SlideShare a Scribd company logo
Painless Container Management
with Google Container Engine & Kubernetes
Matthew Feigal
Cloud Platform Solutions Engineer
feigal@google.com, @mattfgl
Jorrit Salverda
Technical Architect
jsalverda@travix.com
2
Online Travel Agency focused on air travel
● 28 countries
● 2.5+ million passengers / year
● 150 flight searches / second
● 100 applications
● 13 scrum teams
Travix: About Us
3
Early 2015
On-premise only
May 2015
First site in Kubernetes (still Alpha)
Today
50 applications in Kubernetes
Travix: Our Journey to Kubernetes
4
Deployments are very fast and reliable
● Deploy or rollback in half a minute
● Autoscale during deployment
Learnings
● Graceful termination to finish request
Travix: Our Experiences
5
Automatic restarts of rogue applications
● Less alerts and manual actions
● See restart count the next morning
Learnings
● Getting resource limits right takes time
Travix: Our Experiences
6
Pods group tightly coupled containers
Handle cross-cutting concerns in separate
containers
For example
● HAProxy for TLS termination
Travix: Our Experiences
7
Travix: Demo
8
What Kubernetes brought us today
● Insanely fast deployments
● Less alerts, less manual actions
● Simpler application code
What’s next?
● Canary deployments
● PetSets for stateful apps
Travix: Recap
9

More Related Content

Painless container management with Container Engine and Kubernetes

  • 1. Painless Container Management with Google Container Engine & Kubernetes Matthew Feigal Cloud Platform Solutions Engineer feigal@google.com, @mattfgl Jorrit Salverda Technical Architect jsalverda@travix.com
  • 2. 2 Online Travel Agency focused on air travel ● 28 countries ● 2.5+ million passengers / year ● 150 flight searches / second ● 100 applications ● 13 scrum teams Travix: About Us
  • 3. 3 Early 2015 On-premise only May 2015 First site in Kubernetes (still Alpha) Today 50 applications in Kubernetes Travix: Our Journey to Kubernetes
  • 4. 4 Deployments are very fast and reliable ● Deploy or rollback in half a minute ● Autoscale during deployment Learnings ● Graceful termination to finish request Travix: Our Experiences
  • 5. 5 Automatic restarts of rogue applications ● Less alerts and manual actions ● See restart count the next morning Learnings ● Getting resource limits right takes time Travix: Our Experiences
  • 6. 6 Pods group tightly coupled containers Handle cross-cutting concerns in separate containers For example ● HAProxy for TLS termination Travix: Our Experiences
  • 8. 8 What Kubernetes brought us today ● Insanely fast deployments ● Less alerts, less manual actions ● Simpler application code What’s next? ● Canary deployments ● PetSets for stateful apps Travix: Recap
  • 9. 9