Environments for Sync

There are four environments :

dev: development environment composed of three VM: storage, reg and sreg. The configuration files for this environment can be freely changed. The servers are maintained and modified by the developer.
sandbox: development environment composed of three VM: storage, reg and sreg. The configuration files are maintained by puppet.
stage: stage environment composed of three VM: storage, reg and sreg. The configuration files are maintained by puppet and are similar to the sandbox files.
prod: prod environment.

Continuous integration

Hudson builds the RPMs in the dev environment on every commit
Hudson runs the functional tests :
- against the dev environment on every commit
- against the sandbox environment every other hour
- against the stage environment every other hour
- against the production environment once per day

developers should be able to view at any time how sandbox, stage or prod is configured in Puppet. For this, the repositories that contains puppet files should be accessible in read-only mode by developers
Ops should be able to see the configuration files and their default state in each code repository, and ask for upgrades of their initial state.

There are two ways to upgrade the Sync servers:

Once we've reached a dev/sandbox/stage/prod stability, we need to do regular releases in time boxes. Every week seems to be a good start.

every Monday, the developer tags a release for every package when there are changes.
Ops upgrade stage on Tuesday ask for QA to check it
Ops upgrade prod on Wednesday

See the Deploying a new version section for more details.

Urgent fixes might be applied by Ops in prod environment, then backported to code for the next regular release.

XXX process TBD

the developer tags each repository using rpm-MAJOR.MINOR-release where release is the rpm release number in the spec file and MAJOR.MINOR the python version.
the developer deploys the RPMs in the sandbox and validates that it works
the developer possibly change some configuration files and ask to the Ops to push them in puppet
once everything works, the developer validates that there are no failures in Hudson against the sandbox environment,then tells the Ops to upgrade the stage environment.