Struct ContainerManager 

pub struct ContainerManager { /* private fields */ }

Container lifecycle manager.

When the containers feature is enabled, this uses bollard for Docker API access. Without the feature, all operations return ContainerError::DockerNotAvailable.

Implementations

impl ContainerManager

pub fn new(config: ContainerConfig) -> Self

Create a new container manager.

Does NOT attempt to connect to Docker at construction time. Connection is deferred to the first operation that needs it.

pub fn simulated(config: ContainerConfig) -> Self

A manager whose start_container is a simulated transition (no engine runs anything). For tests of the registry / health wiring only; ContainerManager::new refuses to start.

pub fn is_simulated(&self) -> bool

Whether this manager fakes starts (ContainerManager::simulated).

pub fn set_chain_manager(&mut self, cm: Arc<ChainManager>)

Attach a chain manager for exochain event logging.

pub fn config(&self) -> &ContainerConfig

Get the container configuration.

pub fn configure( &self, spec: ManagedContainer, ) -> Result<String, ContainerError>

Configure and validate a container image, registering it for management.

Validates the specification (image name must be non-empty, port numbers must be valid, container name must be non-empty) and registers it in the Stopped state. Returns the container name as its identifier.

Errors

Returns ContainerError::InvalidConfig when validation fails.

pub fn register(&self, spec: ManagedContainer)

Register a container specification for management.

This does not start the container; it only registers it for tracking. Call start_container to actually start it.

pub fn start_container(&self, name: &str) -> Result<(), ContainerError>

Start a managed container by transitioning its state to Running.

Only a ContainerManager::simulated manager starts anything, and that is a simulated transition, kept so the integration between ContainerManager and the kernel ServiceRegistry / HealthSystem can be tested without a container runtime installed. Workloads (cogs) do not start here: they run through the real engine adapters in crate::workload_runtime::container (Apple container, Docker / OrbStack, Podman), governed and chained by WorkloadHost (mesh-placement-09).

Errors

Returns ContainerError::DockerNotAvailable on a manager built with ContainerManager::new (no engine is wired), ContainerError::ContainerNotFound if the name is not registered, or ContainerError::StartFailed if the container is in a state that cannot be started.

pub fn stop_container(&self, name: &str) -> Result<(), ContainerError>

Stop a managed container.

Transitions the container from any active state to Stopped.

pub fn container_state(&self, name: &str) -> Option<ContainerState>

Get the state of a managed container.

pub fn list_containers(&self) -> Vec<(String, ContainerState)>

List all managed containers with their states.

pub fn health_check(&self, name: &str) -> Result<HealthStatus, ContainerError>

Health check for a specific container, returning a HealthStatus.

pub fn container_health( &self, name: &str, ) -> Result<ContainerHealth, ContainerError>

Detailed health report for a specific container.

pub fn stop_all(&self)

Stop all managed containers.

pub fn len(&self) -> usize

Get the number of managed containers.

pub fn is_empty(&self) -> bool

Check whether any containers are managed.