The driver exposes measurements of its internal behavior through the popular Dropwizard Metrics library. Developers can access these metrics and choose to export them to a monitoring tool.
The driver depends on Metrics 3.2.x, but is compatible with newer versions of Dropwizard Metrics. For using Metrics 4.x with the driver, see Metrics 4 Compatibility.
Metric names are path-like, dot-separated strings. Metrics are measured at the Cluster
-level,
thus the driver prefixes them with the name of the Cluster
they are associated with (see withClusterName
for how to configure this), suffixed by -metrics
. For example:
cluster1-metrics.connected-to
cluster1-metrics.connection-errors
...
By default, metrics are enabled and exposed via JMX as MXBeans.
Some users may find that they don't want the driver to record and expose metrics. To disable metrics collection, use the withoutMetrics builder method, i.e.:
Cluster cluster = Cluster.builder()
.withoutMetrics()
.build();
Note that if you decide to disable metrics, you may also consider excluding metrics as a dependency. To do this in a maven project:
<dependency>
<groupId>com.scylladb</groupId>
<artifactId>scylla-driver-core</artifactId>
<version>3.11.2.0</version>
<exclusions>
<exclusion>
<groupId>io.dropwizard.metrics</groupId>
<artifactId>metrics-core</artifactId>
</exclusion>
</exclusions>
</dependency>
Alternatively, one may not want to expose metrics using JMX. Disabling JMX reporting is simple as using the withoutJMXReporting builder method, i.e.:
Cluster cluster = Cluster.builder()
.withoutJMXReporting()
.build();
Cluster
metrics may be accessed via the getMetrics method. The Metrics class offers
direct access to all metrics recorded for the Cluster
via getter methods. Refer to
the Metrics javadoc for more details about the metrics offered.
It is very common for applications to record their own metrics. You can add all metrics
recorded for a Cluster
to your applications' MetricRegistry in the following manner:
MetricRegistry myRegistery = new MetricRegistry();
myRegistry.registerAll(cluster.getMetrics().getRegistry());
Dropwizard Metrics offers a variety of Reporters for exporting metrics. To enable reporting,
access the Cluster
's metrics via the getMetrics method. For example, to enable CSV reporting
every 30 seconds:
import com.codahale.metrics.*;
import java.io.File;
import java.util.concurrent.TimeUnit;
Metrics metrics = cluster.getMetrics();
CsvReporter csvReporter = CsvReporter.forRegistry(metrics.getRegistry())
.convertDurationsTo(TimeUnit.MILLISECONDS)
.convertRatesTo(TimeUnit.SECONDS)
.build(new File("."));
csvReporter.start(30, TimeUnit.SECONDS);
While the driver depends on Metrics 3.2.x, it also works with Metrics 4, with some caveats.
In Metrics 4, JMX reporting was moved to a separate module, metrics-jmx
. Because of this you are
likely to encounter the following exception at runtime when initializing a Cluster
:
Exception in thread "main" java.lang.NoClassDefFoundError: com/codahale/metrics/JmxReporter
at com.datastax.driver.core.Metrics.<init>(Metrics.java:103)
at com.datastax.driver.core.Cluster$Manager.init(Cluster.java:1402)
at com.datastax.driver.core.Cluster.init(Cluster.java:159)
at com.datastax.driver.core.Cluster.connectAsync(Cluster.java:330)
at com.datastax.driver.core.Cluster.connectAsync(Cluster.java:305)
at com.datastax.durationtest.core.DurationTest.createSessions(DurationTest.java:360)
....
Caused by: java.lang.ClassNotFoundException: com.codahale.metrics.JmxReporter
at java.base/jdk.internal.loader.BuiltinClassLoader.loadClass(BuiltinClassLoader.java:582)
at java.base/jdk.internal.loader.ClassLoaders$AppClassLoader.loadClass(ClassLoaders.java:185)
at java.base/java.lang.ClassLoader.loadClass(ClassLoader.java:496)
... 8 more
To fix this, use withoutJMXReporting when constructing your Cluster
. If you still desire JMX
reporting, add metrics-jmx
as a dependency:
<dependency>
<groupId>io.dropwizard.metrics</groupId>
<artifactId>metrics-jmx</artifactId>
<version>4.0.2</version>
</dependency>
Then create your Cluster
and JmxReporter
in the following manner:
Cluster cluster = Cluster.builder()
.withoutJMXReporting()
.build();
JmxReporter reporter =
JmxReporter.forRegistry(cluster.getMetrics().getRegistry())
.inDomain(cluster.getClusterName() + "-metrics")
.build();
reporter.start();