Speaker Aurynn Shaw
Time 2019-08-03 09:20
Conference PyCon Au 2019
Talk details Link

Eiara, DevOps

How to put stuff in prod without it catching fire.

Difficulty in getting new technologies used in production.

Negative language, emotions, used to describe people who disagree. Frustration, irritation, etc. Cultural issue.

Some things I’ve learned by talking to your boss


Eternal September. AOL linked to Usenet.

Linus Torvalds

“PHP is Bad”

Framework to understand system.

Communication unpleasant to outsiders. We need framework to enable us to ask questions.

The business does not care about technology.

The technology matters to us. So cool.

Experience: New technology will save business lots of money, or get better outcome, but we are ignored.

Business is focusing on predictability. Existing technology is understood, can be budgeted for, will allow staff to get paid on schedule. Risk.

Consequences. Communications gap. This gap is not the fault of the business.

We stand on the shoulders of mistakes. We are not trusted. We are sidelined, we exist to be worked around.

We build tools for ourselves. These tools build hostilities. No good build tools for Windows. Microsoft had to do this themselves. This is because so many communities consider Windows in low regard. Not only did we hold Windows in low regard, but also programmers, and users too.

10x engineer.

We have pushed ourselves to the sidelines. When we provide input concerning deadlines, etc, we get ignored. This is not the business fault. This is our fault.

We have to speak to the business, and do so on the terms of the business. We have to consider them as equals, not lessers then us. They have learnt over the users to ignore our inputs because we do treat them as lessers.

The business does care about technology.

…but I spent the entire first act saying the business doesn’t care about technology.

They do care about knowability. Defined boundaries. Engineers just want to be unnecessarily hardware.

The do care about technology in terms of “Well done. Good job team.”

Anxiety. Weeks when business is losing money when something not working correctly. Or regulatory issue. Observable outcomes.

Bad experience may have happened at previous employment. When moving to new job we bring new practices, but bring old anxieties.

Critical path. These people often hostile. “That guy” makes you feel small and ignorant. Generate impostors syndrome. These people generate resistance to changing or taking risks.

The cloud. Test, experiment, refine.

Questions. Business doesn’t care that we can spin up new instance in minutes. We are bad at saying how this affects risk, what the business cares about. This creates anxiety. We won’t tell business what they need to know about stability.

Change is a risk. Some times change is required. For various reasons.

Change can fail. Why? Communications gap. Why is this change valuable? What risks? How does it alter power dynamics?

Story. Roll out Docker. Having problems with release manager. Who wants to ensure build is stable. Release gets delayed. Lots of changes get incorporated into release. Was very risky.

Sat down with release manager. What are you concerned about? Listened to responses. Never: “No your wrong!” - which is part of the contempt culture. Instead: “Here is what we can do for you”. Plus “What did we miss”. Response: “I understand what we can do to reduce risk.”

“This is nonsense” - Different ideas from our own.

“This is the coolest thing since slice bread” - ideas do align with our own.

What we think we do: Work on technology.

What we should do: Work with business and political requirements.

What we actually do: We ignore business.

What we are saying is not working. Communicate risks, values, and benefits.

To talk about politics. First step, look at organisation chart. No, what is the real organisation chart. Real chart is based on who owns the deadlines. We can’t change the process without the ability to try the new process. When things catch fire the safest fail-back is to fall back to the old process.

Need guarantee that thing X will take exactly x hours. Impossible to guarantee. This means impossible to experiment as no time to experiment.

Concerns and needs. They will be defensive, as they will expect us to make their lives miserable. We have to go to them. They will not come to us.

Isolation. Info tech team in corner. Not working. Contempt culture. Security issues not being reported. Info sec team moved to center of office, so everything would have to walk past them. Measurement number of security events vs weight of lollies in jar. People stopped to take a lolly and have a chat. This made conversation safe. This meant security issues got reported more often. This improved conversation where we need to be involved.

Risk/benefit. Road maps.

How much data are you willing to loose? This tells out what level of backups we need. How long can we cope with network outages?

No such thing as perfect implementation. Just what meets our needs today, and can be expanded to meet our needs tomorrow.

We cannot build something for the future. If we try, we will be wrong. Wrong: “If only we had done this right.” Puts down all previous work done before, without understanding context and constraints with previous solution.

Linus stepped down. Everything I have done was bad. Now I understand.