Skip to content

Latest commit

 

History

History
54 lines (44 loc) · 2.1 KB

DEPLOYMENT-PROCESS.md

File metadata and controls

54 lines (44 loc) · 2.1 KB
Raw

HAPI COMPONENTS

  1. HAPI-SCHEMA release 0.8.17

  2. HAPI-PIPELINES plays dev role: main branch is for (code) -> db-export branch (CSV daily depends on HAPI-SCHEMA release 0.8.17)

  3. HAPI-PIPELINES-PROD plays prod role: main (code) <- PRs from HAPI-PIPELINES -> db-export (CSV daily depends on HAPI-SCHEMA release 0.8.17)

  4. HWA 0.0.7 depends on HAPI-SCHEMA -> imports db-export (CSV daily depends on HAPI-SCHEMA release 0.8.17) from HAPI-PIPELINES-PROD / HAPI-PIPELINES -> postgres

  5. HAPI API 0.7.0 depends on HAPI-SCHEMA release 0.8.17 serves data from postgres

  6. HAPI SMOKE TESTS depend on endpoint schema, which in turn, depends on HAPI SCHEMA

PROCESS:

  1. HAPI-SCHEMA has a new release ( ex: deletes a column from an existing table, adds a new table ), 0.9.0

  2. HAPI-PIPELINES use HAPI-SCHEMA 0.9.0 -> produces 0.9.0 CSVs

  3. [Ticket] HAPI API to use new 0.9.0 HAPI-SCHEMA

  4. [Ticket] HWA to use new 0.9.0 HAPI-SCHEMA. Special care need to be taken when creating alembic revisions

    • Enum changes are not automatically detected by alembic
    • nullable=False constraints can pose problems with existing data
    • changes to primary keys can pose problems to existing data

    NOTE: 3 and 4 can be implemented in parallel

  5. Deploy on dev server:

    • Deploy [dev] HWA
    • Run in parallel:
      • Run HWA ingest
      • Deploy [dev] HAPI
  • Manual Testing
  1. [Ticket] HAPI SMOKE TESTS update
  2. Test HAPI SMOKE TESTS run successfully on dev server
  3. HWA new version 0.0.8
    • make sure to check for enum changes when creating alembic patch
  4. HAPI API new version 0.8.0
  5. PR from HAPI-PIPELINES to HAPI-PIPELINES-PROD -> generate new 0.9.0 CSVs
  6. Deploy on prod server:
    • Deploy [0.0.8] HWA
    • Run in parallel:
      • Run HWA ingest
      • Deploy [0.8.0] HAPI
    • Manual Testing
    • Run smoke tests on prod

IDEAS FOR A BRIGHTER FUTURE:

  • HWA to create a new postgres schema for each new HAPI SCHEMA release
  • Stop using postgres, use something that supports faceting out of the box (solr, elastic, mongo)