Back merge main -> dev

VERSION

@@ -1 +1 @@
-1.21.3
+1.21.4

config/manager/kustomization.yaml

@@ -3,4 +3,4 @@ resources:
 images:
   - name: controller
     newName: cr.eu-north1.nebius.cloud/soperator/slurm-operator
-    newTag: 1.21.3
+    newTag: 1.21.4

config/manager/manager.yaml

@@ -87,7 +87,7 @@ spec:
             value: "false"
           - name: SLURM_OPERATOR_WATCH_NAMESPACES
             value: "*"
-        image: controller:1.21.3
+        image: controller:1.21.4
         imagePullPolicy: Always
         name: manager
         securityContext:

config/soperatorchecks/kustomization.yaml

@@ -4,7 +4,7 @@ resources:
 images:
   - name: controller
     newName: cr.eu-north1.nebius.cloud/soperator/soperatorchecks
-    newTag: 1.21.3
+    newTag: 1.21.4
 patches:
   # Protect the /metrics endpoint by putting it behind auth.
   # If you want your controller-manager to expose the /metrics

fluxcd/environment/nebius-cloud/dev/bootstrap/flux-kustomization.yaml

@@ -11,7 +11,7 @@ spec:
     name: nebius-cloud
   postBuild:
     substitute:
-      soperator_version: 1.21.3
+      soperator_version: 1.21.4
   path: "./fluxcd/enviroment/nebius-cloud/dev"
   prune: true
   timeout: 1m

fluxcd/environment/nebius-cloud/prod/bootstrap/flux-kustomization.yaml

@@ -10,7 +10,7 @@ spec:
     name: nebius-cloud
   postBuild:
     substitute:
-      soperator_version: 1.21.3
+      soperator_version: 1.21.4
   path: "./fluxcd/enviroment/nebius-cloud/prod"
   prune: false
   timeout: 1m

helm/nodeconfigurator/Chart.yaml

@@ -13,9 +13,9 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 1.21.3
+version: 1.21.4
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to
 # follow Semantic Versioning. They should reflect the version the application is using.
 # It is recommended to use it with quotes.
-appVersion: "1.21.3"
+appVersion: "1.21.4"

helm/nodeconfigurator/values.yaml

@@ -22,7 +22,7 @@ rebooter:
   env: []
   image:
     repository: "cr.eu-north1.nebius.cloud/soperator/rebooter"
-    tag: "1.21.3"
+    tag: "1.21.4"
     pullPolicy: IfNotPresent
   nodeSelector: {}
   resources: {}

helm/slurm-cluster-storage/Chart.yaml

@@ -2,5 +2,5 @@ apiVersion: v2
 name: helm-slurm-cluster-storage
 description: A Helm chart for Kubernetes
 type: application
-version: "1.21.3"
-appVersion: "1.21.3"
+version: "1.21.4"
+appVersion: "1.21.4"

helm/slurm-cluster/Chart.yaml

@@ -2,6 +2,6 @@ apiVersion: v2
 name: helm-slurm-cluster
 description: A Helm chart for Kubernetes
 type: application
-version: "1.21.3"
-appVersion: "1.21.3"
+version: "1.21.4"
+appVersion: "1.21.4"
 kubeVersion: ">=1.29.0-0"

helm/slurm-cluster/templates/slurm-scripts-cm.yaml

@@ -40,6 +40,9 @@ data:
   health_checker.sh: |
 {{ tpl (.Files.Get "slurm_scripts/health_checker.sh") . | indent 4 }}
 
+  boot_disk_full.sh: |
+{{ tpl (.Files.Get "slurm_scripts/boot_disk_full.sh") . | indent 4 }}
+
 {{- if .Values.slurmScripts.custom }}
 {{- range $name, $content := .Values.slurmScripts.custom }}
   {{ $name }}: |

helm/slurm-cluster/values.yaml

@@ -510,16 +510,16 @@ telemetry: {}
 #   otelCollectorPort: 8429
 
 images:
-  slurmctld: "cr.eu-north1.nebius.cloud/soperator/controller_slurmctld:1.21.3-jammy-slurm24.11.5"
-  slurmrestd: "cr.eu-north1.nebius.cloud/soperator/slurmrestd:1.21.3-jammy-slurm24.11.5"
-  slurmd: "cr.eu-north1.nebius.cloud/soperator/worker_slurmd:1.21.3-jammy-slurm24.11.5"
-  sshd: "cr.eu-north1.nebius.cloud/soperator/login_sshd:1.21.3-jammy-slurm24.11.5"
-  munge: "cr.eu-north1.nebius.cloud/soperator/munge:1.21.3-jammy-slurm24.11.5"
-  populateJail: "cr.eu-north1.nebius.cloud/soperator/populate_jail:1.21.3-jammy-slurm24.11.5"
-  ncclBenchmark: "cr.eu-north1.nebius.cloud/soperator/nccl_benchmark:1.21.3-jammy-slurm24.11.5"
-  slurmdbd: "cr.eu-north1.nebius.cloud/soperator/controller_slurmdbd:1.21.3-jammy-slurm24.11.5"
-  soperatorExporter: "cr.eu-north1.nebius.cloud/soperator/soperator-exporter:1.21.3-jammy-slurm24.11.5"
-  sConfigController: cr.eu-north1.nebius.cloud/soperator/sconfigcontroller:1.21.3
+  slurmctld: "cr.eu-north1.nebius.cloud/soperator/controller_slurmctld:1.21.4-jammy-slurm24.11.5"
+  slurmrestd: "cr.eu-north1.nebius.cloud/soperator/slurmrestd:1.21.4-jammy-slurm24.11.5"
+  slurmd: "cr.eu-north1.nebius.cloud/soperator/worker_slurmd:1.21.4-jammy-slurm24.11.5"
+  sshd: "cr.eu-north1.nebius.cloud/soperator/login_sshd:1.21.4-jammy-slurm24.11.5"
+  munge: "cr.eu-north1.nebius.cloud/soperator/munge:1.21.4-jammy-slurm24.11.5"
+  populateJail: "cr.eu-north1.nebius.cloud/soperator/populate_jail:1.21.4-jammy-slurm24.11.5"
+  ncclBenchmark: "cr.eu-north1.nebius.cloud/soperator/nccl_benchmark:1.21.4-jammy-slurm24.11.5"
+  slurmdbd: "cr.eu-north1.nebius.cloud/soperator/controller_slurmdbd:1.21.4-jammy-slurm24.11.5"
+  soperatorExporter: "cr.eu-north1.nebius.cloud/soperator/soperator-exporter:1.21.4-jammy-slurm24.11.5"
+  sConfigController: cr.eu-north1.nebius.cloud/soperator/sconfigcontroller:1.21.4
   mariaDB: docker-registry1.mariadb.com/library/mariadb:11.4.3
 # Configuration for slurm scripts.
 slurmScripts:

helm/soperator-activechecks/Chart.yaml

@@ -2,5 +2,5 @@ apiVersion: v2
 name: helm-soperator-activechecks
 description: A Helm chart for Kubernetes
 type: application
-version: "1.21.3"
-appVersion: "1.21.3"
+version: "1.21.4"
+appVersion: "1.21.4"

helm/soperator-activechecks/values.yaml

@@ -41,6 +41,6 @@ activeCheck:
     setCondition: false
     drainSlurmNode: false
 images:
-  slurmJob: "cr.eu-north1.nebius.cloud/soperator/slurm_check_job:1.21.3-jammy-slurm24.11.5"
+  slurmJob: "cr.eu-north1.nebius.cloud/soperator/slurm_check_job:1.21.4-jammy-slurm24.11.5"
   k8sJob: "cr.eu-north1.nebius.cloud/soperator/ubuntu:jammy"
-  munge: "cr.eu-north1.nebius.cloud/soperator/munge:1.21.3-jammy-slurm24.11.5"
+  munge: "cr.eu-north1.nebius.cloud/soperator/munge:1.21.4-jammy-slurm24.11.5"

helm/soperator-crds/Chart.yaml

@@ -2,6 +2,6 @@ apiVersion: v2
 name: helm-soperator-crds
 description: A Helm chart for Kubernetes
 type: application
-version: 1.21.3
-appVersion: "1.21.3"
+version: 1.21.4
+appVersion: "1.21.4"
 kubeVersion: ">=1.29.0-0"

helm/soperator-dcgm-exporter/Chart.yaml

@@ -2,5 +2,5 @@ apiVersion: v2
 name: helm-soperator-dcgm-exporter
 description: A Helm chart for Nvidia DCGM Exporter
 type: application
-version: 1.21.3
-appVersion: "1.21.3"
+version: 1.21.4
+appVersion: "1.21.4"

helm/soperator-fluxcd/Chart.yaml

@@ -13,9 +13,9 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 1.21.3
+version: 1.21.4
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to
 # follow Semantic Versioning. They should reflect the version the application is using.
 # It is recommended to use it with quotes.
-appVersion: "1.21.3"
+appVersion: "1.21.4"

helm/soperator-fluxcd/values.yaml

@@ -183,7 +183,7 @@ observability:
     enabled: true
     interval: 5m
     timeout: 5m
-    version: 1.21.3
+    version: 1.21.4
     namespace: soperator
     releaseName: soperator-dcgm-exporter
     values:
@@ -202,7 +202,7 @@ slurmCluster:
   enabled: true
   interval: 5m
   timeout: 5m
-  version: 1.21.3
+  version: 1.21.4
   namespace: soperator
   releaseName: soperator
   values: null
@@ -218,7 +218,7 @@ soperator:
   enabled: true
   interval: 5m
   timeout: 5m
-  version: 1.21.3
+  version: 1.21.4
   namespace: "soperator-system"
   releaseName: soperator-controller
   values:

helm/soperator/Chart.yaml

@@ -2,8 +2,8 @@ apiVersion: v2
 name: helm-soperator
 description: A Helm chart for Kubernetes
 type: application
-version: 1.21.3
-appVersion: "1.21.3"
+version: 1.21.4
+appVersion: "1.21.4"
 kubeVersion: ">=1.29.0-0"
 dependencies:
   - name: kruise

helm/soperator/values.yaml

@@ -38,7 +38,7 @@ controllerManager:
       slurmOperatorWatchNamespaces: '*'
     image:
       repository: cr.eu-north1.nebius.cloud/soperator/slurm-operator
-      tag: 1.21.3
+      tag: 1.21.4
     imagePullPolicy: Always
     resources:
       limits:

helm/soperatorchecks/Chart.yaml

@@ -13,9 +13,9 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 1.21.3
+version: 1.21.4
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to
 # follow Semantic Versioning. They should reflect the version the application is using.
 # It is recommended to use it with quotes.
-appVersion: "1.21.3"
+appVersion: "1.21.4"

helm/soperatorchecks/values.yaml

@@ -32,7 +32,7 @@ checks:
       slurmOperatorWatchNamespaces: '*'
     image:
       repository: cr.eu-north1.nebius.cloud/soperator/soperatorchecks
-      tag: 1.21.3
+      tag: 1.21.4
     imagePullPolicy: Always
     resources:
       limits:

internal/consts/version.go

@@ -2,5 +2,5 @@
 package consts
 
 const (
-	VersionCR = "1.21.3"
+	VersionCR = "1.21.4"
 )