Skip to content

Latest commit

 

History

History
125 lines (85 loc) · 3.23 KB

README.md

File metadata and controls

125 lines (85 loc) · 3.23 KB

Making an Internal Kubernetes Offering Generally Available

Background Slides

Speakers

  • James Wen, Spotify

Spotify is currently migrating their internal services from Helios to Kubernetes. This talk covers their experience and learnings they have obtained during this process.

Operation Solidarity

Multicluster Strategy: Management

Offload cluster management to an automation tool (e.g. Terraform) or managed service if possible.

Multicluster Strategy: Deployment

Keep it simple for developers to reason about.

They have a deployment abstraction layer (Tugboat) that can deploy to their previous platform or Kubernetes. They also have a service they call Compass that determines which Kubernetes cluster(s) applications should be deployed to (e.g. sticky certain apps to particular clusters, deploy the app to N clusters, etc.)

Operational Metrics/Alerts

Establish trust through monitoring.

  • Monitoring on integration components
    • Error
    • Latency
    • Saturation
  • Smoke test —> periodic deploys
  • Metrics for:
    • Available cluster capacity
    • Status of backups
    • Scheduling

DiRTing

Plan for and simulate failures.

  • Disaster Recovery Testing
    • Planned and executed DiRTs
      • Cluster deletions
      • Master being down
    • Validation and improvement of
      • Metrics + Alerting
      • Playbooks
      • Backups

Business Impact

Track enough data to answer questions.

Metrics Tracking

Metrics/Dashboard to track:

  • Resource Utilization
  • Cost
  • % apps migrated
  • Developer productivity

Cost Tracking

Cost tracking was an effective motivator for many teams to adopt Kubernetes on their own.

Self-Service

Self-Service Experience: Namespaces

Take complexity for your developers.

Allow clients to self-service create namespaces

Self-Service Experience: Apps

Use existing tools/abstractions.

Their existing applications had a configuration file for Helios and tooling (Tugboat) that reads their Helios config to deploy. They extended Tugboat to also support Kubernetes yaml files, allowing their development teams to migrate without changing their deployment process.

Documentation

You as owners are not the audience.

Shared Support

Support can be mutually beneficial.

  • Developers will always have questions
  • US East Coast + Europe
  • Support shifts covering both, where people on these shifts expect to be interrupted to handle client inquiries/issues

Feature Parity

Developers need feature parity or better.

The new Kubernetes offering must have feature parity or more with the predecessor offering

Spotify's GA Requirements

  • Operational Solidity
  • Business Impact
  • Self Service Experience
  • Important Feature Parity

Migration & GA Learnings

  • Be mindful and deliberate in dictating terminology
  • Migrate in stages/series of goals that increase in scope
  • Devs will adopt and/or migrate to your Kubernetes offering if it is easy, complete, and compelling
  • Talk to other companies about how they do infrastructure/solve infrastructure problems