
+        with:
+          build-args: |
+            COMMIT=${{ github.sha }}
+            CREATION_DATE={{ env.NOW }}
+          context: slurm/slurmd
+          push: true
+          tags: alphaunito/slurmd:21.08.5

README.md

@@ -0,0 +1,17 @@
+# Docker for HPC
+
+This repository collects Dockerized versions of the most widespread HPC stacks. In particular, each folder contains:
+
+- A set of `Dockerfiles` that can be used standalone or manually assembled in a microservices application;
+- A `docker-compose.yml` file that simulates an HPC cluster orchestrated by the selected stack.
+
+This repository's main purpose is to provide the HPC community with Docker-based simulated environments of HPC stacks, which will be used mainly for small-scale experiments, debugging, and CI/CD pipelines.
+
+All Docker containers are published on DockerHub under the `alphaunito` organization. At the moment, this repository contains Dockerized versions of:
+
+- [Slurm](./slurm/README.md)
+
+All images support the explicit definition of opencontainers [annotations](https://github.com/opencontainers/image-spec/blob/main/annotations.md) through two build args:
+
+- The `COMMIT` build arg should contain the digest of the commit the image is being created from. It populates the `org.opencontainers.image.revision` label
+- The `CREATION_DATE` build arg should contain the creation time of the image in `RFC 3339` format. It populates the `org.opencontainers.image.created` label

slurm/README.md

@@ -0,0 +1,54 @@
+# Slurm on Docker
+
+This folder contains a fully Dockerized version of the [Slurm](https://slurm.schedmd.com/) queue manager. The version of SLURM shipped is the one downloaded from the apt repository, which is currently `21.08.5`
+
+This repository contains the source code of different container images:
+
+- `alphaunito/slurmctld:21.08.5`, which runs the Slurm control plane
+- `alphaunito/slurmd:21:08.5`, which runs a Slurm compute node
+
+Plus, it also contains a [docker-compose.yml](./docker-compose.yml) file that can deplyo an entire Slurm cluster with a single controller and a set of compute nodes. All these components are detailed below
+
+## Slurmctld
+
+The `slurmctld` process is the central management daemon of Slurm. It constitutes the control plane of the Slurm queue manager. The `slurmctld` Docker image can be build and published on DockerHub using the following commands
+
+```bash
+docker build -t alphaunito/slurmctld:21.08.5 slurmctld
+docker push alphaunito/slurmctld:21.08.5
+```
+
+To correctly populate the `slurm.conf` file, a `slurmctld` container needs 3 environment variables:
+
+- The `SLURMCTLD_HOSTNAME` variable should contain the hostname of the container. It is populated by default using the `hostname` command
+- The `SLURMD_NODES` variable must contain the number of compute nodes that Slurm should manage. If this variable is not set, the container displays an error message and terminates
+- The `SLURMD_HOSTNAME_PREFIX` variable should contain the prefix of the hostname used to identify compute nodes. If this variable is not set, the container displays an error message and terminates
+
+Note that all the compute nodes in the simulated HPC cluster should have a reachable hostname equal to `"${SLURMD_HOSTNAME_PREFIX}${X}"`, where `X` is an integer in the range `[1, ${SLURMD_NODES}]`
+
+## Compute
+
+The `slurmd` process is the compute node daemon for Slurm. It monitors all tasks running on the compute node , accepts work (tasks), launches tasks, and kills running tasks upon request. The `slurmd` Docker image can be build and published using the following commands
+
+```bash
+docker build -t alphaunito/slurmd:21.08.5 slurmd
+docker push alphaunito/slurmd:21.08.5
+```
+
+To correctly connect to a `slurmctld` node, a `slurmd` container needs a `SLURMCTLD_HOSTNAME` variable that should contain the hostname of the target `slurmctld` container. If this variable is not set, the container displays an error message and terminates
+
+## Cluster
+
+The `slurmctld` and `slurmd` images described above can be used to set up a Docker-based Slurm cluster with a single controller and many compute nodes. This task can be achieved either manually or starting from the [docker-compose.yml](./docker-compose.yml) file contained in this repository
+
+Note that the `slurmctld` node should have an identifiable hostname, as compute nodes must register with the control plane to be addressable. In Docker Compose, an explicit hostname can be set for a given service using the `hostname` keyword.
+
+Slurm relies on [MUNGE](https://dun.github.io/munge/) for authentication. To setup a munge cluster, all the involved nodes should share a common key located in the the `/etc/munge` fodler. Given that, the `/etc/munge` folder should be mounted as a shared volume among all the containers in the Slurm cluster.
+
+To allow for unprivileged workloads, an `hpcuser` has been configured inside the images. Commands can be executed by explicitly impersonating this user, through the `--user hpcuser` flag. For example
+
+```bash
+docker exec -it --user hpcuser slurmctld bash
+```
+
+In order to simulate an HPC facility, where the home folder is commonly mounted on a shared parallel file system, users may want to mount the `/home/hpcuser` folder as a shared volume among all the containers in the Slurm cluster.

slurm/docker-compose.yml

@@ -0,0 +1,32 @@
+version: "3.8"
+services:
+  slurmctld:
+    image: alphaunito/slurmctld:21.08.5
+    environment:
+      SLURMD_HOSTNAME_PREFIX: ${COMPOSE_PROJECT_NAME}-slurmd
+      SLURMD_NODES: 2
+    hostname: slurmctld
+    networks:
+      - slurmnet
+    volumes:
+      - home:/home/hpcuser
+      - munge:/etc/munge
+      - mysql:/var/lib/mysql
+  slurmd:
+    image: alphaunito/slurmd:21.08.5
+    deploy:
+      mode: replicated
+      replicas: 2
+    environment:
+      SLURMCTLD_HOSTNAME: slurmctld
+    networks:
+      - slurmnet
+    volumes:
+      - home:/home/hpcuser
+      - munge:/etc/munge
+networks:
+  slurmnet:
+volumes:
+  home:
+  munge:
+  mysql:

slurm/slurmctld/Dockerfile

@@ -0,0 +1,68 @@
+FROM ubuntu:jammy
+
+ARG COMMIT="none"
+ARG CREATION_DATE="none"
+
+LABEL org.opencontainers.image.authors="Iacopo Colonnelli <[email protected]>"
+LABEL org.opencontainers.image.base.name="docker.io/ubuntu:jammy"
+LABEL org.opencontainers.image.created="${CREATION_DATE}"
+LABEL org.opencontainers.image.licenses="OpenSSL"
+LABEL org.opencontainers.image.revision="${COMMIT}"
+LABEL org.opencontainers.image.ref.name="alphaunito/slurmctld"
+LABEL org.opencontainers.image.source="https://github.com/alpha-unito/docker-for-hpc"
+LABEL org.opencontainers.image.title="Slurm management daemon"
+LABEL org.opencontainers.image.version="21.08.5"
+
+RUN apt update                                                                             \
+ && apt install -y --no-install-recommends                                                 \
+        ca-certificates                                                                    \
+        hwloc                                                                              \
+        mailutils                                                                          \
+        munge                                                                              \
+        openmpi-bin                                                                        \
+        openssh-server                                                                     \
+        slurm-client                                                                       \
+        slurmctld                                                                          \
+        supervisor                                                                         \
+        wget                                                                               \
+ && rm -rf /var/lib/apt/lists/*                                                            \
+ && wget -O /bin/gosu https://github.com/tianon/gosu/releases/download/1.17/gosu-amd64     \
+ && chmod +x /bin/gosu                                                                     \
+ && gosu nobody true                                                                       \
+ && mkdir -p                                                                               \
+        /run/munge                                                                         \
+        /run/sshd                                                                          \
+        /var/log/munge                                                                     \
+        /var/log/slurm                                                                     \
+        /var/run/slurm                                                                     \
+        /var/spool/slurmctld                                                               \
+ && chown munge:munge                                                                      \
+        /run/munge                                                                         \
+        /var/log/munge                                                                     \
+ && chown slurm:slurm                                                                      \
+        /var/log/slurm                                                                     \
+        /var/run/slurm                                                                     \
+        /var/spool/slurmctld                                                               \
+ && adduser                                                                                \
+        --disabled-password                                                                \
+        --gecos ""                                                                         \
+        hpcuser
+
+COPY config/cgroups.conf                                                                   \
+     config/slurm.conf                                                                     \
+     /etc/slurm/
+
+COPY config/supervisord.conf                                                               \
+     /etc/supervisor/conf.d/
+
+COPY scripts/run-slurmctld                                                                 \
+     /bin/
+
+HEALTHCHECK --start-period=15s CMD                                                         \
+    test "$(supervisorctl status | awk '{print $2}' | sort | uniq)" == "RUNNING"
+
+EXPOSE 22 6817
+
+WORKDIR /home/hpcuser
+
+ENTRYPOINT supervisord

slurm/slurmctld/config/cgroups.conf

@@ -0,0 +1,7 @@
+CgroupAutomount=yes
+CgroupReleaseAgentDir="/etc/slurm/cgroup"
+ConstrainCores=yes
+ConstrainDevices=no
+ConstrainRAMSpace=yes
+ConstrainSwapSpace=yes
+IgnoreSystemd=yes

slurm/slurmctld/config/slurm.conf

@@ -0,0 +1,35 @@
+AccountingStorageType=accounting_storage/none
+ClusterName=docker-slurm
+InactiveLimit=0
+JobAcctGatherFrequency=30
+JobAcctGatherType=jobacct_gather/none
+JobCompType=jobcomp/none
+KillWait=30
+MinJobAge=300
+MpiDefault=none
+NodeName=__SLURMD_HOSTNAME_PREFIX__-[1-__SLURMD_NODES__] CPUs=__SLURM_CPUS_ON_NODE__ Sockets=__SLURM_SOCKETS_ON_NODE__ CoresPerSocket=__SLURM_CORES_PER_SOCKET_ON_NODE__ ThreadsPerCore=__SLURM_THREADS_PER_CORE__ State=UNKNOWN
+PartitionName=docker Nodes=ALL Default=YES MaxTime=INFINITE State=UP
+ProctrackType=proctrack/linuxproc
+ReturnToService=1
+SchedulerType=sched/backfill
+SelectType=select/cons_res
+SelectTypeParameters=CR_Core
+SlurmctldDebug=info
+SlurmctldHost=__SLURMCTLD_HOST__
+SlurmctldLogFile=/var/log/slurm/slurmctld.log
+SlurmctldParameters=enable_configless
+SlurmctldPidFile=/var/run/slurm/slurmctld.pid
+SlurmctldPort=6817
+SlurmctldTimeout=120
+SlurmdDebug=info
+SlurmdLogFile=/var/log/slurm/slurmd.log
+SlurmdPidFile=/var/run/slurm/slurmd.pid
+SlurmdPort=6818
+SlurmdSpoolDir=/var/spool/slurmd
+SlurmdTimeout=300
+SlurmdUser=root
+SlurmUser=slurm
+StateSaveLocation=/var/spool/slurmctld
+SwitchType=switch/none
+TreeWidth=65533
+Waittime=0

slurm/slurmctld/config/supervisord.conf

@@ -0,0 +1,25 @@
+[supervisord]
+nodaemon=true
+logfile=/dev/null
+logfile_maxbytes=0
+
+[program:munge]
+command=gosu munge /usr/sbin/munged --foreground
+autostart=true
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+redirect_stderr=true
+
+[program:slurmctld]
+command=run-slurmctld
+autostart=true
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+redirect_stderr=true
+
+[program:sshd]
+command=/usr/sbin/sshd -D
+autostart=true
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+redirect_stderr=true