It is well known that ActiveObjects technology does not play well with multiple schemas in database.
Atlassian advice is to have separate database for each JIRA instance:
https://confluence.atlassian.com/jirakb/action-in-jira-server-fails-with-could-not-find-column-in-previously-parsed-query-error-in-log-315326688.html
There are also similar problem with Bitbucket:
https://community.developer.atlassian.com/t/critical-issue-with-active-objects-postrgresql-and-multiple-data-schema-migrations/24888
Recently we have started to use ActiveObjects technology for database model migrations on cloud environments where the library deals with thousands of schemas of PostgreSQL >9.6 database.
We were faced with improving the library to handle multiple schemas. It turned out to be pretty simple for PostgreSQL. In DatabaseMetaDataReaderImpl.java class getFields method jdbc driver DatabaseMetadata.getColumns method had passed null as schema name. Funny, because several lines further DatabaseMetadata.getPrimaryKeys method passed schema name as result of databaseProvider.getSchema().
Improving getting columns data by passing databaseProvider.getSchema() as schema name resolved the problem.
Why this improvement was not implemented by Atlassian? Does it conflict some database type/version?
