contiv-experimental/cluster: cluster management infrastructure using serf, collins and ansible! · GitHub

##Overview Contiv Cluster is an integrated collection of software that simplifies deployment, management, and maintenance of clustering systems. At the heart of Contiv Cluster is Cluster Manager, a RESTful API service providing programmatic access to the Contiv Cluster.

Contiv Cluster intends on supporting a range of clustering software such as Docker Swarm, Kubernetes, Apache Mesos, and others. Currently, Contiv Cluster supports the following cluster formation:

• Cluster Type: Docker [Unified Control Plane] (https://www.docker.com/products/docker-universal-control-plane) | Docker Swarm | Kubernetes
• Cluster OS: CentOS 7.2
• Cluster Target: Bare Metal | Vagrant

##Definitions To better understand Contiv Cluster, you should know a few system definitions.

###Node A physical or virtual server with predefined/discovered compute, memory, storage and networking capabilities.

###Cluster A collection of one or more nodes running clustering software such as Docker Swarm. Nodes may or may not be homogeneous in their capabilities.

###Image An operating system with a minimal set of pre-installed packages used by a node. For instance, node automated discovery requires a cluster membership service like Serf to be pre-installed and started on the node's operating system during bootstrapping.

###Bootstrap

Node Bootstrap: The process of installing the node image and booting the node for the first time.

Cluster Bootstrap: The process of adding one or more bootstrapped nodes to the cluster for the first time.

###Lifecycle A collection of well-defined states that a node or cluster transitions through based on system events.

###Infra Service A userspace application or a kernel module that runs on a node and provides a support service to the clustering system software. Examples include plugins, drivers, or components for networking and storage, key/value stores, etc..

##Features

Contiv Cluster provides the following features. Not all features are implemented, or will be implemented in the initial release. Please review the Roadmap or continue reading to understand individual feature status.

• Node discovery to simplify cluster bootstrapping and expansion
• Cluster management
  • Inventory
    • Discovery
    • Node state database
  • Node image management
    • Install
    • Upgrade
  • Node configuration management
    • Clustering software components (swarm-agent, etc)
    • Infrastructure software components (ovs, serf, consul/etcd, etc)
    • Contiv infrastructure services (netmaster/netplugin and volmaster/volplugin)
  • Bootstrap
    • Node
    • Cluster
  • Lifecycle
    • Node
    • Cluster

The rest of the the document details the cluster management features that were briefly listed above. Use the Design Guide to better understand the technical details of Contiv Cluster or use the Installation and Operations Guide to start using the system.

##Inventory Inventory management provides:

• Cluster Membership Management: Automatically discovers nodes and tracks discovery status.

• Node State Database: Tracks the current state of the node within the cluster. Node states consist of the following:

  • New: A node that has not completed the bootstrap and discovery process.
  • Discovered: A node has completed bootstrapping and notifies the Cluster Manager. The node waits to be Commissioned.
  • Commissioned: A node has been configured to participate in the cluster.
  • Upgraded: One or more node components have been successfully upgraded and the node is ready to participate in the cluster.
  • Decommissioned: All configuration and software components have been removed and the node no longer participates in the cluster.

##Node Image Management Node Image Management provides:

• Image Repository: A central location reachable by nodes, where images are hosted. Images are delivered to nodes using mechanisms such as PXE.
• Image Installation: Automates the image installation of new nodes.
• Upgrades: Automates the process of upgrading the node image. Upgrades can be automatically triggered and can be performed cluster-wide or rolling.

Note: Node Image Management features are not included in the initial release of Contiv Cluster. Operators are responsible for node bootstrapping, including image provisioning.

##Node Configuration Management Node Configuration Management provides:

• Configuration Repository: A central location for hosting node configuration files. The configuration is used to automate the deployment of nodes.
• Configuration Push: The configuration of nodes is pushed from Cluster Manager to nodes.
• Configuration Cleanup: Configuration files, service, packages, etc. are removed from nodes.
• Configuration Verification: Checks to ensure that the node configuration is truly functional.
• Configuration Upgrade: Automates the process of upgrading node software components. Upgrades can be automatically triggered and can be performed cluster-wide or rolling.
• Role/Group-Based Configuration: Nodes can be assigned a group or role. Role/Group-Based configuration selectively manages services on nodes statically by the operator or dynamically based on service availability policy.

Note: Dynamic role assignment is not supported in initial release of Contiv Cluster.

##Bootstrap Bootstrap involves provisioning a node or cluster for the first time.

###Node Contiv Cluster provides the following node bootstrapping capabilities:

• Installing the base image
• Performing initial configurations such as disk partitioning
• Assigning an IP address
• Configuring user credentials and permissions to perform configuration management tasks
• Starting infrastructure services such as Serf

Note: To start simple this feature is not provided as part of initial release of Contiv Cluster.

Operators are responsible for node bootstrapping, including image provisioning.

###Cluster The cluster bootstrap installs and configures clustering software components such as swarm-master (for Docker Swarm) to the first bootstrapped node with parameters such as:

• Configuration management parameters such as user information, configuration repository
• Inventory management parameters such as database url

##Lifecycle Lifecycle management involves integrating multiple cluster and node management tasks at a central location to simplify monitoring and administration of the cluster.

###Node Node lifecycle management provides:

• Bootstrap: Remotely provision a node's image from a central location. As described above this feature is not provided in initial release of the Contiv Cluster.
• Cluster Membership: Automatically track cluster's membership of reachable and unreachable nodes.
• Commission: Remotely provision a node's clustering software, configuration, etc.. Optionally commission newly discovered nodes automatically.
• Upgrade: Remotely upgrade the image or a software component of a commissioned node. Optionally upgrade nodes when a configuration repository changes.
• Decommission: Automatically remove configuration, software components, etcd from a commissioned node.
• Batch operations: Commission, upgrade or decommission all or a subset number of nodes.
• Reloads: In event event of a node reload or unreachability event, verify the node configuration and automate corrective actions.

###Cluster The following Cluster management capabilities are provided by Contiv Cluster:

• Bootstrap: Remotely bootstrapping a cluster.
• High Availability: The cluster manager service is available as long as one node is running in the cluster.

• Installation and Operations Guide: Deploy Contiv Cluster to bare metal servers or a Vagrant development environment.
• Troubleshooting Guide: Troubleshoot Contiv Cluster workflow failures.
• Design Documentation: Learn more about the design details of Contiv Cluster.

Use the following resources to get support or create a GitHub Issue:

• Website: http://contiv.github.io/
• Slack: contiv.slack.com