Skip to content

Latest commit

 

History

History
59 lines (40 loc) · 3.59 KB

readme.md

File metadata and controls

59 lines (40 loc) · 3.59 KB
Raw

molten-core

Provides the core functionalities for Molten.

Reactor initialization

In order to have proper logging and context propagation the following should be invoked before any other Reactor calls in your application.

com.hotels.molten.core.MoltenCore.initialize();

It sets Slf4j for logging and enables internal Reactor metrics if enabled via MoltenCore.setMetricsEnabled(boolean).

Context propagation

Context propagation in reactive flows is not always trivial. To be able to pass around implicit contexts (which are usually handled via ThreadLocal) extra steps should be taken.

There are mainly two scenarios where context propagation should happen:

  • tasks submitted to async schedulers
  • returning to reactive flows via callbacks from non-reactor managed threads

Schedulers

To propagate contexts for scheduled tasks, one should register a onSchedule hook via reactor.core.scheduler.Schedulers.onScheduleHook. See com.hotels.molten.core.mdc.MoltenMDC.MDCCopyingAction for an example hook.

Callbacks

Publishers with custom threading solutions often calls back to reactive flows from a thread which is not managed by Reactor. These kind of scenarios are also found in non-blocking libraries (e.g. netty).

While Reactor provides a way to transform flows at each operator using reactor.core.publisher.Hooks.onEachOperator(String, Function) it is usually unnecessary everywhere and can also be expensive. Molten provides a generic way to register context propagators for flows with such callbacks, only invoked where needed.

Calling com.hotels.molten.core.MoltenCore.registerContextPropagator(String, Function) with a transformer function will add it to registered context propagators. Wherever explicit context propagation is needed, com.hotels.molten.core.MoltenCore.propagateContext() should be used as a transformer. e.g. aMono.transform(MoltenCore.propagateContext()) This will apply all registered context propagators.

To create a custom context propagator you can use com.hotels.molten.core.common.AbstractNonFusingSubscription as a base class. See com.hotels.molten.core.mdc.MoltenMDC.propagate() for an example.

MDC

Using org.slf4j.MDC to propagate diagnostic information (logging, request correlation, tracing, etc.) is a common solution. Support for MDC is provided via com.hotels.molten.core.mdc.MoltenMDC. Call com.hotels.molten.core.mdc.MoltenMDC.initialize() to set up required hooks and context propagators for MDC. This should be done as soon as possible, before any Reactor calls.

Metrics

Molten supports both dimensional and hierarchical metrics.
This can be controlled via com.hotels.molten.core.metrics.MoltenMetrics. To enable dimensional metrics one should invoke MoltenMetrics.setDimensionalMetricsEnabled(true) or set system property MOLTEN_METRICS_DIMENSIONAL_METRICS_ENABLED=true. When dimensional metric names are enabled, the hierarchial name can also be added as a label with MoltenMetrics.setGraphiteIdMetricsLabelEnabled(true) or setting system property MOLTEN_METRICS_GRAPHITE_ID_LABEL_ADDED=true.

Request collapsing

There are generic solutions to make more effective service invocations with request collapsing:

A complete integration pattern can be found here