Add init containers, shared namespaces, per-volume sizing, and user/label support (#997)

Reviewed-on: #997
Co-authored-by: Prathamesh Musale <prathamesh.musale0@gmail.com>
Co-committed-by: Prathamesh Musale <prathamesh.musale0@gmail.com>

stack_orchestrator/data/stacks/test/deploy/commands.py

@@ -21,7 +21,7 @@ from stack_orchestrator.deploy.deploy_util import VolumeMapping, run_container_c
 from pathlib import Path
 
 default_spec_file_content = """config:
-    test-variable-1: test-value-1
+    test_variable_1: test-value-1
 """
 
 

stack_orchestrator/deploy/k8s/cluster_info.py

@@ -273,19 +273,25 @@ class ClusterInfo:
         result = []
         spec_volumes = self.spec.get_volumes()
         named_volumes = self._all_named_volumes()
-        resources = self.spec.get_volume_resources()
-        if not resources:
-            resources = DEFAULT_VOLUME_RESOURCES
+        global_resources = self.spec.get_volume_resources()
+        if not global_resources:
+            global_resources = DEFAULT_VOLUME_RESOURCES
         if opts.o.debug:
             print(f"Spec Volumes: {spec_volumes}")
             print(f"Named Volumes: {named_volumes}")
-            print(f"Resources: {resources}")
+            print(f"Resources: {global_resources}")
         for volume_name, volume_path in spec_volumes.items():
             if volume_name not in named_volumes:
                 if opts.o.debug:
                     print(f"{volume_name} not in pod files")
                 continue
 
+            # Per-volume resources override global, which overrides default.
+            vol_resources = (
+                self.spec.get_volume_resources_for(volume_name)
+                or global_resources
+            )
+
             labels = {
                 "app": self.app_name,
                 "volume-label": f"{self.app_name}-{volume_name}",
@@ -301,7 +307,7 @@ class ClusterInfo:
             spec = client.V1PersistentVolumeClaimSpec(
                 access_modes=["ReadWriteOnce"],
                 storage_class_name=storage_class_name,
-                resources=to_k8s_resource_requirements(resources),
+                resources=to_k8s_resource_requirements(vol_resources),
                 volume_name=k8s_volume_name,
             )
             pvc = client.V1PersistentVolumeClaim(
@@ -353,9 +359,9 @@ class ClusterInfo:
         result = []
         spec_volumes = self.spec.get_volumes()
         named_volumes = self._all_named_volumes()
-        resources = self.spec.get_volume_resources()
-        if not resources:
-            resources = DEFAULT_VOLUME_RESOURCES
+        global_resources = self.spec.get_volume_resources()
+        if not global_resources:
+            global_resources = DEFAULT_VOLUME_RESOURCES
         for volume_name, volume_path in spec_volumes.items():
             # We only need to create a volume if it is fully qualified HostPath.
             # Otherwise, we create the PVC and expect the node to allocate the volume
@@ -384,6 +390,10 @@ class ClusterInfo:
                     )
                     continue
 
+            vol_resources = (
+                self.spec.get_volume_resources_for(volume_name)
+                or global_resources
+            )
             if self.spec.is_kind_deployment():
                 host_path = client.V1HostPathVolumeSource(
                     path=get_kind_pv_bind_mount_path(volume_name)
@@ -393,7 +403,7 @@ class ClusterInfo:
             spec = client.V1PersistentVolumeSpec(
                 storage_class_name="manual",
                 access_modes=["ReadWriteOnce"],
-                capacity=to_k8s_resource_requirements(resources).requests,
+                capacity=to_k8s_resource_requirements(vol_resources).requests,
                 host_path=host_path,
             )
             pv = client.V1PersistentVolume(
@@ -446,12 +456,16 @@ class ClusterInfo:
     ) -> tuple:
         """Build k8s container specs from parsed compose YAML.
 
-        Returns a tuple of (containers, services, volumes) where:
+        Returns a tuple of (containers, init_containers, services, volumes)
+        where:
         - containers: list of V1Container objects
+        - init_containers: list of V1Container objects for init containers
+          (compose services with label ``laconic.init-container: "true"``)
         - services: the last services dict processed (used for annotations/labels)
         - volumes: list of V1Volume objects
         """
         containers = []
+        init_containers = []
         services = {}
         global_resources = self.spec.get_container_resources()
         if not global_resources:
@@ -563,6 +577,7 @@ class ClusterInfo:
                     volume_mounts=volume_mounts,
                     security_context=client.V1SecurityContext(
                         privileged=self.spec.get_privileged(),
+                        run_as_user=int(service_info["user"]) if "user" in service_info else None,
                         capabilities=client.V1Capabilities(
                             add=self.spec.get_capabilities()
                         )
@@ -571,15 +586,29 @@ class ClusterInfo:
                     ),
                     resources=to_k8s_resource_requirements(container_resources),
                 )
-                containers.append(container)
+                # Services with laconic.init-container label become
+                # k8s init containers instead of regular containers.
+                svc_labels = service_info.get("labels", {})
+                if isinstance(svc_labels, list):
+                    # docker-compose labels can be a list of "key=value"
+                    svc_labels = dict(
+                        item.split("=", 1) for item in svc_labels
+                    )
+                is_init = str(
+                    svc_labels.get("laconic.init-container", "")
+                ).lower() in ("true", "1", "yes")
+                if is_init:
+                    init_containers.append(container)
+                else:
+                    containers.append(container)
         volumes = volumes_for_pod_files(
             parsed_yaml_map, self.spec, self.app_name
         )
-        return containers, services, volumes
+        return containers, init_containers, services, volumes
 
     # TODO: put things like image pull policy into an object-scope struct
     def get_deployment(self, image_pull_policy: Optional[str] = None):
-        containers, services, volumes = self._build_containers(
+        containers, init_containers, services, volumes = self._build_containers(
             self.parsed_pod_yaml_map, image_pull_policy
         )
         registry_config = self.spec.get_image_registry_config()
@@ -650,6 +679,7 @@ class ClusterInfo:
             metadata=client.V1ObjectMeta(annotations=annotations, labels=labels),
             spec=client.V1PodSpec(
                 containers=containers,
+                init_containers=init_containers or None,
                 image_pull_secrets=image_pull_secrets,
                 volumes=volumes,
                 affinity=affinity,
@@ -668,7 +698,10 @@ class ClusterInfo:
         deployment = client.V1Deployment(
             api_version="apps/v1",
             kind="Deployment",
-            metadata=client.V1ObjectMeta(name=f"{self.app_name}-deployment"),
+            metadata=client.V1ObjectMeta(
+                name=f"{self.app_name}-deployment",
+                labels={"app": self.app_name, **({"app.kubernetes.io/stack": self.stack_name} if self.stack_name else {})},
+            ),
             spec=spec,
         )
         return deployment
@@ -695,8 +728,8 @@ class ClusterInfo:
         for job_file in self.parsed_job_yaml_map:
             # Build containers for this single job file
             single_job_map = {job_file: self.parsed_job_yaml_map[job_file]}
-            containers, _services, volumes = self._build_containers(
-                single_job_map, image_pull_policy
+            containers, init_containers, _services, volumes = (
+                self._build_containers(single_job_map, image_pull_policy)
             )
 
             # Derive job name from file path: docker-compose-<name>.yml -> <name>
@@ -722,6 +755,7 @@ class ClusterInfo:
                 ),
                 spec=client.V1PodSpec(
                     containers=containers,
+                    init_containers=init_containers or None,
                     image_pull_secrets=image_pull_secrets,
                     volumes=volumes,
                     restart_policy="Never",

stack_orchestrator/deploy/k8s/deploy_k8s.py

@@ -122,9 +122,9 @@ class K8sDeployer(Deployer):
             return
         self.deployment_dir = deployment_context.deployment_dir
         self.deployment_context = deployment_context
-        self.kind_cluster_name = compose_project_name
-        # Use deployment-specific namespace for resource isolation and easy cleanup
-        self.k8s_namespace = f"laconic-{compose_project_name}"
+        self.kind_cluster_name = deployment_context.spec.get_kind_cluster_name() or compose_project_name
+        # Use spec namespace if provided, otherwise derive from cluster-id
+        self.k8s_namespace = deployment_context.spec.get_namespace() or f"laconic-{compose_project_name}"
         self.cluster_info = ClusterInfo()
         # stack.name may be an absolute path (from spec "stack:" key after
         # path resolution). Extract just the directory basename for labels.
@@ -204,6 +204,93 @@ class K8sDeployer(Deployer):
             else:
                 raise
 
+    def _delete_resources_by_label(self, label_selector: str, delete_volumes: bool):
+        """Delete only this stack's resources from a shared namespace."""
+        ns = self.k8s_namespace
+        if opts.o.dry_run:
+            print(f"Dry run: would delete resources with {label_selector} in {ns}")
+            return
+
+        # Deployments
+        try:
+            deps = self.apps_api.list_namespaced_deployment(
+                namespace=ns, label_selector=label_selector
+            )
+            for dep in deps.items:
+                print(f"Deleting Deployment {dep.metadata.name}")
+                self.apps_api.delete_namespaced_deployment(
+                    name=dep.metadata.name, namespace=ns
+                )
+        except ApiException as e:
+            _check_delete_exception(e)
+
+        # Jobs
+        try:
+            jobs = self.batch_api.list_namespaced_job(
+                namespace=ns, label_selector=label_selector
+            )
+            for job in jobs.items:
+                print(f"Deleting Job {job.metadata.name}")
+                self.batch_api.delete_namespaced_job(
+                    name=job.metadata.name, namespace=ns,
+                    body=client.V1DeleteOptions(propagation_policy="Background"),
+                )
+        except ApiException as e:
+            _check_delete_exception(e)
+
+        # Services (NodePorts created by SO)
+        try:
+            svcs = self.core_api.list_namespaced_service(
+                namespace=ns, label_selector=label_selector
+            )
+            for svc in svcs.items:
+                print(f"Deleting Service {svc.metadata.name}")
+                self.core_api.delete_namespaced_service(
+                    name=svc.metadata.name, namespace=ns
+                )
+        except ApiException as e:
+            _check_delete_exception(e)
+
+        # Ingresses
+        try:
+            ings = self.networking_api.list_namespaced_ingress(
+                namespace=ns, label_selector=label_selector
+            )
+            for ing in ings.items:
+                print(f"Deleting Ingress {ing.metadata.name}")
+                self.networking_api.delete_namespaced_ingress(
+                    name=ing.metadata.name, namespace=ns
+                )
+        except ApiException as e:
+            _check_delete_exception(e)
+
+        # ConfigMaps
+        try:
+            cms = self.core_api.list_namespaced_config_map(
+                namespace=ns, label_selector=label_selector
+            )
+            for cm in cms.items:
+                print(f"Deleting ConfigMap {cm.metadata.name}")
+                self.core_api.delete_namespaced_config_map(
+                    name=cm.metadata.name, namespace=ns
+                )
+        except ApiException as e:
+            _check_delete_exception(e)
+
+        # PVCs (only if --delete-volumes)
+        if delete_volumes:
+            try:
+                pvcs = self.core_api.list_namespaced_persistent_volume_claim(
+                    namespace=ns, label_selector=label_selector
+                )
+                for pvc in pvcs.items:
+                    print(f"Deleting PVC {pvc.metadata.name}")
+                    self.core_api.delete_namespaced_persistent_volume_claim(
+                        name=pvc.metadata.name, namespace=ns
+                    )
+            except ApiException as e:
+                _check_delete_exception(e)
+
     def _create_volume_data(self):
         # Create the host-path-mounted PVs for this deployment
         pvs = self.cluster_info.get_pvs()
@@ -473,11 +560,13 @@ class K8sDeployer(Deployer):
         self.skip_cluster_management = skip_cluster_management
         self.connect_api()
 
+        app_label = f"app={self.cluster_info.app_name}"
+
         # PersistentVolumes are cluster-scoped (not namespaced), so delete by label
         if volumes:
             try:
                 pvs = self.core_api.list_persistent_volume(
-                    label_selector=f"app={self.cluster_info.app_name}"
+                    label_selector=app_label
                 )
                 for pv in pvs.items:
                     if opts.o.debug:
@@ -490,9 +579,14 @@ class K8sDeployer(Deployer):
                 if opts.o.debug:
                     print(f"Error listing PVs: {e}")
 
-        # Delete the deployment namespace - this cascades to all namespaced resources
-        # (PVCs, ConfigMaps, Deployments, Services, Ingresses, etc.)
-        self._delete_namespace()
+        # When namespace is explicitly set in the spec, it may be shared with
+        # other stacks — delete only this stack's resources by label.
+        # Otherwise the namespace is owned by this deployment, delete it entirely.
+        shared_namespace = self.deployment_context.spec.get_namespace() is not None
+        if shared_namespace:
+            self._delete_resources_by_label(app_label, volumes)
+        else:
+            self._delete_namespace()
 
         if self.is_kind() and not self.skip_cluster_management:
             # Destroy the kind cluster

stack_orchestrator/deploy/spec.py

@@ -149,9 +149,48 @@ class Spec:
             self.obj.get(constants.resources_key, {}).get(constants.volumes_key, {})
         )
 
+    def get_volume_resources_for(self, volume_name: str) -> typing.Optional[Resources]:
+        """Look up per-volume resource overrides from spec.yml.
+
+        Supports two formats under resources.volumes:
+
+        Global (original):
+            resources:
+              volumes:
+                reservations:
+                  storage: 5Gi
+
+        Per-volume (new):
+            resources:
+              volumes:
+                my-volume:
+                  reservations:
+                    storage: 10Gi
+
+        Returns the per-volume Resources if found, otherwise None.
+        The caller should fall back to get_volume_resources() then the default.
+        """
+        vol_section = (
+            self.obj.get(constants.resources_key, {}).get(constants.volumes_key, {})
+        )
+        if volume_name not in vol_section:
+            return None
+        entry = vol_section[volume_name]
+        if isinstance(entry, dict) and (
+            "reservations" in entry or "limits" in entry
+        ):
+            return Resources(entry)
+        return None
+
     def get_http_proxy(self):
         return self.obj.get(constants.network_key, {}).get(constants.http_proxy_key, [])
 
+    def get_namespace(self):
+        return self.obj.get("namespace")
+
+    def get_kind_cluster_name(self):
+        return self.obj.get("kind-cluster-name")
+
     def get_annotations(self):
         return self.obj.get(constants.annotations_key, {})