5.1.5 Git

GIT (man 1 git) is used to publish meta-data and to log operations performed on these files. There is one gitbare repository for each server one each host. This repository handle one module for the local server’s meta-data, and one for each collection on the “master” servers only. Remote servers access the “master” (or so called “bare”) GIT collection’s repository by using Ssh (see Ssh). The Apache (see Apache) configuration files and the Ssh’s public keys are also distributed among servers using GIT. The GIT configuration and collection modules are setup respectively by init-remove-purge (see init-remove-purge) and new-free-clean (see new-free-clean) scripts.

Process conceptual model:

• Events in

  $ git init --bare

  $ git config

  from init-remove-purge (see init-remove-purge) or new-free-clean (see new-free-clean): initialise a GIT repository.

  $ git clone

  from new-free-clean (see new-free-clean): checkout a GIT collection module.

  $ git config

  $ git commit

  $ git pull

  $ git push

  from upgrade-commit-pull-push (see upgrade-commit-pull-push): synchronise a GIT module.

• Processing

  Synchronise the meta-data files between servers.

• Events out

  $ ssh

  to Ssh (see Ssh): access the GIT bare repository.

• Data

  /var/lib/mediatex/mdtx/mdtx

  /var/lib/mediatex/mdtx/mdtx-COLL

  The GIT bare repositories

  ~/mdtx/git/mdtx

  ~/mdtx/git/mdtx-COLL

  The GIT cloned modules.

The configuration of the client repository is manage by MEDIATEX upgrade (see Scenario E).

Code:

scripts/lib/git.sh