Monitoring event performance

Event processing metrics are internal counts maintained on cell performance categories in connection with event processing. These metrics count the following items:

Number of events received, including erroneous ones.
Number of events containing errors.
Number of events dropped by rules (Filter, Regulate).
Number of events removed from the event repository during cleanup.
Number of events propagated, including sendto.
Number of events added to the event repository, that is, entering the permanent context.

Calculation is performed on a 60-second basis, so every minute the counters are restarted. Counts from the last five minutes are retained. Running counters are reset only on demand.

The resulting metrics are:

Short term—total count of the last complete one-minute interval
Medium ter —sum of the five last, completed one-minute intervals
Long term—running total

Metrics are stored in MC_CELL_METRIC data objects, one object instance for each metric. Each metric displays the subject. For each of the short-, medium-, and long-term results, it contains the length of the interval, in seconds, and the total count. An average per second is also provided, rounded to an integer. Other averages per second, minute, or hour can be calculated by the application from this information, if needed.

A configuration parameter, CellMetricsEnabled, determines whether metrics are collected or not.

The mcontrol CLI is used to switch metric collection on and off, and to reset the counters. The duration of the short- and medium-term metrics are configurable using the cell configuration parameters. For more information, see Cell configuration parameters. Short- and medium-term metrics are reset whenever metrics are disabled. Metrics can be retrieved through rules by data object access, or through a command. The mgetinfo CLI can use that command.

The received event counter does not include incoming messages that cannot be parsed as events. It does include events of nonexistent classes or events with erroneous slots. These are added to the erroneous event counter. Internally generated events are counted as received events. Dropped events include those that are dropped when an event with the same universal ID exists.

The following table lists the metrics data objects MC_CELL_METRIC slots.

MC_CELL_METRIC slots

Slot	Description
`description`	Metric description
`long_average`	Long-term average, per second
`long_interval`	Long-term interval lengths, in seconds
`long_total`	Long-term total count
`medium_average`	Medium-term average, per second
`medium_interval`	Medium-term interval lengths, in seconds
`medium_total`	Medium-term total count
`short_average`	Short-term average, per second
`short_interval`	Short-term interval lengths, in seconds
`short_total`	Short-term total count
`subject`	Metric subject name

Available subject names are

ReceivedEvents
ErrorEvents
DroppedEvents
StoredEvents
RemovedEvents
PropagatedEvents

CLI commands for collecting metrics are:

mcontrol metrics on|off|reset
mcontrol metrics interval 60 amount 5
mgetinfo [-v] metrics

Page tree

Monitoring event performance