Speeding Up Infrastructure Provisioning with CloudForms Jason Dillaman Principal Architect, Red Hat Nick Lane Consultant, Red Hat
Agenda Where do clouds come from? Cloud Compute Capacity Scale-Out DevOps Provisioning Demo
Where do clouds come from?
Lessons from the Wild Derived from real-world CloudForms deployments Driven by efforts to make I.T. more responsive to business needs Data center compute capacity scale-out DevOps provisioning Laying the groundwork for efficient Cloud capabilities It s a journey
The Illusion of Infinite Capacity Insight into current capacity and utilization is required Converged infrastructure provides the foundation for automation Cisco/NetApp FlexPod one such example Not cost effective to support all possible infrastructure combinations Quickly (re-)allocate servers based on where they are needed Quickly (re-)allocate VMs based on where capacity is available
DevOps Collaboration between software developers and I.T. I.T. needs to expose the tools necessary for developer selfservice Goal is to reduce turnaround from weeks or days to minutes Eliminate the touch points where possible I.T. still needs to be able to enforce policy Right-sizing recommendations to avoid waste
CloudForms to the Rescue Red Hat CloudForms provides the glue to tie infrastructure components together Repeatable patterns to help modernize and standardize operations Automate all the Things * * within reason
Red Hat Cloud Portfolio
Red Hat CloudForms
Complete Cloud Service Lifecycle Automated Provisioning Simple/Multi-Tier, Full Stack Self-Service, Service Catalog Automated Provisioning Delegated Operations Power Operations, Console Reconfiguration Intelligent Optimization CPU, Memory & Storage Demand-Driven Scaling Horizontal & Vertical Start/Stop or Provision/Destroy Scheduled Retirement Fully Automated Multi-Phase Scheduled Retirement Demand- Driven Scaling Delegated Operations Intelligent Optimization
Cloud Compute Capacity Scale-out
Requirements Rapidly provision racks of Cisco UCS blades for infrastructure modernization Integrate with existing kickstart provisioning infrastructure Integrate with existing infrastructure for IP and DNS management Minimize the amount of pre-configuration required
Hypervisor Provisioning Focus is on RHEV hypervisor scale-out Cisco UCS, Infoblox DDI, and Red Hat Satellite Pattern can be applied to OpenStack Compute scale-out as well Scale-up and down with ease Admin initiated via the CloudForms service catalog Automatically initiated based upon utilization alert event Automation uses infrastructure tags and other heuristics to provision
Workflow Initiation Policy Enforcement Requests Service Catalog RBAC Policy CLOUDFORMS Intelligent Provisioning Role-Based Access Controls Approval Workflow
Example Service Catalog Item
Hypervisor State Machine
Infoblox DDI DNS and IP Address Management IP subnets are tagged with extensible attributes Location (IAD, RDU), Environment (DEV, TEST, QA, PRD), Service (Infra, Tier 1-3), Network location (Internal vs DMZ) REST API or Perl library to locate available IP address, available DNS name, and allocate a host entry Repeat if IP address / DNS name collision
Infoblox Workflow Details
Cisco UCS Manager Management of all hardware components in the Cisco UCS Servers are associated to Service Profiles Software definition of the server and is LAN and SAN connectivity Utilize Service Profile Templates to construct Service Profiles Avoids hard-coding configuration within CloudForms Associated with server pool to automatically allocate physical server ondemand REST API calls to create new Service Profile from Service Profile Template, control power state, and query status
Cisco UCS Manager Workflow Details
Red Hat Satellite Responsible for provisioning and configuration management CloudForms can perform basic provisioning Re-use existing infrastructure Kickstart profile for RHEV-H or RHEL+KVM Activation keys to subscribe to correct channels XMLRPC calls to create new PXE provisioning request and query status
Red Hat Satellite Workflow Details
Red Hat Enterprise Virtualization Newly provisioned hypervisors are connected to an existing RHEV-M cluster CloudForms tags drive automation's selection of RHEV-M and cluster Cisco UCS networks are mapped to RHEV-M logical networks by VLAN or name REST calls to create new host, configure host networking and power management, query status, and activate
Red Hat Enterprise Virtualization Workflow Create Host Get Install Status Deactivate CLOUDFORMS Update Power / Nets Activate RHEV-M
DevOps Provisioning
Requirements Need to transition existing applications to cloudy -like model Not feasible to leave existing software investment behind Automate the placement and installation/configuration of multi-tier applications Integrate with existing infrastructure for IP and DNS management Minimize the amount of user-input required
DevOps IaaS+ Provisioning Bridge between IaaS and PaaS Multi-tier applications are tied together using CloudForms and Puppet Scale-up and down with ease End-user initiated via the CloudForms service catalog Automatically initiated based upon utilization alert event Automation uses infrastructure tags and other heuristics to provision
Workflow Initiation Policy Enforcement Requests Service Catalog RBAC Policy CLOUDFORMS Intelligent Provisioning Role-Based Access Controls Approval Workflow
Example Service Catalog Item
Application State Machine
Red Hat Satellite & Infoblox Similar integration into Red Hat Satellite for VM provisioning and Infoblox for IP address and DNS management
Puppet Configuration Management Stand-alone Puppet Master or Satellite 6 (Foreman) CloudForms assigns new VMs to host group representing platform Override parameters as required based on service dialog inputs Discovery of peer VMs via Puppet manifest or injected by CloudForms REST API for Satellite 6; Git or SSH+CLI for Puppet Enterprise
Puppet Workflow Details Configure Host Puppet Settings CLOUDFORMS Retrieve Configuration Satellite 6 / Foreman / Puppet Application Stack
Demo
Summary
In Conclusion Real-world success with CloudForms IaaS+ Embrace automation where feasible Avoid the repetitive tasks, eliminate touch points, expose self-service Goal is to design-in scalability to meet future needs Compute capacity Application horizontal scaling Welcome to the cloud model
Questions?