7. Operations and Life Cycle Management

To create an Infrastructure as a Service (IaaS) cloud requires the provisioning and deployment of the underlying infrastructure (compute, networking and storage) and deployment, configuration and management of the necessary software on the infrastructure; in the process of deploying the software, configuration of the infrastructure may also need to be performed.

Instead of deploying the infrastructure components and services manually, the current best practice is to write code (Infrastructure as Code, IaC) to define, provision, deploy, configure and manage the IaaS cloud infrastructure and services. IaC tools allow the entire provisioning, configuration and management processes to be automated. The desired state of the infrastructure and services is represented in a set of human readable, machine executable, and version-controlled files. With version control, it is easy to roll back to an older version and have access to the history of all committed changes.

The provisioning of the infrastructure is typically performed by provisioning tools while the deployment of the software and the configuration of the software, and where needed the infrastructure, falls in the domain of configuration management tools. A single tool may support both provisioning and configuration management.

Operators may choose certain paradigms with respect to how they provision and configure their IaaS cloud. These paradigms will drive the selection of the provisioning and configuration tools. In this chapter we will discuss the capabilities of provisioning and configuration management systems; some open-source tools may be mentioned but their capabilities are beyond the scope of this chapter.

7.1. Procedural versus Declarative code

The procedural style IaC tools require code that specifies how to achieve the desired state. Whilst the declarative style IaC tools require code that specifies the desired state (what not how). The major difference between the two styles emerges when changes to the desired state are required. In the procedural style, the change is coded in terms of the difference between the desired and current states while in the declarative style the new desired state is specified. In the procedural style since the state difference has to be coded, a new code file has to be created for each change; in the declarative style the existing code file is updated with the new state information. In the declarative style knowledge of the current state is not required. In the procedural style, knowledge of the current state has to be manually figured by tracing the created code files and the order in which they were applied.

7.2. Mutable versus Immutable infrastructure

In the mutable infrastructure paradigm, software updates are made in place. Over time this can lead to configuration drift where each server becomes slightly different from all other servers. In the immutable infrastructure paradigm, new servers are deployed with the new software version and then the old servers are undeployed.

7.3. Cloud Infrastructure provisioning and configuration management

In the Reference Model [1], the “Configuration and Lifecycle Management” chapter defines the functions of Configuration and Life Cycle Management (LCM). To operate and manage a scalable cloud, that minimises operational costs, requires tools that incorporates systems for automated provisioning and deployment, and managing configurations that ensures the correctness and integrity of the deployed and configured systems.

7.3.1. Underlying resources provisioning

This section deals with automated provisioning of the Cloud Infrastructure; for example, provisioning the servers, switches, routers, networking (e.g., subnets, routing tables, load balancers, etc.), databases and all required operating systems (Servers, switches, etc.).

The following are the minimum tasks that need to be performed by automation:

  • Pre-boot configuration such as BIOS/RAID/IPMI settings: Hardware manufacturers typically have their proprietary interface for these tasks but standards such as Redfish are being increasingly utilised. Consider using tooling to ensure consistency across all infrastructure components.

  • Bootloader installation of base Network Operating System (NOS) on networking equipment or the Operating System (OS) should be performed using PXE; again consider tooling to ensure consistency across all infrastructure components.

To ensure operational efficiency and save cost and time, the lifecycle management for physical and virtual servers must be automated using tools which will handle the repetitive tasks like provisioning, configuration, and monitoring. Foreman [127] is commonly used to automate the provisioning and management of bare metal infrastructure. Foreman is an open-source project, base of several commercial products. Foreman provides the full management of PXE configuration and the installation for many Operating Systems (CentOS, Fedora, Ubuntu, Debian, Red Hat Enterprise Linux, OpenSUSE, etc.). Foreman service can be installed by Ansible playbooks [128]. Ansible playbooks are basic tools for the automation of the infrastructure virtualisation layer deployments.

7.3.2. VIM deployment

When the underlying resources are installed and configured, the VIM software is deployed. Automated deployment is highly recommended for the same reasons of efficiency. Open-source installers are available to perform the deployments of the OpenStack services. A subset of these tools is described below.

  • OpenStack TripleO [84], “OpenStack on OpenStack”

    TripleO is an official OpenStack project which allows to deploy and manage a production cloud onto bare metal hardware using a subset of existing OpenStack components. The first step of deployment is the creation of an “undercloud” or deployment cloud. The undercloud contains the necessary OpenStack components to deploy and manage an “overcloud”, representing the deployed cloud. The architecture document [129] describes the solution. Nova and Ironic are used in the undercloud to manage the servers in bare metal environment. TripleO leverages on Heat tempates.

  • Airship v2 [82]

    Airship is supported by the OpenStack Foundation. It is a collection of interopable open-source components for declarative automation of cloud provisioning. The configurations are defined by YAML documents. All services run on containers. Airship v2 is aligned with maturing CNCF projects such as Kubernetes, Kubectl, Kubeadmin, Argo, Cluster API, Kustomize, and Metal3. Airship v2.1, released in November 2021, leverages on Kubernetes 1.21. It includes cloud provisioning at edge and for 3rd party cloud. The use of the OpenStack-Helm project allows the deployment of OpenStack on top of Kubernetes. Airship is not only a provisioning tool, but also a also a configuration management system.

  • StarlingX [83]

    StarlingX is dedicated to cloud infrastructure deployment at the edge, taking into account the specific edge use cases requirements for low latency and precision clock synchronisation. It aims to install a containerised version of OpenStack services, leveraging on Kubernetes, Docker registry, Airship Armada, and Helm.

    OpenStack-Helm is used as a starting point. OpenStack is installed and managed as an Armada application. Armada Applications are a set of one or more interdependent Application Helm charts. In the case of StarlingX, there is generally a Helm chart for every OpenStack service.

7.3.3. Configuration Management

The configuration management system ensures the correctness and integrity of the deployed and configured systems. The tools provide the assurance that the expected software is running with the expected configurations on correctly configured nodes that continue to be configured correctly.

Configuration Management is composed of the following activities:

  • Desired (Target) State: a version of the software and hardware and their configurations. Depending upon the configuration management system these configurations are specified in cookbooks, playbooks, manifests, etc. The configuration specifications in these artefacts are used to configure the different types of nodes, BIOS, operating systems, hypervisor and OpenStack services (through settings within their config files such as nova.conf, etc.).

  • Current State: the current configuration of software and hardware as provided by monitoring systems

  • State variance mitigation: The CM system, on discovering a variance between the desired and current states, acts to drive the state to the desired state. Each CM system accomplishes the task in different ways.

7.4. Cloud Infrastructure and VIM Maintenance

Cloud Infrastructure and VIM Maintenance activities can be classified as

  1. Deployment of additional infrastructure components (or removal of infrastructure components)

  2. Cloud Infrastructure Configuration changes

  3. VIM Configuration changes

  4. Version changes (upgrade) of Cloud Infrastructure software (for example, Host Operating System, Hypervisor, etc.)

  5. Version changes of VIM Software (or component services)

Deployment (or removal) of infrastructure components

In declarative tools, the code with the specified desired state (for example, number of compute servers) is modified to the new desired state. The IaC tool then ensures that the desired state is achieved. In procedural tools, the step-by-step code to deploy (remove) infrastructure components needs to be specified. Existing code can be cloned, and appropriate changes made to get to the desired state.

Configuration and Version Changes

Configuration and Version Changes are made in a similar fashion to the “Deployment of infrastructure components” except that the IaC tools used may be different.

7.5. Logging, Monitoring and Analytics

  • Logging

  • Monitoring

  • Alerting

  • Logging, Monitoring, and Analytics (LMA) Framework

7.5.1. Logging

A log, in the context of computing, is the automatically produced and time-stamped documentation of events relevant to a particular system. All software, including operating systems, middleware and applications produce log files. Enterprises and vendors may have custom monitoring and logging solutions. The intent of logging and monitoring is to capture events and data of interest to the Cloud Infrastructure and workloads so that appropriate actions can be taken. For example,

  • Operating systems and web servers maintain an access log of all access requests, session details and file access.

  • Databases maintain a transaction log of all transaction executed including added, changed and deleted data.

  • Audit logs record chronological documentation of any activities that could have affected a particular operation or event. Data typically includes resources accessed, destination and source addresses, and a timestamp and login information for the person who accessed the resources.

Some of the data is to support the metrics collection specified in the Reference Model [1].

Logs have multiple operational uses including for:

  1. Regulatory Compliance and Security Information and Event Management (SIEM) featuring the automated gathering, analysis and correlation of log data across all systems and devices across an operator to provide real-time analysis, event prioritisation, reporting, notification and alerting.

  2. Monitoring across systems in real-time to detect particular log events, patterns, anomalies or inactivity to gauge system and application health

  3. Identify system and application performance and configuration issues

  4. Root cause analysis for system and application failures and errors

  5. Ensuring that operational objectives and SLAs are met

7.5.2. Monitoring

Monitoring is the process of collecting, aggregating, and analysing values that improve awareness of the components’ characteristics and behavior. The data from various parts of the environment are collected into a monitoring system that is responsible for storage, aggregation, visualisation, and initiating automated responses when the values meet specific threshold.

Monitoring systems fulfil many related functions. Their first responsibility is to accept and store incoming and historical data. While values representing the current point in time are useful, it is almost always more helpful to view those numbers in relation to past values to provide context around changes and trends.

7.5.3. Alerting

Alerting is the responsive component of a monitoring system that performs actions based on changes in metric values. Alert definitions are composed of two components: a metrics-based condition or threshold, and an action to perform when the values fall outside of the acceptable conditions.

While monitoring systems are incredibly useful for active interpretation and investigation, one of the primary benefits of a complete monitoring system is letting administrators disengage from the system. Alerts allow the specification of situations that make sense to actively manage, while relying on the passive monitoring of the software to watch for changing conditions.

7.5.4. Logging, Monitoring, and Analytics (LMA) Framework

In this section, a possible framework utilising Prometheus, Fluentd, Elasticsearch and Kibana is given as an example only.

Monitoring and Logging Framework

Figure 7.1 Monitoring and Logging Framework

The monitoring and logging framework (figure above) leverages Prometheus as the monitoring engine and Fluentd for logging. In addition, the framework uses Elasticsearch to store and organise logs for easy access. Prometheus agents pull information from individual components on every host. Fluentd, an Open Source data collector, unifies data collection and consumption for better use and understanding of data. Fluentd captures the access, application and system logs.