Recently i migrated a testing instance of Bitbucket from Postgres to MSSQL, and this move was a real breeze. Login to BB, click migrate data, enter credentials, do some small tests and BOOM data was migrated without issues and minimum downtime.
However when it comes to Jira and Confluence it's more of a complete reinstall, with all things like copying attachments, application links, reinstalling plugins (sometimes running versions that are no longer the version on the market place, meaning manual installs), customization of files for reverse proxies.
I keep asking myself, why can't it be as easy as it was with Bitbucket? What stops us doing a mysqldump, run a conversions script that creates a MSSQL dump that we can later just import, change the database driver and credentials and be done?
My main reason to ask is because i fear an as extensive rollback as the "migration" was, which would render our systems down if it turns out MSSQL would create performance issues and the fear that we can never get back again. Sure things can be tested in it's essense but simulating the same load our production systems have is not a small task, the proof will always be in the pooding.
Questions:
Have anyone tried to simply convert the dumps and change the database configuration? What issues did you run into? Can it be done? What stops us from doing this?
