Abstract: Eighteen months ago my company was asked to serve as the primary change agent during the transformation of a large, civilian agency towards agile/CD/DevOps practices. During the sales cycle our early impressions of the the client were fantastic: they had experience with agile delivery; a CIO with a history of IT innovation; leadership that already worked well with each other and a large, highly-visible NextGen effort making everyone eager to embrace any practices that could make them more efficient. What could go wrong? We learned that even under the most favorable circumstances, enterprise-wide change is not easy. It takes significant commitment from leadership and a willingness to continually revisit the people, processes and tools involved with delivering software.
This talk provides brief background on the state of the organization when we first engaged, and highlights the major successes (and growing pains!) we encountered while building an agile/CD platform, on-boarding over 100 projects and moving the agency up the DevOps maturity curve:
The Backdrop…
• Pressure to Deliver Suite of NextGen Applications
• Highly Visible Legacy Blow-Ups (CM, Deployment)
• Infrastructure & Networking issues
• Various methodologies/technologies
• No central source for CM, Development, Test or Deployment of Systems
• No shared priorities=organizational conflict
Building the Platform…
• Introducing Open-Source & Retiring Expensive Licensed Products
• Integrating Agency Tools (Selenium, LoadRunner, WebInspect…)
• Networking Considerations (DNS Names, Publicly Accessible)
• Standardizing Environments
• Moving towards the Cloud
Creating the “Platform Playbook”…
• Platform Onboarding & Project Maturation Process
• Inter-Organization Touch-points
• Piloting Migrations to the Platform
• Refactoring Process at we Learn
Shoring Up Agile Practices…
• Agile Assessments/Metrics
• Tailored Best Practices, Standards
• Training
• Codified Enforcement of Standards
• Automated Builds
Aligning the Organization…
• Ensuring Shared Priorities
o Quarterly CIO Summit (Roadmap)
o Monthly Stakeholder (Objectives)
o Weekly Planning (Stories)
• Communication (newsletter, website, brown-bags, Kanban wall)
Victims of our own Success…
• One-Size-Fits-All Pilot no longer works at more teams onboard
o Expanded Training for users at lower levels of process expertise
o Increased support for diverse technologies via plug-ins and customizations
o Large Diversity in Technologies & Target Environments
o Revised Bundling Technique
• Rising fixed costs for Administration
o Moved to LDAP Groups to offload user access
o Created Administrative Reports
o Automated Platform Deployments
• Difficulty Policing Platform Customers
o Implemented Templates (Jenkins)
o Automated Compliance & Administrative Reports
• Heavy Load at Peak Periods
o Jenkins Slaves
o Labeling
Finally, automating Deployments…
• Release bundles
o Environment Agnostic
o Release Documentation
o Traceability
• Automated deployments
o Jenkins/Scripts
o Puppet
o Ansible
Knowledge Transfer and Signs of Growing DevOps
• Puppet & Ansible can live together
• Co-Location
• Shared Dashboards
Closing\Q&A
Learning Outcomes: - Attendees will leave with an understanding that:
- • An enterprise level shift towards DevOps is difficult even in the most favorable situation
- • People, Process and Tools are all important during transformation
- • Process and tools become more important in organizations (i.e., Government) where contracted staff is always in transition.
- • There will be growing pains as you move from pilot to enterprise-wide adoption. What works today will not work tomorrow. Plan for it.
- • A solid foundation in Agile best practices is critical to the success of a lasting CD/DevOps practice
Attachments:
