Graph management

Graphs are handled by at least two endpoints : a storage endpoint for the global policy and a RRD endpoint for the physical file management. To manage graphs, you usually only need to interact with storage endpoints (either through configuration variables or by database update).

Graph creation

index_data

Before any graph creation, an entry in the index_data table must exist. Such entry is linked to either a service or a host. Centreon Broker does not insert data in this table, so it must exist prior to any graph operation. If you use Centreon, don’t worry : it will insert an entry in index_data as soon as you create a new host or service.

metrics

The metrics table contains the list of metrics managed by Centreon Broker. Each entry in this table will create a physical graph. Centreon Broker insert and update this table according to performance data provided by check plugins. The performance data format is explained in the Centreon Engine documentation.

data_bin

This last table store all performance data. Each entry is linked to an entry in the metrics table. This table is used mainly for graph rebuild as it contains all historical data of all graphs.

Graph update

Every time a plugin is executed, its perfdata is provided to a storage endpoint. This endpoint usually inserts the data in data_bin and then generates an event for use by one or multiple RRD endpoints.

Graph deletion

Physical graph deletion

If the graph (.rrd file) has been removed from the hard drive, it will automatically be recreated by the RRD endpoint. It will obviously be empty (no historical data) but can be rebuild using the standard procedure.

In DB graph deletion

Currently, Centreon Broker only handle whole host or service deletion (index_data). You cannot delete only one selected metric. To delete a host or service, flag its corresponding entry in the index_data table by setting the trashed flag to ‘1’. Then make Centreon Broker aware of this change (restart or SIGHUP). Centreon Broker will then remove entries from index_data, metrics and data_bin as requested. It will also generate deletion events for use by RRD endpoints. As a consequence all metrology history of the host or service will be deleted.