This is a new feature in version 6.2.
This is only typically needed when support is investigating performance issues
on the server. Support may ask that you activate this when thread or
memory dumps or cpu samples are needed from the two server components. We already have
standard self-monitoring which returns host machine metrics and JVM metrics so
this was added to further analyze a under-performing a problematic component.
This can disabled this in the dynaTrace server
settings under the Settings link and the last tab 'General', there is a
checkbox for Use Agent-base
Self-Monitoring. This will stop the agents and internal collector
on the server once the server has been restarted. This is the only way to
stop/restart these agents and collector as they are spawned by the Dynatrace
Since this is more of an internal support feature and not normally used unless Dynatrace Support asks you to enable this feature, there is no User level documentation on its usage directly. But you can find references to it, such as in the Alerting and Incidents doc.
Dynatrace Vendor support teams always have available a very rich set of metrics to perform their own local support activities. Check out the dashboards available under the Deployment Health section of Status Overview. In the event a problem occurs with the AppMon product and Dynatrace Support team needs deeper metrics, the "Agent-Based Self-Monitoring" can also be enabled, however the resulting data is likely to be of no value to someone who didn't write the Dynatrace Component code itself.
Please advise DT support not to forget to disable "Agent Based Self-Monitoring". If the the agents and internal collector heap size together with the DT Server and Frontend Server heap sizes bigger than Physical Memory available there will be measurable paging and significant performance degradation of DT infrastructure.