Merge pull request #94 from dciangot/feature/startup-probe-support

feat: add support for Kubernetes startup probes

Author: dciangot

.gitignore

@@ -1,5 +1,7 @@
 envs.sh
 bin
+vendor/
+
 # Eclipse IDE
 .project
 .settings

examples/startup-probe-example.yaml

@@ -0,0 +1,42 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  name: startup-probe-test-pod
+  namespace: vk
+spec:
+  containers:
+  - name: app-container
+    image: nginx:alpine
+    ports:
+    - containerPort: 80
+    startupProbe:
+      httpGet:
+        path: /
+        port: 80
+        scheme: HTTP
+      initialDelaySeconds: 300
+      periodSeconds: 10
+      timeoutSeconds: 5
+      successThreshold: 1
+      failureThreshold: 3
+    readinessProbe:
+      httpGet:
+        path: /
+        port: 80
+        scheme: HTTP
+      initialDelaySeconds: 5
+      periodSeconds: 10
+      timeoutSeconds: 5
+      successThreshold: 1
+      failureThreshold: 3
+    livenessProbe:
+      httpGet:
+        path: /
+        port: 80
+        scheme: HTTP
+      initialDelaySeconds: 15
+      periodSeconds: 20
+      timeoutSeconds: 5
+      successThreshold: 1
+      failureThreshold: 3
+  restartPolicy: Always

pkg/slurm/Create.go

@@ -148,14 +148,15 @@ func (h *SidecarHandler) SubmitHandler(w http.ResponseWriter, r *http.Request) {
 		)
 
 		// Process probes if enabled
-		var readinessProbes, livenessProbes []ProbeCommand
+		var readinessProbes, livenessProbes, startupProbes []ProbeCommand
 		if h.Config.EnableProbes && !isInit {
-			readinessProbes, livenessProbes = translateKubernetesProbes(spanCtx, container)
-			if len(readinessProbes) > 0 || len(livenessProbes) > 0 {
+			readinessProbes, livenessProbes, startupProbes = translateKubernetesProbes(spanCtx, container)
+			if len(readinessProbes) > 0 || len(livenessProbes) > 0 || len(startupProbes) > 0 {
 				log.G(h.Ctx).Info("-- Container " + container.Name + " has probes configured")
 				span.SetAttributes(
 					attribute.Int("job.container"+strconv.Itoa(i)+".readiness_probes", len(readinessProbes)),
 					attribute.Int("job.container"+strconv.Itoa(i)+".liveness_probes", len(livenessProbes)),
+					attribute.Int("job.container"+strconv.Itoa(i)+".startup_probes", len(startupProbes)),
 				)
 			}
 		}
@@ -168,6 +169,7 @@ func (h *SidecarHandler) SubmitHandler(w http.ResponseWriter, r *http.Request) {
 			isInitContainer:  isInit,
 			readinessProbes:  readinessProbes,
 			livenessProbes:   livenessProbes,
+			startupProbes:    startupProbes,
 			containerImage:   image,
 		})
 	}

pkg/slurm/Status.go

@@ -212,7 +212,7 @@ func (h *SidecarHandler) StatusHandler(w http.ResponseWriter, r *http.Request) {
 						}
 						for _, ct := range pod.Spec.Containers {
 							// Check probe status for container readiness
-							readinessCount, _, err := loadProbeMetadata(path, ct.Name)
+							readinessCount, _, _, err := loadProbeMetadata(path, ct.Name)
 							isReady := true
 							if err != nil {
 								log.G(h.Ctx).Debug("Failed to load probe metadata for container ", ct.Name, ": ", err)
@@ -287,7 +287,7 @@ func (h *SidecarHandler) StatusHandler(w http.ResponseWriter, r *http.Request) {
 						}
 						for _, ct := range pod.Spec.Containers {
 							// Check probe status for container readiness
-							readinessCount, _, err := loadProbeMetadata(path, ct.Name)
+							readinessCount, _, _, err := loadProbeMetadata(path, ct.Name)
 							isReady := true
 							if err != nil {
 								log.G(h.Ctx).Debug("Failed to load probe metadata for container ", ct.Name, ": ", err)

pkg/slurm/func.go

@@ -98,6 +98,12 @@ func NewSlurmConfig() (SlurmConfig, error) {
 
 			SlurmConfigInst.Tsockspath = path
 		}
+
+		// Set default ContainerRuntime if not configured
+		if SlurmConfigInst.ContainerRuntime == "" {
+			SlurmConfigInst.ContainerRuntime = "singularity"
+		}
+
 		// Check if a supported container runtime is configured (supported: singularity, enroot)
 		if SlurmConfigInst.ContainerRuntime != "singularity" && SlurmConfigInst.ContainerRuntime != "enroot" {
 			err := fmt.Errorf("container runtime %q is not supported. Please configure a supported one (singularity, enroot)", SlurmConfigInst.ContainerRuntime)

pkg/slurm/prepare.go

@@ -836,15 +836,15 @@ highestExitCode=0
 	// Generate probe cleanup script first if any probes exist
 	var hasProbes bool
 	for _, containerCommand := range commands {
-		if len(containerCommand.readinessProbes) > 0 || len(containerCommand.livenessProbes) > 0 {
+		if len(containerCommand.readinessProbes) > 0 || len(containerCommand.livenessProbes) > 0 || len(containerCommand.startupProbes) > 0 {
 			hasProbes = true
 			break
 		}
 	}
 	if hasProbes && config.EnableProbes {
 		for _, containerCommand := range commands {
-			if len(containerCommand.readinessProbes) > 0 || len(containerCommand.livenessProbes) > 0 {
-				cleanupScript := generateProbeCleanupScript(containerCommand.containerName, containerCommand.readinessProbes, containerCommand.livenessProbes)
+			if len(containerCommand.readinessProbes) > 0 || len(containerCommand.livenessProbes) > 0 || len(containerCommand.startupProbes) > 0 {
+				cleanupScript := generateProbeCleanupScript(containerCommand.containerName, containerCommand.readinessProbes, containerCommand.livenessProbes, containerCommand.startupProbes)
 				stringToBeWritten.WriteString(cleanupScript)
 				break // Only need one cleanup script
 			}
@@ -898,7 +898,7 @@ highestExitCode=0
 		}
 
 		// Generate probe scripts if enabled and not an init container
-		if config.EnableProbes && !containerCommand.isInitContainer && (len(containerCommand.readinessProbes) > 0 || len(containerCommand.livenessProbes) > 0) {
+		if config.EnableProbes && !containerCommand.isInitContainer && (len(containerCommand.readinessProbes) > 0 || len(containerCommand.livenessProbes) > 0 || len(containerCommand.startupProbes) > 0) {
 			// Extract the image name from the singularity command
 			var imageName string
 			for i, arg := range containerCommand.runtimeCommand {
@@ -922,12 +922,12 @@ highestExitCode=0
 
 			if imageName != "" {
 				// Store probe metadata for status checking
-				err := storeProbeMetadata(path, containerCommand.containerName, len(containerCommand.readinessProbes), len(containerCommand.livenessProbes))
+				err := storeProbeMetadata(path, containerCommand.containerName, len(containerCommand.readinessProbes), len(containerCommand.livenessProbes), len(containerCommand.startupProbes))
 				if err != nil {
 					log.G(Ctx).Error("Failed to store probe metadata: ", err)
 				}
 
-				probeScript := generateProbeScript(Ctx, config, containerCommand.containerName, imageName, containerCommand.readinessProbes, containerCommand.livenessProbes)
+				probeScript := generateProbeScript(Ctx, config, containerCommand.containerName, imageName, containerCommand.readinessProbes, containerCommand.livenessProbes, containerCommand.startupProbes)
 				stringToBeWritten.WriteString("\n")
 				stringToBeWritten.WriteString(probeScript)
 			}
@@ -1381,7 +1381,7 @@ func prepareRuntimeCommand(config SlurmConfig, container v1.Container, metadata
 			singularityOptions = singOpts
 		}
 
-		// See https://github.com/interTwin-eu/interlink-slurm-plugin/issues/32#issuecomment-2416031030
+		// See https://github.com/interlink-hq/interlink-slurm-plugin/issues/32#issuecomment-2416031030
 		// singularity run will honor the entrypoint/command (if exist) in container image, while exec will override entrypoint.
 		// Thus if pod command (equivalent to container entrypoint) exist, we do exec, and other case we do run
 		singularityCommand := ""