What is DevOps?

DevOps is NOT

• A set of tools
• A process / set of processes
• Something you make
• What you do
• What you are
• Devs pushing to Production with no tests

So if it's not these........

DevOps Is

Simply put:

"DevOps is a philosophy."

A concept that promotes:

• Transparency
• Collaboration
• Breaking down "silos"
• Lean Thinking
• Agile Methods
• Visibility

and so on, and so on,...

DevOps Battles Inhumanity!

Ticketing Systems as Tools of Oppression?

I created a ticket! It's your problem now!

On-Call Sucks!

Woken up in the middle of the night for something out of your control?

On call twice (or more) per month?

One-man Ops Team?

Agile Methods

More work with less people!

Responsibility with no Authority

We chose this soution, look after it! - Whoops it sucks!

The CAMS model

• Culture
• Automation
• Measurement
• Sharing

But...

Changing

Culture is hard

Start simple

DevOps is Laying the SMAC Down!

DevOps Starts with Sharing

• Blog posts
• Presentations
• Books
• Practices
• Attend MeetUps

Start the conversation!

If I Were to Keynote...

DevOps and Fighter Jets

DevOps scares people. - Why?

DevOps empowers
Continuous Improvement

Whether it's a goal or a by-product,
the philosophy of DevOps leads to
Continuous Improvement in those that adopt it.

The key to Continuous Improvement is lowering cycle time.

Low cycle time makes better fighter jets!

Let's pretend we're dogfighting

Even equal time:

• You can make and respond to 3 decisions.
• I can make and respond to 7 decisions.

Shorter cycle time = Quicker to learn and adapt

DevOps Makes Better Fighter Jets!

Reducing the time and reducing the barriers between
discovery, analysis, feedback and response
leads to quicker learning!

Continuing to Continuously Improve

"Do Not Fear Failure."

Failure Testing

Failure Testing = Taking control of failure!

Failure WILL happen

• Bugs in 'exceptional' code paths
• Systems won't recover as quickly as expected
• What is 'normal' when things aren't?

Inject Failure

1. Create a controlled test (Stop a service, restart a server, etc)
2. Alert and respond as "normal"
3. Analyze the results
   - What worked, what needs improvement?

SHARE THE RESULTS!

Successful Failure

System and Operational knowledge sharing

How it's supposed to work vs. How it acutally works

Highlights "untestable" systems

If you can't test it, it's not fault tolerant

How do we begin?

• Start small
• Automation can come later

Fail often...but fail TOGETHER!

Working With a Distributed Team

It's all about CAP.

• Consistency
• Availability
• Partition Tolerant

Consistency

Information and Transparency

Availability

Overlap and Face Time

Flexible Work Hours, Make yourself available!

Partition Tolerant

Small, flexible teams

Self-Organizing

Independent

Tooling?

Communication

• Ephemeral
  HipChat, Skype, Goto Meeting
• Persistent
  Email
• Permanent
  Wiki, JIRA, Pull Requests

Tooling?

Solving the "Over the Shoulder" Problem

• Screen sharing
• tmux/wemux-ing
• Consistent Tooling
  IntelliJ, rvm/ruby, SQL Developer

Anything Else?

Keep Up Process!
Standups, Retros, 1-on-1's

Share Metrics!
All about transparency

Culture - Like the people!

Keep the Sharing Going!

IGNITE Talks

IGNITE Talks are quick, organized presentations on a small, but complete topic

• 20 slides
• 5 minutes
• Slides auto-advance every 15s

Open Space Concept

Stimulate diverse, AdHoc conversations between particpants

Participants submit a list of topics they want to discuss

The group votes for the popularity / interest in each topic

The group decides how to split up the selected topics
For example, three concurrent groups, 45 minutes each, three time blocks

Participants can join whichever discussion they want, for however long they want.

Conversations are unstructured and informal.

Tools

Docker

Container technology is very new, but very exciting!

"Containerizing" tools and processes creates a division of tasks.

Docker Containers for Workstations - Each container has the tools and setup to accomplish a specific task.

Container != VM

• SoliumScripts
• Wrecker Dev Setup
• "Phoenix" Sensu clients

Ansible

Light-weight configuration management tool.

Shallow learning curve.

Wiki Setup Page of Doom? --> Ansible playbook.

Concentric Shell Scripts? --> Ansible Playbook

serverspec

Test your infrastructure!

Validate your Configuration Management.

We would want it for PAAS.

Leads to "Test Driven Infrastructure".

Decoupled from Configuration Management.

3rd Party Services

PagerDuty

Still no.

DataDog

Platform monitoring tool.

It's not NewRelic - Focused on 'Capacity Monitoring'.

No application traces - Not for code optimization.

Java / JMX Compatible.

Integrations for Nagios (in) and OpsGenie (out).

Open sourced plugins - We could write our own!

Continuous Devilry

It is NOT:

Continuous Deployment

Automating ALL THE THINGS!

It IS:

Eliminating bottlenecks in delivering software!

Some blockers:

• Releases disrupt customers
• Operations can handle increased releases
• Ops and Dev can't support more releases
• Releases tied to DB changes

Continuous Delivery is about removing blockers!

Guerilla DevOps Warfare

Conquering the Impossible Project!

1. What comes first Culture or Technology?
2. Getting others to see the Whole System.
3. Architecture Matters - Everywhere!

Culture or Technology?

Neither!

Reflexive Feedback Loops - Changing one (or two) changes the others!

Seeing The System

Map out your processes to see where bottlenecks exists!

Define your end goals!
Especially when choosing technology!

Did I mention SHARING?

Architecture Matters

EVERYWHERE!

How TELUS is Doing It

TELUS (~43K Employees) is starting to employ DevOps principles

They use a "Transformation Team" to:

• Evaluate existing workflows
• Empower teams to breakdown Dev and Ops barriers
• Promote "Change Champions" to carry the message

How are they doing?

Two years into a five year plan.

"Planted the seeds" into various small groups of TELUS.

Starting to see small returns in small groups - Koodo

Executive Buy-in! - from the CEO!

In Conclusion

Lessons Learned

Our deployments are pretty good!
"Two button deployment!"

We are leading the charge on ChatOps!
Deploy spikes, alerts, LiveBot, Front-End Business

There's no alternative to Liquibase.
You're using it or you're not
Everyone is as bad at schema migrations as us

We're doing DevOps with the Business!
JIRA Workflows, ChatOps, Dashboards, Post-mortems

Takeaways

We should be running PaaS.

Let's amp up our serverspec efforts.
Even for manually created infrastructure

Logstash - nohup is good hup.

Start sharing with Dev!
Guerilla DevOps Warfare

Moar MeetUps!

More Takeaways

Give DataDog a shot.

Dockerize sensu-clients.

What are our barriers for delivery?

lita.io for LiveBot?

Fun Facts

The Cloud = SAAS + Amazon + 'A bunch of stuff that doesn't really matter'

A small box of business cards looks like a juice box in an airport baggage scanner...