The harvesting engine does not store harvesting results. This implies that if the server is restarted the last results are lost.
Changes to the harvesting parameters (for example privileges and categories) are taken into account in the next harvesting run.
During harvesting, site icons are harvested and local copies updated. Icons are propagated to new nodes as soon as these nodes harvest from this one.
The metadata uuid is taken from the info.xml file of the MEF bundle. Any uuid stored inside the metadata will be overwritten with this one.
The same metadata could be harvested several times by different harvesting nodes. Anyway, this is not a good practice because every copy of the metadata will have a different uuid and the system will fill with different copies of the same metadata.
If the dct:modified element is missing from the GetRecords response the metadata will be always harvested.
Any exception during getRecordById operation is discarded and the metadata skipped.
The id of the remote server must be a uuid. If not, metadata can be harvested but during hierarchical propagation id clashes could corrupt harvested metadata.
During harvesting, GeoNetwork will try to autodetect the schema of each metadata. If the schema is not supported the metadata is skipped.