Mapping Events to Elasticsearch

Overview of Index Mapping

In Meridian, event table entries contain references to associated node, asset, service, and journal message tables. In Elasticsearch, we must flatten these entries into a single index entry for each insertion. Thus, each index entry contains more context information than would be found in the actual Meridian event. This context information includes the associated node and asset table information which was current when (but may have changed since) the event was archived.

The table of index mappings below has an example event JSON entries retrieved using a sense command. The table illustrates how Meridian saves data in Elasticsearch.

Internal Elasticsearch fields always begin with an underscore character. The internal fields _id, _index and _type are combined to give the unique identifier for an entry as described above under index definitions. All of the fields under _source represent the stored event (Elasticsearch documentation refers to source entries as indexed documents). The ID of each event is included in the _source id field and also duplicated in the internal _id.

Events in the Meridian events table (i.e., those corresponding to logs or traps) are copied directly to the opennms-events-raw-`indexes. In Meridian events can contain `parameters that are key-value pairs referencing additional data stored when the event is created. In Elasticsearch these parameters are always stored in separate fields in the index, with names beginning with p_.

Events have severity fields defined as integers (long) and also corresponding severity_text fields that give the text equivalent (Critical, Major, Minor, Normal, Cleared).