DockerCon2018

JP Penny

• WANT: agile & lead, etc ms/cloud benefits. Cam and catalog service.
• microservices in immutable infra. IaC, check in scripts, create and scale infra on demand
  • allows polyglot persistence, good for their domain. Database replication +pic of their architecture
• No big bang release
• 30+ services came in 6 months after good results.
• ISSUES w/ Docker 1.11: connectivity loss b/t services, services spinning up at the same time, major outage. MITIGATE:
  • slow down onboatding
  • multiple docker clusters (one for f/e & b/e, for parallel deployments in 2 diff clusters)
  • vertically scaled k/v store
  • accerlate upgrade plan
  • Docker EE v17.03 - gove-live 3 mon, open issues resolved, faster deployment, scaling; self-healing, stable; IPSec support
  • scaled cluster to 4x needs, perf test well
  • entire ecommerce platform migrated in 2 years.
    • 120 microservinces, 100k deployments, 120mil business, 0 issue 2017 peak, onboarding store services
    • wrote docker deployment API. triggered by jenkins, talk to their deployment API
    • process visible to all stakeholders in dashboard
    • Jenkinsfile and deployment spec files (pic): joint managed by dev and ops, 1 spec for environment going to and services version. ALL use same specification

ADDL HIGHLIGHTS:

• on-demand releases, dynamic scaling, improved time to market, reduced cost

KEY TAKEAWAYS:

• controlled on-boarding of services to prove it worked and not take down system. Multi-cluster design helped, move to one first, reduce risk *- reduce friction - cloud agnostic helped running locally for devs fixing bugs! Frictionless upgrade from underlying platform
• Empower teams - break down walls b/t dev and ops, self-service / all tools needed, training, higher up support
• Plan for failure - failure isolation, think about failure when designing app
• Talent acquisition -

?

• how much training did you need to give your people?
• how much time to market was reduced?
• how much cost savings?