Write a Runtime Plugin

Runtime Plugin

The Runtime plugin implements the life-cycle management for FDU, is has to strictly follow the state machine defined for the FDU and provide handles for the state management and monitoring for the running FDUs.

This picture represent the state machine that an FDU has to follow.

States

Each FDU can be in one of those nine states:

• UNDEFINED
• DEFINED
• CONFIGURED
• RUNNING
• PAUSED
• SCALING
• MIGRATING
• TAKE_OFF
• LANDING

UNDEFINED

In this state the FDU is registered in the catalog but there are no instances in the system

DEFINED

In this state a FDU instance was created in a node, but it is only created, meaning that an instance UUID was assigned to it.

CONFIGURE

The FDU instance was configured in a node, the image was fetched, resources allocated, and the instance was defined in the underlying hypervisor/container engine. Or in case of native application the namespace was created.

RUNNING

The FDU instance is started, the instance is started in the underlying hypervisor/container engine or the native application process is started. Monitoring of the instance begin at this stage.

MIGRATING

The FDU instance is instructed for migration, this leads to two different states on the source and destination node. Source node instance transits to TAKE_OFF, when it is ready for migration, while destination instance transits to LANDING when ready for migration. After migration the instance in the source node is destroyed, while in destination node the instance is in RUNNING state. Migration is supported only for collaborative FDUs.

PAUSE

The execution of the instance is freezed. This state is supported only for certain FDUs.

SCALING

The instance is scaled up or down. This state is supported only for certain FDUs.

State transitions

The Runtime plugin has to implement state transition for the FDU is able to manage. The Runtime plugin can access to functionalities of the OS and Network Manager plugin using the os and nm objects.

State transitions are defined by the RuntimePluginFDU interface and by the implementation of the following functions.

define_fdu

It takes as parameter the FDU instance record and create the instance in the node.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

undefine_fdu

Takes the instance UUID as parameter and destroy the instance from the node.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

configure_fdu

Takes the instance UUID as parameter and create all the resources needed for the execution of the FDU, creates the instance in the hypervisor/container engine if needed, creates the needed networking components by calling the Network Manager plugin.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

clean_fdu

Takes the instance UUID as parameter and cleans the configured FDU, removes images, instance from hypervisor/container engine if needed, networking components, releases the resources.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

run_fdu

Take the instance UUID and starts the instance UUID, calls APIs from hypervisors/container engine in needed or start the native application.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

stop_fdu

Takes the instance UUID and stops the instance running, calls APIs from hypervisors/container engine in needed or send a signal the native application.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

pause_fdu

Takes the instance UUID and pauses the execution of the running FDU if it is possible.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

resume_fdu

Takes the instance UUID and resumes the execution of the paused FDU.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

scale_fdu

Takes the instance UUID and scaling parameters and scales the running FDU if it is possible.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

migrate_fdu

Takes the instance UUID and migrates the FDU from a source node to a destination node if it is possible.

In case of success updates the status using update_fdu_status

In case of error returns an error using write_fdu_error

The interface for a Runtime plugin is available in Go and Python3 more implementation are coming soon.

See example implementations from the LXD Plugin