Multiple Projects in Single Database
I wanted to be able to generate reports (show me all of my tickets) across all of my trac projects. With each project having its own database, that's problematic. What I wanted was a single database for all of my projects, but I didn't want to spend a lot of time changing the way Trac used the database.
What I did was create a single postgres database for all projects. I modified all of trac's tables to include a db_user column as part of the primary key. I also renamed all of the tables to have a
master_ prefix. Then I created an updatable view for each
master_ table that looks identical to the original trac table. The view only selects rows where db_user is equal to the current database user, and the view modifies INSERT statements to include the name of the current database user.
It appears to be working. I haven't tested everything yet, so there might still be some problems.
- Windows 2000
- Trac 0.9
- PostgreSQL 8.1.0-2
- Python 2.4
- Install mxDateTime (I think this is required by pyPgSQL. You might not need it for a different driver)
- Install pyPgSQL
- Install PostgreSQL
- As the postgres user, create your database (I called mine
- OPTIONAL: Create a
tracschema. This is useful for keeping trac's tables separated from other projects.
- As the
postgresif you didn't create the user/schema), execute trac_ddl.sql to create the tables and views.
That's it for the database setup.
Creating a Project
- Create a postgres user for the project. Each project must have a unique database user. If you created a
tracschema above, be sure to set the
search_pathfor the user to look in that schema first.
- If you created a
tracschema, grant the new user USAGE privileges on the schema.
- Grant the new user SELECT, UPDATE, INSERT, and DELETE on all views and the two sequences in the database. I use trac_create_user.sql for this.
- As the new database user, load the initial project data into the database using trac_data.sql. Normally
trac-admin initenvwould do this, but that command also tries to create the database structure so we can't use it.
- Initialize your new Trac environment with
trac-admin initenv. IMPORTANT: Don't try to initialize the postgres database. I use a temporary (the default) sqlite database during this step.
- In your new environment, edit
conf/trac.inito point to the postgres database.
- There is no step 7.
- I'm pretty sure automatic upgrades won't work with this setup. Buyer beware and all that.
- Since multiple projects are using the same ticket id sequences, there will be gaps in the ticket numbers for each project. I don't think this is a big deal, but I thought I'd mention it.
- I don't know how this affects version control features, since I have those turned off right now.
Using this as inspiration I also create a different user for each trac project in the postgres database but I then create a schema for each user. Any tables in the database which need to be different for each project stay in the user schema while all the common ones (tickets, etc) stay in the public schema. This has the advantage that initenv and possibly upgrade all work with the caveat that you may need to go in and remove tables from the user schema that get created. I have this mostly working on a standard Centis 4.3 system using trac /trunk (0.1dev). I can now select which parts of trac are per project and which are global. greg@…
- You could also simply use an rss agregator that links on all timeline's RSS
trac_ddl.sql (21.5 KB) - added by 10 years ago.
SQL script to create tables and views
trac_create_user.sql (1.9 KB) - added by 10 years ago.
SQL script to create and provision a new database user
trac_data.sql (149.9 KB) - added by 10 years ago.
SQL script for populating a new project database
Download all attachments as: .zip