Difference between revisions of "Senlin"

Latest revision as of 05:52, 6 May 2019

Overview

Senlin is a clustering service for OpenStack clouds. It creates and operates clusters of homogeneous objects exposed by other OpenStack services. The goal is to make orchestration of collections of similar objects easier.

Senlin interacts with other OpenStack services so that clusters of resources exposed by those services can be created and operated. These interactions are mostly done through the via profile plugins. Each profile type implementation enable Senlin to create, update, delete a specific type of resources.

A Cluster can be associated with different Policy objects that can be checked/enforced at varying enforcement levels. Through service APIs, a user can dynamically add Node to and remove node from a cluster, attach and detach policies, such as creation policy, deletion policy, load-balancing policy, scaling policy, health policy etc. Through integration with other OpenStack projects, users will be enabled to manage deployments and orchestrations large-scale resource pools much easier.

Senlin is designed to be capable of managing different types of objects. An object’s lifecycle is managed using Profile Type implementations, which are plugins that can be dynamically loaded by the service engine.

Features Highlight

The primary features of the Senlin service are listed below:

• A generic clustering/collection service for managing groups of homogeneous cloud objects on OpenStack.
• A set of APIs for managing cluster membership, e.g. add/remove nodes.
• A plugin-based object profile management enabling the creation and management of any object pools.
• A plugin-based policy enforcement framework featuring flexible policy customization for cluster management.
• A plugin-based event notification that allows for dumping or pumping cluster actions to storage or downstream software.
• A asynchronous execution engine for ensuring the state consistency of clusters and nodes.

Architecture

The following picture depicts the high-level system architecture of the Senlin service:

The main components in the picture are:

• The senlinclient component provides the command line interface (CLI) for users to interact with the Senlin service;
• The senlin-api service provides the RESTful APIs for the senlinclient component or other services;
• The senlin-engine service sits behind the senlin-api service and operates on the clusters/nodes, using profiles and policies that are loaded as plugins.

Usage Scenarios

Below are some typical usage scenarios of the Senlin service. More scenarios will be added later.

• Managing Node Affinity
• AutoScaling using Ceilometer/Aodh
• AutoScaling using Heat Resource Types

Documentation

• Installation Guide
• Tutorial
• User References
• Developer Documents
• API Version 1.0 Reference

Related Presentations

• Enhancing high availability in the context of OpenStack, Atlanta Summit, 2014

 - A taxonomy of architecting high-availability into an OpenStack cloud.
 - The then current status of OpenStack support to HA.
 - A proposed service (HAaaS) which layed the foundation for Senlin's cluster design.

• Suning cloud and Heat, Paris Summit, 2014

 - Experience sharing on using OpenStack for an online retailer
 - Requirements on orchestrating enterprise applications: clustering, scaling, HA ...
 - Gaps identified in the then current status of OpenStack services

• VM HA and cross-region autoscaling, Paris Summit, 2014

 - How VM HA can be achieved by integrating OpenStack services.
 - How to scale a VM cluster across multiple regions.

• Senlin design discussion, Vancouver Design Summit, 2015

 - A discussion on Heat design summit about senlin service design
 - How the service is architected, the terminology behind it
 - How auto-scaling can be achieved in this design

• Integration Magnum and Senlin for container autoscaling, Tokyo Summit, 2015

 - Experience sharing on integrating Magnum and Senlin
 - Magnum provisions the container orchestration engines (COEs)
 - Senlin is responsible to scale the container clusters and the underlying VM clusters.

• Senlin deep dive for v1.0.0 release, Austin Summit, 2016

 - An overview of the Senlin service design as of the 1.0.0 release.
 - A summary of progress made during Mitaka cycle.
 - An overlook of features planned for Newton cycle.

• OpenStack native container clustering, Austin Summit, 2016

 - A technology preview of the container clustering support in Senlin
 - The discussion focuses on the on-going effort that builds and manages a docker cluster on OpenStack

• Elastic, resilient and load-balanced cluster in 5 mins, Austin Summit, 2016

 - Experiences sharing how elasticity, resiliency and load-balancing can be achieved using Senlin
 - The demo shows how the above requirements are met using a Heat template consisting of Senlin resource types.

• on-building-an-auto-healing-resource-cluster-using-senlin [1], Barcelona, 2016
• managing-clusters-of-thousands-of-vms-using-senlin [2], Barcelona, 2016
• integration-of-enterprise-monitoring-product-senlin-and-mistral-for-auto-healing [3], Boston, 2017

• high-availablity-and-scalability-management-of-vnf [4], Boston, 2017

• senlin-an-ideal-bridge-between-nfv-orchestrator-and-openstack[5], Boston, 2017

• standing-up-kubernetes-on-openstack-is-cool-then-what [6], Sydeny Summit, 2017

Get Involved