How it works

The publisher hooks into the Spring Boot lifecycle to read and upload the database schema once the application is ready, without affecting business logic or startup time.

Startup flow

1. DbSchemaPublisherEventListener listens for the Spring ApplicationReadyEvent.
2. It calls DbSchemaPublisher.publishDatabaseSchemaAsync(), which is @Async on a dedicated single-thread task executor (dbSchemaPublisherTaskExecutor). The upload therefore runs in the background and never blocks startup.
3. DatabaseModelReader opens a JDBC connection from the application DataSource and reads the schema (named by jeap.archrepo.database.schema-name, default data) from DatabaseMetaData.
4. The result is wrapped in a CreateOrUpdateDbSchemaDto (the system component name is spring.application.name) and posted to the archrepo at POST /api/dbschemas.
5. The operation is optionally wrapped by TracingTimer in a Micrometer span (publish-db-schema) and timer (jeap-publish-database-schema, tagged status=success|error) when a Tracer and MeterRegistry are present.

The whole upload is best-effort: any exception is caught and logged as Failed to publish database schema; the application keeps running.

The schema model

DatabaseModelReader (module jeap-db-schema-publisher-model-reader) maps JDBC metadata into a set of immutable records:

Record	Fields
DatabaseSchema	name, version, tables
Table	name, columns, foreignKeys, primaryKey
TableColumn	name, type, nullable
TablePrimaryKey	name, columnNames
TableForeignKey	name, columnNames, referencedTableName, referencedColumnNames

Foreign keys that span multiple columns are grouped by foreign-key name. The version field is the application version resolved by AppVersionProvider from BuildProperties, then GitProperties (git.build.version), falling back to na if neither is available.

Related

• Getting started
• Configuration reference
• Authentication
• jeap-db-schema-publisher