Skip to content

Commit

Permalink
Merge pull request #179 from nebius/dev
Browse files Browse the repository at this point in the history 
Release 1.15.2
  • Loading branch information
@asteny
asteny authored Nov 12, 2024
2 parents 8766049 + 4ceb478 commit ee4743f
Show file tree
Hide file tree
Showing 20 changed files with 120 additions and 89 deletions.
21 changes: 0 additions & 21 deletions .github/workflows/auto-assign-review.yml
View file

This file was deleted.

1 change: 1 addition & 0 deletions .github/workflows/one_job.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@ on:
paths-ignore:
- '.github/**'
- 'docs/**'
- 'CODEOWNERS'
- 'LICENSE'
- 'PROJECT'
- 'README.md'
Expand Down
2 changes: 2 additions & 0 deletions CODEOWNERS
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
# Global code owners (applies to the whole repo)
* @dstaroff @asteny @rdjjke @Uburro
2 changes: 1 addition & 1 deletion VERSION
View file
Original file line number Diff line number Diff line change
@@ -1 +1 @@
1.15.1
1.15.2
14 changes: 13 additions & 1 deletion api/v1/slurmcluster_types.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -134,6 +134,12 @@ type PopulateJail struct {
// +kubebuilder:validation:Optional
// +kubebuilder:default=false
Overwrite bool `json:"overwrite"`

// AppArmorProfile defines the AppArmor profile for the Slurm node
//
// +kubebuilder:validation:Optional
// +kubebuilder:default="unconfined"
AppArmorProfile string `json:"appArmorProfile,omitempty"`
}

// PeriodicChecks define the k8s CronJobs performing cluster checks
Expand Down Expand Up @@ -204,6 +210,12 @@ type NCCLBenchmark struct {
//
// +kubebuilder:validation:Required
K8sNodeFilterName string `json:"k8sNodeFilterName"`

// AppArmorProfile defines the AppArmor profile for the Slurm node
//
// +kubebuilder:validation:Optional
// +kubebuilder:default="unconfined"
AppArmorProfile string `json:"appArmorProfile,omitempty"`
}

// NCCLArguments define nccl settings for periodic nccl benchmark
Expand Down Expand Up @@ -749,7 +761,7 @@ type NodeContainer struct {
// +kubebuilder:validation:Optional
SecurityLimitsConfig string `json:"securityLimitsConfig,omitempty"`

// AppArmorProfile defines the AppArmor profile for the Slurm worker node
// AppArmorProfile defines the AppArmor profile for the Slurm containers
//
// +kubebuilder:validation:Optional
// +kubebuilder:default="unconfined"
Expand Down
34 changes: 22 additions & 12 deletions config/crd/bases/slurm.nebius.ai_slurmclusters.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1123,6 +1123,11 @@ spec:
in seconds
format: int64
type: integer
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm node
type: string
enabled:
default: true
description: Enabled defines whether the CronJob should be
Expand Down Expand Up @@ -1219,6 +1224,11 @@ spec:
description: PopulateJail defines the k8s Job that performs initial
jail file system population
properties:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile for
the Slurm node
type: string
image:
description: Image defines the populate jail container image
type: string
Expand Down Expand Up @@ -1524,7 +1534,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
enabled:
type: boolean
Expand Down Expand Up @@ -8289,7 +8299,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -8376,7 +8386,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -8532,7 +8542,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -8579,7 +8589,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -9072,7 +9082,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -9123,7 +9133,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -9612,7 +9622,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -9664,7 +9674,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -9993,7 +10003,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -10057,7 +10067,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down Expand Up @@ -10103,7 +10113,7 @@ spec:
appArmorProfile:
default: unconfined
description: AppArmorProfile defines the AppArmor profile
for the Slurm worker node
for the Slurm containers
type: string
image:
description: Image defines the container image
Expand Down
2 changes: 1 addition & 1 deletion config/manager/kustomization.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,4 +3,4 @@ resources:
images:
- name: controller
newName: cr.eu-north1.nebius.cloud/soperator/slurm-operator
newTag: 1.15.1
newTag: 1.15.2
2 changes: 1 addition & 1 deletion config/manager/manager.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -82,7 +82,7 @@ spec:
value: "false"
- name: SLURM_OPERATOR_WATCH_NAMESPACES
value: "*"
image: controller:1.15.1
image: controller:1.15.2
imagePullPolicy: Always
name: manager
securityContext:
Expand Down
4 changes: 2 additions & 2 deletions helm/slurm-cluster-storage/Chart.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,5 +2,5 @@ apiVersion: v2
name: helm-slurm-cluster-storage
description: A Helm chart for Kubernetes
type: application
version: "1.15.1"
appVersion: "1.15.1"
version: "1.15.2"
appVersion: "1.15.2"
4 changes: 2 additions & 2 deletions helm/slurm-cluster/Chart.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,5 +2,5 @@ apiVersion: v2
name: helm-slurm-cluster
description: A Helm chart for Kubernetes
type: application
version: "1.15.1"
appVersion: "1.15.1"
version: "1.15.2"
appVersion: "1.15.2"
3 changes: 1 addition & 2 deletions helm/slurm-cluster/templates/slurm-cluster-cr.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -40,8 +40,7 @@ spec:
- name: {{ .name | quote }}
{{- omit . "name" | toYaml | nindent 6 }}
{{- end }}
secrets:
sshdKeysName: {{ include "slurm-cluster.secret.sshdKeysName" . }}
secrets: {{ toYaml .Values.secrets | nindent 4 }}
populateJail:
image: {{ required "populateJail image" .Values.images.populateJail | quote }}
imagePullPolicy: {{ default "IfNotPresent" .Values.populateJail.imagePullPolicy | quote }}
Expand Down
24 changes: 12 additions & 12 deletions helm/slurm-cluster/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -61,9 +61,9 @@ volumeSources:
# readOnly: false

# Secret references needed for Slurm cluster operation
secrets:
# Secret reference required for login sshd. If secret name empty - operator generate own secret with keys
sshdKeysName: ""
secrets: {}
# Secret reference required for login sshd. If secret name empty - operator generate own secret with keys
# sshdKeysName: ""
# Job performing initial jail file system population
populateJail:
imagePullPolicy: "IfNotPresent"
Expand Down Expand Up @@ -380,13 +380,13 @@ telemetry: {}
# otelCollectorPort: 8429

images:
slurmctld: "cr.eu-north1.nebius.cloud/soperator/controller_slurmctld:1.15.1-jammy-slurm24.05.2"
slurmrestd: "cr.eu-north1.nebius.cloud/soperator/slurmrestd:1.15.1-jammy-slurm24.05.2"
slurmd: "cr.eu-north1.nebius.cloud/soperator/worker_slurmd:1.15.1-jammy-slurm24.05.2"
sshd: "cr.eu-north1.nebius.cloud/soperator/login_sshd:1.15.1-jammy-slurm24.05.2"
munge: "cr.eu-north1.nebius.cloud/soperator/munge:1.15.1-jammy-slurm24.05.2"
populateJail: "cr.eu-north1.nebius.cloud/soperator/populate_jail:1.15.1-jammy-slurm24.05.2"
ncclBenchmark: "cr.eu-north1.nebius.cloud/soperator/nccl_benchmark:1.15.1-jammy-slurm24.05.2"
slurmdbd: "cr.eu-north1.nebius.cloud/soperator/controller_slurmdbd:1.15.1-jammy-slurm24.05.2"
exporter: "cr.eu-north1.nebius.cloud/soperator/exporter:1.15.1-jammy-slurm24.05.2"
slurmctld: "cr.eu-north1.nebius.cloud/soperator/controller_slurmctld:1.15.2-jammy-slurm24.05.2"
slurmrestd: "cr.eu-north1.nebius.cloud/soperator/slurmrestd:1.15.2-jammy-slurm24.05.2"
slurmd: "cr.eu-north1.nebius.cloud/soperator/worker_slurmd:1.15.2-jammy-slurm24.05.2"
sshd: "cr.eu-north1.nebius.cloud/soperator/login_sshd:1.15.2-jammy-slurm24.05.2"
munge: "cr.eu-north1.nebius.cloud/soperator/munge:1.15.2-jammy-slurm24.05.2"
populateJail: "cr.eu-north1.nebius.cloud/soperator/populate_jail:1.15.2-jammy-slurm24.05.2"
ncclBenchmark: "cr.eu-north1.nebius.cloud/soperator/nccl_benchmark:1.15.2-jammy-slurm24.05.2"
slurmdbd: "cr.eu-north1.nebius.cloud/soperator/controller_slurmdbd:1.15.2-jammy-slurm24.05.2"
exporter: "cr.eu-north1.nebius.cloud/soperator/exporter:1.15.2-jammy-slurm24.05.2"
mariaDB: "docker-registry1.mariadb.com/library/mariadb:11.4.3"
4 changes: 2 additions & 2 deletions helm/soperator-crds/Chart.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,5 +2,5 @@ apiVersion: v2
name: helm-soperator-crds
description: A Helm chart for Kubernetes
type: application
version: 1.15.1
appVersion: "1.15.1"
version: 1.15.2
appVersion: "1.15.2"
Loading

0 comments on commit ee4743f

Please sign in to comment.