How it works
Every record in the database has a corresponding entry in the Repligard table, and an associated unique guid. See Table 14-1 and Figure 14-2.
Table 14-1. The Repligard table
| guid | Globally unique id |
| id | id of the object in its table |
| realm | table name |
| changed | Unix timestamp for the last change of the record with id id in table realm |
| updated | Unix timestamp for the last update of the object by Repligard |
| action | The last action performed on the record with id id in table realm |
| sitegroup | Sitegroup id |
The basic principle is that, if changed is greater than updated, then either the record hasn't been replicated, or it has been updated after the last replication.
Data integrity is preserved thanks to the repligard.xml file. Please, notice the rlink element in the blobs and parameter (record_extension) sections. Beware that repligard.xml has to be modified following the same model if, for instance, you use any of the article extra fields with one such level of indirection.
repligard.conf is used to select which part of the database you want replicated:
The login username and password attributes allow you to restrict replication to a Sitegroup, or to a specific resource tree.
The replicate element enables further granularity in replication.