Infrastructure · operations · Current

Modernizing an enterprise GIS: cloud migration, a major version upgrade and 100+ shared services

Modernized a legacy ArcGIS Enterprise deployment: moved it from on-premises servers to the cloud, brought it up to current practice with all code in versioned repositories, carried it through a major version upgrade, and rightsized it to save $30K a year. It serves 100+ map and feature services, holding 150+ feature classes, to GIS and non-GIS departments.

Organization
FilmLA
Year
Current
$30K
saved a year by rightsizing
99.98%
service uptime
Real-time
web-based boundary editing
500K+
service requests a month

Results

  • Moved to the cloud and rebuilt to current practice, with all code in versioned repositories.
  • A major version upgrade, with custom plugins and infrastructure integrations migrated.
  • One-click releases; a rejected release rolls back and goes back to dev for the full cycle.
  • One set of authoritative layers, with role-based access, instead of copies kept by each team.

Stack

  • ArcGIS Enterprise
  • Enterprise geodatabase
  • PostgreSQL
  • ArcGIS REST services
  • Cloud migration
  • Versioned code repositories
  • Chef / Cinc
  • Apache JMeter
  • Jenkins CI/CD
  • Role-based access
  • Monitoring
  • IT service management
Diagram: a legacy on-premises ArcGIS Enterprise deployment moved to the cloud and rebuilt with all code in versioned repositories, into separate dev, test and prod environments, then taken through a major version upgrade. Changes are built in dev, tested in test and, once approved through IT change management, deployed to prod in one click. A rejected release rolls prod back to the last release and goes back to dev to run the full cycle again. Patching, upgrades and monitoring with alerting run alongside.
Built in dev, checked in test, released to prod; a rejected release goes back to dev for the full cycle.

Modernized, not just moved

Moving off the on-premises servers was the chance to bring the deployment up to current practice. All code and configuration went into versioned repositories, the servers are built from code (Chef, through its open-source build Cinc) instead of by hand, and every change runs through dev and test before prod. Approved changes deploy in one click. A rejected release rolls prod back to the last release and goes back to dev to run the full cycle again.

A major version upgrade

The upgrade crossed a major version, so the custom plugins and infrastructure integrations had to be migrated with it. Test was built and migrated in full before prod moved.

  • Written down. Eight runbooks document the upgrade.
  • Revalidated. About ten web apps and widgets were checked before release.
  • Load tested. Apache JMeter ran the new servers under load before users moved over.

Day-to-day operations

  • Nightly refreshes of outside data, such as EV charger locations from NREL.
  • Data copies between environments, with a schema compare first.
  • Rotating PostgreSQL backups, Jenkins CI/CD, and bulk-update tools built on the Portal REST API.
  • Monitoring and alerting on the services.

One platform instead of silos

The GIS supports FilmLA’s film permit process and connects the agencies involved, and their data, into one process. Each dataset lives once in the enterprise geodatabase and is published as web services, so teams use the same layer instead of keeping their own copies. Role-based access decides who can view or edit each one.

Diagram: partner agency and in-house layers stored once in an enterprise geodatabase and published as 100+ map and feature services holding 150+ feature classes. Role-based access sends each role to the apps it uses, each in its own colour: the public views web maps, editors view and edit in apps, staff view reports, and analysts read the data warehouse. The roles are illustrative. The platform supports FilmLA’s film permit process.
Stored once, published as shared services, and each role sees what it needs.

The service catalog

Each service is published twice, as a map service for display and a feature service for queries. About half are partner data from city, county, state and federal agencies, refreshed as those sources change; the rest are derived in house. They cover airspace, parks and natural areas, transportation (streets, traffic, bikes, parking and EV charging), jurisdictions and parcels, hazards such as fire zones and wildfire perimeters, and events and restrictions.