Why continuous delivery needs
devops, and why devops needs
infrastructure-as-code
Sriram Narayan @sriramnarayan
about me
• Part of ThoughtWorks Studios Go team
• Have consulted as Tech Principal, Agile Coach, organization
development coach
• Been a manager, leadership development coach and
innovation facilitator
• Written production code in Java, C#, Ruby and Python
2010
About ThoughtWorks Studios
Training on agile foundations, project management, testing, continuous delivery www.thoughtworks-studios.com/services/agile-workshops
Executive summary
• Why does CD need devops?
– Because you can’t have CD with a siloed org
– devops helps blur the strict boundary between dev and ops.
• Why does devops need IaC?
– IaC helps create a common currency between dev and ops
Continuous delivery is
an approach to delivering software that reduces the cost, time, and risk
of delivering incremental changes to users.
ThoughtWorks Anthology, March 2008
I think of it as agile software delivery/release instead of agile software development
1. Your software is releasable throughout its lifecycle
2. Your team prioritizes keeping the software releasable over working on new features
3. Anybody can get fast, automated feedback on the production readiness of your systems any
4. You can perform push-button releases of any version of the software on demand.
It is a high bar
devops
theagileadmin.com/what-is-devops/
system administrators participating in an agile development process alongside developers and
using a many of the same agile techniques for their systems work
wikipedia
a response to the interdependence of software development and IT operations. It aims to help an organization rapidly produce software products and services.
Q. Why does CD need devops?
A. Because you can’t have CD with a siloed org and devops helps blur the strict boundary between dev and ops.
A typical setup (pre-devops)
A typical anti-pattern (post-devops)
Silo trouble
• What is a silo?
• Why are silos bad for CD? • What causes silos?
– Understand the causes so that we can take preventive measures
What is a silo?
http://wiki.answers.com/Q/What_are_organizational_silos
Organizational silos…are (often) dysfunctional units or departments within an enterprise, characterized by their tendency to protect themselves...they are sometimes the result of unchecked "empire
building" by middle management.
A less harsh attempt…
Anything that tends to localize a (technical)
competency and create handoffs in a delivery value stream.
Why are silos bad for CD?
• Silos mean handoffs • Handoffs
– Cannot be continuous or real-time
– Communication protocol (e.g. SPOC, minutes) – Work queues, turn around times, follow ups,
priority escalations
• Can only work with large batch size
We all want shorter cycle times
• This requires small batch size
• Handoffs double when batch size halves
– Too much communication overhead
– Just not responsive enough • Ergo, we reduce hand-offs
Things that promote a silo
• Org structure, reporting hierarchy • Geographic separation
• Speciality tools
– Limited licenses to go around
– e.g. salesforce accounts, expensive deployment tools
1. Extreme case: dev team has no people from other functions (7 silos)
2. Many orgs now merge development, build and testing into one team (5 silos)
3. devops tries to merge operations into what then becomes
the IT team
Towards autonomous cross-functional teams
7 silos 5 silos no silos
ARCHITECTURE ARCHITECTURE DEVELOPMENT BUILD DATABASE TESTING DATABASE OPERATIONS OPERATIONS
Silo mitigation – team and org design
• Organize along business outcomes
• Create semi-autonomous teams that own
outcomes and minimize geographic separation • Distribute specialists among the teams rather
than create a speciality team
– Responsiveness over Efficiency • It’s ok if specialist utilization falls
Silo mitigation – choice of tools
• Use tools & techniques that blur boundaries between specialists rather than define them
– Same source code repo for app, build and deploy – Same wiki tool, work management tool
• One over-arching tool for the entire continuous delivery value stream
http://www.thoughtworks.com/radar
Q. Why does devops need
infrastructure-as-code?
A. To create a common currency between dev and ops, which in turn helps mutual
infrastructure-as-code
• Infrastructure described via code
– Server and network config, base builds, packages, environments, services and config files
• A domain model for infrastructure
• First class infrastructure primitives that allow for a declarative description of desired state • The phoenix test
The radar on infrastructure-as-code
July 2011
March 2012
http://wiki.opscode.com/display/chef/Chef+Concepts+as+UML
From Bryan Beery http://faodata.blogspot.in/2012/04/chef-testimonial-data.html Chef: resources in a recipe
Chef: data bag
Chef: scripting a master-slave config
Common currency
Application code
The power of text
• Text lends itself better to collaboration and automation
– Non-proprietary version control, diff and merge – So much tooling is available
• Code and config in the form of text descriptors makes transactions between specialists more fluid.
IaC facilitated crossover
What devs learn from ops • Resilience
• Cloud, Virtualization • Data centres
• Inventory
• Backups & Audits • Monitoring
What ops learn from devs • Domain modelling
• Common vocabulary • Continuous integration
• Everything in version control • DRY
Putting it all together
build and unit test acceptance
deploy-Performance deploy-QA
deploy-production
App Source
pkg repo Infra Code
source, tests binaries recipes etc target nodes pipeline trigger publish artifacts
devops
continuous delivery
Infrastructure-as-code
App dev
one cross-functional team of generalizing specialists
text descriptors in version control
One tool for traceability, orchestration, visualization
iterative, not a linear flow from left to right
discover analyze validate architect design develop build test acceptance performance exploratory provision inventory audit deploy upgrade rollback monitor backup