Closed jpolchlo closed 3 years ago
The Docker resources have been restructured. There are now two images for the two distinct tasks, and the structure has been verified in use. This is better, since adding psql
to the spark container was a mess, since that image is now on an out of date debian release for which there are no longer apt profiles. Builds are faster and more maintainable for when we bump to Spark 3.
The purpose of this PR is manifold, but summed up as "simplify the deployment process". The aim is to consolidate as many steps as possible into make rules that can be executed reliably, without fuss.
Knock on benefits of this PR include the ability to finally retire the osmesa-stats-server by importing the necessary materialized view definitions and refresh infrastructure. This PR will also automate the process of rebuilding the changeset metadata from OSM DB extracts. This will also take a step towards eliminating the
deployment/batch
subdirectory, and ultimately allowing us to eliminate the distinction between batch and streaming deployments (though the batch infrastructure still provides the means to create EMR clusters for direct interaction (i.e., not pre-canned operations)).Closes #203 Closes #189