Need Advice an Automated / Pre-Configured Confluence 5.5 Installation.

Tobin Smith

I will need to deploy Conflouence to about 100 servers in a couple of months and am looking for a way to create a mostly preconfigured ond partially dynamically configured installation with the goal of rolling out these Confluence to the servers as quickly, efficiently and with as little manual configuration as possible.

Any suggestions or advice?

Thanks.

Mike Rathwell

Tobin,

100 servers - yikes

Anyway, assuming Linux (I am running Ubuntu 14.04) and running behind Tomcat (no Apache) I have the following process for a rather smaller variant of that where I migrate the production system to a test box to test upgrades or other biggish changes to the installation:

• Pre-reqs:
  • Make sure all target hosts have valid DNS names before the effort
  • Use "wildcard" certificates for SSL
  • For the above two, much simpler if all are on same domain
  • Preconfigure LDAP (if used) with a service account
  • Preconfigure any proxy servers with a service account if necessary (to allow Confluence to see outside your environment)
• When installing the first one, use defaults. They seem to be fine and makes it easy to replicate an installation from one server to another
• Get your first one configured up to the common base level for all 100 installations and thoroughly tested.
• To move to another server:
  1. Shut down Confluence
  2. Dump the database
  3. Tar the Confluence install directory (if default likely /var/atlassian/confluence)
  4. Tar the Confluence data directory (if default likely /opt/atlassian/confluence)
  5. Move the dump file and 2 tar files to the next system
  6. Create unix user running catalina on target system with same UID/GID as source system
  7. Create DB user with required permissions in backing store DB
  8. Restore DB Dump file
  9. Extract tar files in same locations as source system
  10. Start Confluence
  11. Fix the Server Base in the General Configuration

When I follow this process, I get, with little effort, a mirror of the source machine on the target. The only time consuming thing (since I have a largish installation) is moving the files to the target machine. If you're pre-configuring and have basically empty Confluence installations, that won't be as much of an issue. The bigger issue will be getting the host system automagically configured (users, permissions, port mapping, etc).

Maybe investigate deployment with Puppet so can be fully automated after first one is built?