You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Motivation : currently the Metrics SPI expose the measured instances to the SPI implementation when at creation of the corresponding metrics instance. This instance is actually not really used by the implementations (Vertx instance was used by micrometer metrics instance to create an HTTP server but in practice it can start its own embedded instance as there is no much gain to create a new one). One issue providing the Vertx instance is that the provided instance was a racy publication of the Vertx instance complicated to solve (because the MetricsFactory needs to be called early for creating the pool metrics). The metrics event bus initialised callback is also not used in practice (it used to be by hawkular implementation), only by micrometer metrics as a signal to initialise the backend which can be replaced by a simple initialisation in the micrometer metrics factory.
Change : do not provide measured instances when calling the SPI for creating a metrics implementation. Callback methods have also been renamed to contain the type of the metrics created (i.e createMetrics -> createEventBusMetrics, etc...). The event bus initialised callback is removed as well.
The text was updated successfully, but these errors were encountered:
Motivation : currently the Metrics SPI expose the measured instances to the SPI implementation when at creation of the corresponding metrics instance. This instance is actually not really used by the implementations (Vertx instance was used by micrometer metrics instance to create an HTTP server but in practice it can start its own embedded instance as there is no much gain to create a new one). One issue providing the Vertx instance is that the provided instance was a racy publication of the Vertx instance complicated to solve (because the MetricsFactory needs to be called early for creating the pool metrics). The metrics event bus initialised callback is also not used in practice (it used to be by hawkular implementation), only by micrometer metrics as a signal to initialise the backend which can be replaced by a simple initialisation in the micrometer metrics factory.
Change : do not provide measured instances when calling the SPI for creating a metrics implementation. Callback methods have also been renamed to contain the type of the metrics created (i.e
createMetrics
->createEventBusMetrics
, etc...). The event bus initialised callback is removed as well.The text was updated successfully, but these errors were encountered: