Continuous Delivery is a solved problem. Really? A lot of organisations still struggle for many reasons and Confluence was one of them. What if the world your engineering team lives in is not perfect. This talk is all about making continues deployment happen in a real and not so perfect engineering organisation. This talk is all about the options you have a to achieve a goal without trying to solve all the problems at once.
The talk will start with setting the scene of Confluence¨s build and deployment process. There will be some interesting insights in the sheer size of builds, tests and the complexities we have around deployments. Once the stage is set for Confluence¨s environment, the next step is to explore the problems that are preventing the team to move faster both culturally and technically. To name a few: Build structure, build times, flakey tests, lack of automation, wrong culture, dealing with release blockers, test automation, post-deployment verification, monitoring and operating a production system.
With the problems defined, we will go into them in more detail and also learn about the ways the Confluence team resolved them. To wrap it all up we go through the current situation and explain how far we have come with the strategies taken and implemented by the team. As a bonus, a sneak peak of the road ahead for Confluence's Continuous Deployment and Site Reliability Engineering will be shared with the audience.
1 of 93
Download to read offline
More Related Content
Dev opstalks 2018 releasing the monolith on a daily basis
1. VINCENT KOK | ENGINEERING MANAGER, TRELLO | @VINCENTKOK
Releasing the Monolith
On a daily basis
21. ? Big cool statistic
? 2,56
9
? Add-Ons in Marketplace
Culture
https://?ic.kr/p/DsetEn
23. Experience
The way you¨ve been
shaped
Beliefs
Based on your
experience
Actions
Based on your beliefs
Results
The outcome feeding
your experience
The Oz Principle
52. ? Big cool statistic
? 2,56
9
? Add-Ons in Marketplace
Sustainable system
https://?ic.kr/p/awVLLs
53. Early results
Visibility
Create awareness across
the team
Measure
Gather data to allow data
driven decisions
Move the needle
Don¨t boil the ocean, focus
on a set of smaller results
66. The impact of flakes
Build times
Restarts cause build times
to increase drastically
Culture
The culture of green goes
out of the window
Broken windows
It starts with one. Soon, there
will be a lot
67. Retries as a painkiller
Automatic retries relieve some pain.
This is an early result, so see it as a
work-around.
69. ? Big cool statistic
? 2,56
9
? Add-Ons in Marketplace
Fix the root cause
https://?ic.kr/p/J44Nm
76. ? Big cool statistic
? 2,56
9
? Add-Ons in Marketplace
Congestion
https://?ic.kr/p/bRgBZP
77. Ends at master Big bang Releases on
hold
Dog-fooding
becomes hard
Impact
78. For once a free lunch
Fork and Join solved most of our
problems.
79. 12 FACTOR CONFLUENCE
Health check
200 app is alive. 500 app is unhealthy,
destroy the node
Stateless*
Run as many nodes as you need
Expose a port
Only access to the service