make size simpler - just is number of worker nodes

Signed-off-by: vsoch <vsoch@users.noreply.github.com>

Author: vsoch

api/v1alpha1/slurm_types.go

@@ -63,7 +63,7 @@ type SlurmSpec struct {
 	// +optional
 	SlurmVersion string `json:"slurmVersion,omitempty"`
 
-	// Size of the slurm (1 server + (N-1) nodes)
+	// Size is number of worker nodes
 	Size int32 `json:"size"`
 
 	// Interactive mode keeps the cluster running
@@ -230,8 +230,8 @@ func (s *Slurm) SelectorName() string {
 
 // Validate the slurm
 func (s *Slurm) Validate() bool {
-	if s.WorkerNodes() < 1 {
-		fmt.Printf("😥️ Slurm cluster must have at least one worker node, Size >= 2.\n")
+	if s.Spec.Size < 1 {
+		fmt.Printf("😥️ Slurm cluster must have 1 or more worker nodes.\n")
 		return false
 	}
 	// Ensure we have the default image set
@@ -258,12 +258,6 @@ func (s *Slurm) Validate() bool {
 	return true
 }
 
-// WorkerNodes returns the number of worker nodes
-// At this point we've already validated size is >= 1
-func (s *Slurm) WorkerNodes() int32 {
-	return s.Spec.Size - 1
-}
-
 // WorkerNode returns the worker node (if defined) or falls back to the server
 func (s *Slurm) WorkerNode() Node {
 

config/crd/bases/flux-framework.org_slurms.yaml

@@ -244,7 +244,7 @@ spec:
                 description: Resources include limits and requests
                 type: object
               size:
-                description: Size of the slurm (1 server + (N-1) nodes)
+                description: Size is number of worker nodes
                 format: int32
                 type: integer
               slurmVersion:

controllers/slurm/jobset.go

@@ -93,9 +93,8 @@ func (r *SlurmReconciler) newJobSet(
 
 	}
 
-	// Create a cluster (JobSet) with workers (required)
-	workerNodes := cluster.WorkerNodes()
-	workerJob, err := r.getJob(cluster, cluster.WorkerNode(), workerNodes, "w", true)
+	// Create a cluster (JobSet) with workers
+	workerJob, err := r.getJob(cluster, cluster.WorkerNode(), cluster.Spec.Size, "w", true)
 	if err != nil {
 		r.Log.Error(err, "There was an error getting the worker ReplicatedJob")
 		return &jobs, err

controllers/slurm/templates.go

@@ -95,7 +95,7 @@ func generateHostlist(cluster *api.Slurm) string {
 	hosts := ""
 
 	serviceName := cluster.ServiceName()
-	for i := 0; i < int(cluster.WorkerNodes()); i++ {
+	for i := 0; i < int(cluster.Spec.Size); i++ {
 		if hosts == "" {
 			hosts = fmt.Sprintf("%s-w-0-%d.%s.%s.svc.cluster.local", cluster.Name, i, serviceName, cluster.Namespace)
 		} else {

docker/Dockerfile

@@ -39,12 +39,22 @@ RUN set -ex \
        bash-completion \
        vim-enhanced \
        http-parser-devel \
+       hwloc-devel \
+       libevent-devel \
        json-c-devel \
     && yum clean all \
     && rm -rf /var/cache/yum
 
 RUN pip3 install Cython nose
 
+RUN set -x \
+    && wget https://github.com/openpmix/openpmix/releases/download/v3.2.3/pmix-3.2.3.tar.gz \
+    && tar -xvf pmix-3.2.3.tar.gz \
+    && cd pmix-3.2.3 \
+    && ./configure --prefix=/opt/pmix --disable-static \
+    && make -j$(nproc) install \
+    && cd .. && rm -rf pmix-3.2.3*
+
 RUN set -ex \
     && wget -O /usr/local/bin/gosu "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-${ARCH}" \
     && wget -O /usr/local/bin/gosu.asc "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-${ARCH}.asc" \
@@ -58,8 +68,15 @@ RUN set -ex \
 RUN set -x \
     && git clone -b ${SLURM_TAG} --single-branch --depth=1 https://github.com/SchedMD/slurm.git \
     && pushd slurm \
+    && export CFLAGS="-I/usr/include" \
+    && export LDFLAGS="-L/usr/lib/$(uname -m)-linux-gnu" \
     && ./configure --enable-debug --prefix=/usr --sysconfdir=/etc/slurm \
-        --with-mysql_config=/usr/bin  --libdir=/usr/lib64 \
+        --with-mysql_config=/usr/bin \
+        --libdir=/usr/lib \
+        --with-pmix=/opt/pmix \
+        --with-hwloc=/usr \
+        --without-hdf5 \        
+        --with-libevent=/usr \        
     && make install \
     && install -D -m644 etc/cgroup.conf.example /etc/slurm/cgroup.conf.example \
     && install -D -m644 etc/slurm.conf.example /etc/slurm/slurm.conf.example \

docker/Dockerfile.ubuntu

@@ -40,18 +40,35 @@ RUN set -ex \
        build-essential \
        pkg-config \
        gosu \
+       libhwloc-dev \
+       libevent-dev \
     && apt-get clean \
     && rm -rf /var/lib/apt/lists/*
 
 # Note: Ubuntu 24.04 uses PEP 668 (externally managed environments). 
 # If pip install fails, use --break-system-packages or install in a venv.
 RUN pip3 install --break-system-packages Cython nose
 
+RUN set -x \
+    && wget https://github.com/openpmix/openpmix/releases/download/v3.2.3/pmix-3.2.3.tar.gz \
+    && tar -xvf pmix-3.2.3.tar.gz \
+    && cd pmix-3.2.3 \
+    && ./configure --prefix=/opt/pmix --disable-static \
+    && make -j$(nproc) install \
+    && cd .. && rm -rf pmix-3.2.3*
+
 RUN set -x \
     && git clone -b ${SLURM_TAG} --single-branch --depth=1 https://github.com/SchedMD/slurm.git \
-    && cd slurm \
+    && export CFLAGS="-I/usr/include" \
+    && export LDFLAGS="-L/usr/lib/$(uname -m)-linux-gnu" \
+    && cd ./slurm \
     && ./configure --enable-debug --prefix=/usr --sysconfdir=/etc/slurm \
-        --with-mysql_config=/usr/bin --libdir=/usr/lib \
+        --with-mysql_config=/usr/bin \
+        --libdir=/usr/lib \
+        --with-pmix=/opt/pmix \
+        --without-hdf5 \
+        --with-hwloc=/usr \
+        --with-libevent=/usr \        
     && make install \
     && install -D -m644 etc/cgroup.conf.example /etc/slurm/cgroup.conf.example \
     && install -D -m644 etc/slurm.conf.example /etc/slurm/slurm.conf.example \
@@ -96,4 +113,4 @@ EXPOSE 6717
 EXPOSE 6718
 EXPOSE 6719
 
-CMD ["slurmdbd"]
+CMD ["slurmdbd"]

examples/dist/slurm-operator.yaml

@@ -256,7 +256,7 @@ spec:
                 description: Resources include limits and requests
                 type: object
               size:
-                description: Size of the slurm (1 server + (N-1) nodes)
+                description: Size is number of worker nodes
                 format: int32
                 type: integer
               slurmVersion: