Configures this server or client with given exception stats handler.
Configures this server or client to shift user-defined computation (com.twitter.util.Future callbacks and transformations) off of IO threads into a given ExecutorService.
Configures this server or client to shift user-defined computation (com.twitter.util.Future callbacks and transformations) off of IO threads into a given ExecutorService.
By default, Finagle executes all futures in the IO threads, minimizing context switches. Given there is usually a fixed number of IO threads shared across a JVM process, it's critically important to ensure they aren't being blocked by the application code, affecting system's responsiveness. Shifting application-level work onto a dedicated FuturePool or ExecutorService offloads IO threads, which may improve throughput in CPU-bound systems.
As always, run your own tests before enabling this feature.
Configures this server or client to shift user-defined computation (com.twitter.util.Future callbacks and transformations) off of IO threads into a given FuturePool.
Configures this server or client to shift user-defined computation (com.twitter.util.Future callbacks and transformations) off of IO threads into a given FuturePool.
By default, Finagle executes all futures in the IO threads, minimizing context switches. Given there is usually a fixed number of IO threads shared across a JVM process, it's critically important to ensure they aren't being blocked by the application code, affecting system's responsiveness. Shifting application-level work onto a dedicated FuturePool or ExecutorService offloads IO threads, which may improve throughput in CPU-bound systems.
As always, run your own tests before enabling this feature.
Configures this server or client with given label
(default: empty string).
Configures this server or client with given label
(default: empty string).
The label
value is used for stats reporting to scope stats reported
from different clients/servers to a single stats receiver.
Configures this server or client with given util.Monitor (default: com.twitter.finagle.util.DefaultMonitor).
Configures this server or client with given util.Monitor (default: com.twitter.finagle.util.DefaultMonitor).
Monitors are Finagle's out-of-band exception reporters. Whenever an exception is thrown on a request path, it's reported to the monitor. The default monitor implementation, com.twitter.finagle.util.DefaultMonitor, logs these exceptions.
Monitors are wired into the server or client stacks via com.twitter.finagle.filter.MonitorFilter and are applied to the following kinds of exceptions:
Service.apply(request)
Service.apply(request)
respond
, onSuccess
, onFailure
future callbacksmap
, flatMap
, transform
future continuationsPut it this way, we apply Monitor.handle
to an exception if we would otherwise "lose" it,
i.e. when it's not connected to the Future
, nor is it connected to the call stack.
You can compose multiple monitors if you want to extend or override the standard behavior,
defined in DefaultMonitor
.
import com.twitter.util.Monitor val consoleMonitor = new Monitor { def handle(exc: Throwable): Boolean = { Console.err.println(exc.toString) false // continue handling with the next monitor (usually DefaultMonitor) } } $.withMonitor(consoleMonitor)
Returning true
form within a monitor effectively terminates the monitor chain so no
exceptions are propagated down to the next monitor.
Configures the Tunable request timeout
of this server or client (if applying the
Tunable produces a value of None
, an unbounded timeout is used for the request).
Configures the Tunable request timeout
of this server or client (if applying the
Tunable produces a value of None
, an unbounded timeout is used for the request).
If the request has not completed within the Duration resulting from timeout.apply()
,
the pending work will be interrupted via com.twitter.util.Future.raise.
The client request timeout is the maximum amount of time given to a single request (if there are retries, they each get a fresh request timeout). The timeout is applied only after a connection has been acquired. That is: it is applied to the interval between the dispatch of the request and the receipt of the response.
The server request timeout is the maximum amount of time, a server is allowed to spend handling the incoming request. Using the Finagle terminology, this is an amount of time after which a non-satisfied future returned from the user-defined service times out.
Configures the request timeout
of this server or client (default: unbounded).
Configures the request timeout
of this server or client (default: unbounded).
If the request has not completed within the given timeout
, the pending
work will be interrupted via com.twitter.util.Future.raise.
The client request timeout is the maximum amount of time given to a single request (if there are retries, they each get a fresh request timeout). The timeout is applied only after a connection has been acquired. That is: it is applied to the interval between the dispatch of the request and the receipt of the response.
The server request timeout is the maximum amount of time, a server is allowed to spend handling the incoming request. Using the Finagle terminology, this is an amount of time after which a non-satisfied future returned from the user-defined service times out.
Configure a com.twitter.finagle.service.ResponseClassifier which is used to determine the result of a request/response.
Configure a com.twitter.finagle.service.ResponseClassifier which is used to determine the result of a request/response.
This allows developers to give Finagle the additional application-specific knowledge necessary in order to properly classify responses. Without this, Finagle cannot make judgements about application-level failures as it only has a narrow understanding of failures (for example: transport level, timeouts, and nacks).
As an example take an HTTP server that returns a response with a 500 status code. To Finagle this is a successful request/response. However, the application developer may want to treat all 500 status codes as failures and can do so via setting a com.twitter.finagle.service.ResponseClassifier.
ResponseClassifier is a PartialFunction and as such multiple classifiers can be composed together via PartialFunction.orElse.
Response classification is independently configured on the client and server.
For client-side response classification using com.twitter.finagle.builder.ClientBuilder,
see com.twitter.finagle.builder.ClientBuilder.responseClassifier
If unspecified, the default classifier is com.twitter.finagle.service.ResponseClassifier.Default which is a total function fully covering the input domain.
com.twitter.finagle.http.service.HttpResponseClassifier
for some
HTTP classification tools.
Configures this server or client with given stats.StatsReceiver (default: stats.DefaultStatsReceiver).
Configures this server or client with given tracing.Tracer (default: com.twitter.finagle.tracing.DefaultTracer).
Configures this server or client with given tracing.Tracer (default: com.twitter.finagle.tracing.DefaultTracer).
if you supply com.twitter.finagle.tracing.NullTracer, no trace information will be written, but this does not disable Finagle from propagating trace information. Instead, if traces are being aggregated across your fleet, it will orphan subsequent spans.
A collection of methods for configuring common parameters (labels, stats receivers, etc) shared between Finagle clients and servers.
a Stack.Parameterized server/client to configure