This may not be all that relevant given the likelihood of switching to phabricator in the medium term.

In that case we can re-purpose the bug to make Phabricator's database available on labs after the switch happens.

We are also not switching Gerrit to Phabricator yet - just Bugzilla and RT.

No idea what "Phabricator" database this is about. Everything?
Plus no usecase described here at all.
Plus I don't see how to sanitize that data (legal).

I think the original use case (If I recall correctly from IRC) was to get gerrit /and bugzilla/ databases available from labs.

I don't think there's a problem in principle, though it will require a great deal of consideration. (Sanitation in particular - it's not an insurmountable problem since we do exactly that for the project databases).

I mostly intended to keep this bug open as a task tracker for the longer term.

(In reply to Andre Klapper from comment #4)

No idea what "Phabricator" database this is about. Everything?
Plus no usecase described here at all.
Plus I don't see how to sanitize that data (legal).

Yes, everything that is not-private.

Usecase is to create tools like [[mw:Gerrit/Reports]] that don't need to rely on hacks like bug 52329.

In T52422#550927, @Aklapper wrote:

Plus I don't see how to sanitize that data (legal).

The same way we do it for MediaWiki databases?

In T52422#941131, @Krenair wrote:

In T52422#550927, @Aklapper wrote:

Plus I don't see how to sanitize that data (legal).

The same way we do it for MediaWiki databases?

Feel free to provide links to how it's done for MW DBs if it can be done "the same way".

In T52422#941897, @Aklapper wrote:

Feel free to provide links to how it's done for MW DBs if it can be done "the same way".

A couple of code links:

• https://github.com/wikimedia/operations-software-labsdb-auditor
• https://github.com/wikimedia/operations-software/blob/master/dbtools/repl_prepare_schema.sh

I don't believe there's an abundance of documentation about this, but @Springle and @yuvipanda can probably provide more background.

Things you would need:

1. Legal OKying things
2. List of tables that can be fully replicated, without any redaction
3. List of tables that need to have some data redacted, and
   1. Which columns must be fully redacted
   2. Which columns must be conditionally redacted, and what those conditions are

Then we'll have to:

1. Set up triggers to do the conditional and unconditional nulling in the sanitarium DBs
2. Set up views to expose the data to labsdb users.

Lots of manual work + lots of DBA work :)