While the trace events can be viewed via the information schema MV_TRACE_EVENTS table, this buffer is not intended to be directly consumed by the end user. The PROCESS permission is required to query MV_TRACE_EVENTS.

The new information schema table represents semi-structured data to support any arbitrary event that occurs on a node. It also includes a JSON column for details specific to that event type. Queries against this table can also filter events using familiar SQL DML. Note that this is where event filtering happens. The DDL language used to create and configure trace events defines how an event is traced and may, depending on the options selected, result in some events not being traced.

Note that MV_TRACE_EVENTS is a ring buffer where the oldest events are overwritten when the buffer wraps around. To save events, other software (such as the Query History feature)  must gather them before they age out of MV_TRACE_EVENTS.

The rows are returned approximately in ascending order of time, grouped individually by node. As there is no guarantee that an event that occurred at a specific time will be inserted into the ring buffer before an event that occurs later, events are not guaranteed to be returned in chronological order.