The Strangler Pipeline - Delivering Greenfield and Legacy Applications En Masse

622051376c3b381ef1c786036011112d?s=47 Steve Smith
December 13, 2012

The Strangler Pipeline - Delivering Greenfield and Legacy Applications En Masse

The Strangler Pipeline - Delivering Greenfield and Legacy Applications En Masse

622051376c3b381ef1c786036011112d?s=128

Steve Smith

December 13, 2012
Tweet

Transcript

  1. 1.

    The Strangler Pipeline Continuously Delivering Greenfield and Legacy Applications En

    Masse Steve Smith - @agilestevesmith www.stephen-smith.co.uk
  2. 2.

    What is Continuous Delivery? • Improve cycle time via automated

    pipeline • Reduces risk of release, improves time to market • Bridges Development and Operations teams
  3. 3.

    Continuous Delivery Principles “It is essential that everybody in the

    organisation is involved in this process” Dave Farley & Jez Humble
  4. 5.
  5. 6.

    Buy or Build? Continuous Delivery not core business domain, but...

    Build low-cost pipeline based on autonomation
  6. 7.

    Different Start and Stop • Disparate start/stop scripts caused recurring

    issues • Operations start/stop scripts coupled to Tomcat, incompatible with new applications • Application binaries adapted to Binary Interface – thin start/stop/status abstraction for Operations
  7. 8.

    Pattern: Binary Interface Pipeline will not scale if coupled to

    applications Expose binary capabilities, not implementation
  8. 9.

    Different Deployers - Scale • Different applications, technologies, lifecycles... •

    Commit application stage an interface • Commit platform stage creates Aggregate Binaries – manifests of constituent application versions • Deploy Aggregate Binaries incrementally
  9. 11.

    Different Deployers - Strangle • Must support legacy and greenfield

    mechanisms • Wary of imposing upfront uniformity across estate • Stage Strangler de-risks and prioritises application support
  10. 13.

    Continuous Delivery Caveat • Achieved goal of retaining 26 day

    cycle time for legacy platforms, 18-23 days for greenfield platforms • Wait times between silos prevent further improvement • Organisational change necessary to optimise cycle time
  11. 14.

    Spot The System Failure ...but 63 day lead time to

    Production! Messaging 186-13 in pipeline for ~40 seconds...
  12. 15.

    Change Management • Publish pipeline vision • Respect participants in

    value stream • Encourage early users • Gradually permeate silos