Agent

dragent.analyzer

dragent is the main process in the agent that collects and collates data from multiple sources, including syscall events from the kernel in order to generate metrics. The analyzer module that runs in the dragent process does much of the work involved in generating metrics. These internal metrics are used to troubleshoot the health of the analyzer component.

Sysdig Monitor provides the following analyzer metrics:

Metrics

Type

Minimum Agent Version

Description

dragent.analyzer.processes

gauge

0.80.0 or above

The number of processes found by the analyzer.

dragent.analyzer.threads

The number of threads found by the analyzer.

dragent.analyzer.threads.dropped

counter

The number of threads not reported due to thread limits.

dragent.analyzer.containers

gauge

The number of containers found by the analyzer.

dragent.analyzer.javaprocs

The number of java processes found by the analyzer.

dragent.analyzer.appchecks

The number of application checks reporting to the analyzer.

dragent.analyzer.mesos.autodetect

If the agent is configured to autodetect a Mesos environment, value is 1, otherwise is 0.

dragent.analyzer.mesos.detected

If the agent actually found a Mesos environment, value is 1, otherwise, value is 0

dragent.analyzer.fp.pct100

The analyzer flush CPU % (0-100)

dragent.analyzer.fl.ms

The analyzer flush duration (milliseconds)

dragent.analyzer.sr

The current sampling ratio (1=all events, 2= half of events analyzed, 4=one fourth of events analyzed, and so on.

dragent.analyzer.n_evts

The number of events processed

dragent.analyzer.n_drops

The number of events dropped

dragent.analyzer.n_drops_buffer

The number of events dropped due to the buffer being full.

dragent.analyzer.n_preemptions

The number of driver preemptions.

dragent.analyzer.n_command_lines

The number of command lines collected and sent to the collector.

dragent.analyzer.command_line_cats.n_none

dragent.analyzer.n_container_healthcheck_command_lines

0.80.1 or above

The number of command lines identified as container health checks. This metric does not change even if healthcheck command lines are not sent to the collector.