DevOps Explained

DEVOPS
EXPLAINED
Richard Seroter
@rseroter
#Agile2017

WASTED
TIME
WASTED
SPACE
WASTED
EFFORT
“Our greatest fear should not be of failure, but of succeeding at
something that doesn't really matter.”
- D.L. Moody

95%of any process is waste
@rseroter #Agile2017

seven types of waste:
1. Overproduction
2. Waiting
3. Unnecessary transport
4. Over-processing
5. Excess inventory
6. Unnecessary movement
7. Defects

Does this have anything
to do with delivering
technology solutions?
You betcha.

Three observations about state of
tech today:
1. Outages are expensive
2. Demand for faster delivery
3. Talent retention is difficult

We want … more resilient systems that easily
absorb change, from a customer-centric
organization that attracts talent.

We renamed our testing
team and called it DevOps.
Can we fire all our sysadmins and
make devs “full stack engineers”?
This is for startups,
not real businesses.
Our company does ITIL, so
no DevOps for us.
Nobody even knows
what DevOps is.
I bought six units of DevOps from
my vendor. Did I do it right?
People who talk most about
DevOps have never even done it.

What is DevOps? It’s about using
coordination and automation to deliver
higher quality software, faster.

High performers:
Deploy 46x more frequently
Offer 440x faster lead time for changes
Deliver 96x faster mean time to
recovery
Possess a 5x lower change failure rate
- 2017 State of DevOps Report

Core values
Patterns and practices
Toolchain

core values

Empowered individuals
Empowered within the system
Do what’s needed to maintain the
service
Engineers can “pull the cord”
Responsible for proposing
improvements

Accountable
Flip-side of being empowered
Everyone responsible for quality,
everywhere
Part of “you build it, you run it”
mentality
Accountable to teammates as much
(more so?) as management

Teamwork
No lone geniuses
Coordination across functions at *all*
times
Respect for distinct skill sets
Pays off during periods of crisis

Trust
Trust in each other, trust in
management
Requires everyone to be working off
the same values and objectives
Establish trust within and across teams
Ok to “trust but verify”

Transparency
Visibility into objectives, priorities
Clear knowledge of long-term plans
Shared access to metrics, tickets
source code

Continuous learning,
improvement
Requires introspection, identifying root
cause of problems
Sprint and incident retrospectives are
great sources
Learning should clearly make its way
back into the organization
Growing individual and team skills
requires management investment

Feedback loops
How are you learning about apps in
production?
Consider structured way to solicit
feedback from *all* stakeholders
Necessary to have fast-feedback
channels available

Data-driven decisions
Upfront data capture and analysis
capability drives good decisions
Not a replacement for thinking
Drives consensus, prioritization
Only the “right” metrics

Standardization
Care about quality
Prioritize repeatable actions through
automation
Scale through consistency
Consensus use of standards

Customer-centric
Empathy for the customer
Focus on value to the customer
Cross-org, customer-driven goals
Single definition of “done”

patterns & practices

Value stream mapping
Getting a big picture view of the
deployment pipeline
Identifying value and waste
Elevating bottlenecks
Creating “flow”

Balanced teams, engaged
management
Balanced teams with no handoffs
Favor generalist skills set, but don’t
eliminate specialties
Compress management hierarchy,
break up functional silos
Establish new workspace layout

Team standups
Brief opportunity to orient the team
Share milestones, blockers, help
needed
Share continuous improvement
activities with the group
Leaders must set an example

On-call engineering
Rotate team members through on-call
Responsible for service support
Puts the focus on the customer
Encourages instrumentation and
continuous improvement
Keeps the rest of the team focused

Agile planning of small batches
Ship often, get fast feedback
Product owner maintains backlog, team
decides when sprint is “full”
Tasks scoped to max of two days work
Scope doesn’t change
Team always ships at the end of a
sprint

Continuous integration
Daily check-ins to master branch
Some form of TDD is critical
Offers fast feedback, smaller problem
surface to investigate
“Green” builds are good, “red” should
not be feared

Treat infrastructure as code
All interactions with infrastructure done
via source-controlled artifacts
Make infrastructure work visible work
Establishes standardized processes
that are repeatable
Eliminate wasteful snowflake
environments

You build it, you run it
Long-lived service teams vs. project
teams
Develop customer empathy, foster
continuous improvement
Helps prevent overloaded releases
Improves stability of key services

Comprehensive telemetry
No guesswork during service
interruptions
Heavily instrumented environments,
with event-driven notifications
Need baselines to identify anomalies
Over-alerting causes fatigue

Blameless postmortems
(retrospectives)
Not about assigning blame
Understand the situation, timeline
ahead of the meeting
Use “Five Whys” technique to identify
root cause, not just source
Assign owners to action items

Standard tools and processes
Avoid unique toolchain across teams
Empower teams to suggest
improvements to standard tools
Automation is often, but not always, the
answer

Continuous delivery,
deployment
Should be boring process that doesn’t
cause service downtime
Consider what it takes to fully package
software for deployment
Partner with security and compliance
groups
Operations skills come in handy here

Show-and-tell sessions
Help others learn from your success
and failure
Create possibilities for reuse across
teams
Visibility helps other teams coordinate
their own milestones
Encourages experimentation and
continuous learning

Variations of DevOps
SREs maintain large systems at Google,
consult others
Platform Ops is about teams that
improve and operate platforms used by
others

Room for improvement in these core areas.

toolchain

The DevOps toolchain
Collaboration
Planning
Source control
Issue tracking
Configuration management
Development environments
Packaging, artifact management
CI / CD
Application platforms
Monitoring

Collaboration
Connect your teams
Improves situational awareness
Chat rooms, knowledge repositories
Quick collaboration prevents waiting
waste
Examples: Slack, Microsoft Teams,
GitHub Wikis

Planning
Improved visibility of upcoming work
Dynamic tools, not static documents
Easy to see bottlenecks, breakdown in
flow
Also makes sense to include bugs, toil
in planning view
Examples: Pivotal Tracker, Trello, Jira

Source control
Possibly the most important tool in a
DevOps environment
Store code, config, scripts
Avoid transportation waste of moving
code around unnecessarily
Change history is important
Examples: GitHub, GitLab, Bitbucket,
Subversion

Issue tracking
Single way to collect, triage, and
respond to issues
Avoid info lost in transit
Helps improve the feedback loop
Product defects likely tracked
elsewhere
Examples: ZenDesk, GitHub issues,
Jira

Configuration management
Enforce state of compute resources
Treat infrastructure as code, and avoid
configuration drift
Changes applied systematically
Sometimes used as infrastructure build
tools
Examples: Puppet, Chef, Ansible, Salt,
CFEngine, BOSH

Dev Environments
Consistent developer experiences
No more “works on my machine”
Mix of configuration management and
supportive environments
Cloud IDEs a fit here
Examples: Vagrant, Cloud9, Eclipse
Che

Packaging, artifact
management
Typically the output of a CI process
Improves flow, consistency
Multiple ways to package an app
Should package *all* corresponding
components
Examples: Docker, JFrog Artifactory,
BOSH

CI / CD
Keep code “inventory” small and get
fast feedback on code changes
Call out to code scanning, quality
components
Visible status drives accountability
Pipeline-based models are popular
Examples: Concourse, Jenkins,
CircleCI, Bamboo

Application platforms
Modern runtimes that simplify
deployment and ops at scale
Provide APIs for deploying and
managing apps
Designed to reduce toil and automate
lifecycle activities
Examples: Cloud Foundry, AWS
Elastic Beanstalk. Azure App Service

Monitoring
DevOps happiness may hinge on your
approach here
Need views of system health, not just
server/app health
Improves transparency, helps reduce
MTTR
Examples: New Relic, Dynatrace,
Datadog, ELK

Use DevOps principles and approaches
to deliver more customer-centric
software, faster.

DevOps Explained

More Related Content

What's hot

Similar to DevOps Explained

More from Richard Seroter

Recently uploaded

DevOps Explained