Spring Boot 2.2.0 Release Notes
Classes, methods and properties that were deprecated in Spring Boot 2.1 have been removed in this release. Please ensure that you aren’t calling deprecated methods before upgrading.
This release upgrades to Spring Framework 5.2. Please refer to Spring Framework’s upgrade documentation for further details.
As of #16090, JMX is not enabled by default anymore. This feature can be enabled with the configuration property spring.jmx.enabled=true.
If you’re using IDE features to administer your applications, you may want to enable that flag there as well.
Where possible, we have moved from Java EE dependencies with a javax. group ID to the equivalent Jakarta EE dependencies with a jakarta. group ID in Spring Boot’s starters.
Dependency management for the Jakarta EE API dependencies has been added alongside the existing dependency management for the Java EE API dependencies.
The dependency management for the Java EE API dependencies will be removed in the future and all users are encouraged to move to the Jakarta EE API dependencies.
As part of the Java EE to Jakarta EE migration, two dependencies have changed artifact ID in their latest maintenance releases.
com.sun.mail:javax.mail is now com.sun.mail:jakarta.mail and org.glassfish:javax.el is now org.glassfish:jakarta.el.
In the unlikely event that you were using either of these dependencies directly, please update your pom.xml or build.gradle accordingly.
spring-boot-starter-test now provides JUnit 5 by default.
JUnit 5’s vintage engine is included by default to support existing JUnit 4-based test classes so that you can migrate to JUnit 5 when you are ready to do so.
It is also possible to use a mixture of JUnit 4- and JUnit 5-based test classes in the same module.
This allows you to migrate to JUnit 5 gradually if you wish.
This release upgrades to AssertJ 3.12 that contains a breaking API change for assertions related to Iterator.
Please see the AssertJ release notes for details.
This release upgrades to Spring HATEOAS 1.0 which contains some breaking API changes. Please see the blog post that announced the M1 release for details and a script that can be used to ease the migration.
logging.file.max-size and logging.file.total-size-cap are now using the standard DataSize type in replacement of the proprietary FileSize type.
If you are using any of those properties, make sure to align to the standard format.
The DataSource health indicator contains an extra validationQuery attribute with the validation query that was used against the DataSource.
Also, the result of the query that used to be provided by the hello attribute has been renamed to result.
This release upgrades to Elasticsearch 6.7 which contains some breaking API changes. Please see the 6.5, 6.6 and 6.7 breaking changes pages for details.
Both the Elasticsearch transport and Jest clients have been deprecated in favor of other options, such as RestHighLevelClient.
The reference docs on Elasticsearch support have been updated.
Spring Boot now lets Hibernate chose the dialect to use rather than applying a default dialect based on the detected database. If you had a dialect configured previously, you may want to remove your customization.
The Actuator HTTP Trace and Auditing features are not enabled by default anymore, since the default repositories implementations are in-memory and may consume too many resources and are not cluster friendly For robust, production-grade HTTP tracing we recommend the use of Spring Cloud Sleuth or similar.
To enable HTTP tracing, contribute a bean implementing HttpTraceRepository or AuditEventRepository to turn those features back on.
Configuration properties are available to turn those features off even in the presence of those beans (see management.auditevents.enabled and management.trace.http.enabled)
The minimum requirements for Gradle have changed - Spring Boot now requires Gradle 4.10+. Please upgrade accordingly.
server.jetty.accesslog.* configuration properties related to log formatting are no longer available (the ones related to log files are still there).
If your application is making use of those properties, two new keys have been introduced, format and custom-format; format allows you to choose amongst existing formats, custom-format allows you to use your own format:
server.jetty.accesslog.format=NCSA # enum with predefined formats: NCSA, EXTENDED_NCSA
server.jetty.accesslog.custom-format= # String, custom format overriding the former if definedUnsupported options now translate to defining a custom log format and using options as defined in Jetty’s CustomRequestLog.
This release upgrades to Hamcrest 2.1. This should be a drop-in replacement for most users.
If you have a direct dependency on org.hamcrest:hamcrest-core or org.hamcrest:hamcrest-library, please use the new combined module, org.hamcrest:hamcrest, instead.
If you depend on Hamcrest via spring-boot-starter-test no action is required.
This release changes the default template file extension for Freemarker templates.
This aligns Spring Boot with safe configuration defaults for web applications.
Please rename existing templates from *.ftl to *.ftlh when upgrading.
Tomcat’s MBean Registry is now disabled by default, saving approximately 2MB of heap.
If you want to use Tomcat’s MBeans, for example so that they can be used to expose metrics via Micrometer, set the server.tomcat.mbeanregistry.enabled to true.
The <rollingPolicy> in the default Logback configuration has been updated to use a <maxHistory> of 7 days.
It can be tuned using the logging.file.max-history configuration property.
HttpHiddenMethodFilter disabled by default
The filter that handles the _method request parameter is now disabled by default as it causes early consumption of a request body if the body may contain parameters.
This can be restored by setting either spring.webflux.hiddenmethod.filter.enabled or spring.mvc.hiddenmethod.filter.enabled to true.
A number of classes have been deprecated to implement the new health indicator groups feature.
If you are configuring a custom HealthIndicator via CompositeHealthIndicatorConfiguration, you’ll need to update it to use CompositeHealthContributorConfiguration instead.
Micrometer 1.3 has changed the eventType that gets published with each metric to be a fixed value. This aligns with New Relic best practices.
Spring Boot 2.2 will use the value of management.metrics.export.newrelic.event-type as the eventType and add "metricName" and "metricType" attributes for context.
If you don’t explicitly set an event type property then SpringBootSample is used.
If you prefer to use the previous behavor where the meter name was used as the eventType, you can set the management.metrics.export.newrelic.meter-name-event-type-enabled property to true.
The /actuator/health endpoint has changed the resulting JSON format by renaming details to components for the first-level elements.
This helps to differentiate the actual details returned by a HealthIndicator from the component indicators that make up composite health.
As a result of the change, the actuator media type has been bumped from application/vnd.spring-boot.actuator.v2+json to application/vnd.spring-boot.actuator.v3+json.
If you have tools that need to consume the older format, you can use an HTTP Accept: header with the V2 media type, application/vnd.spring-boot.actuator.v2+json.
The preferred location for global DevTools settings is now ~/.config/spring-boot.
Any of the following files can be used:
-
spring-boot-devtools.properties -
spring-boot-devtools.yaml -
spring-boot-devtools.yml
Please see the relevant section of the reference documentation for further details.
The sample projects have been renamed to smoke tests to better reflect their intended purpose. They have also been relocated in the source repository and can now be found here.
|
Tip
|
Check the configuration changelog for a complete overview of the changes in configuration. |
Startup time and memory usage have been reduced by making use of proxyBeanMethods=false in Spring Boot’s @Configuration classes.
proxyBeanMethods is a new attribute on @Configuration introduced in Spring Framework 5.2 M1.
proxyBeanMethods is also available as an attribute on @SpringBootApplication and @SpringBootConfiguration.
When launching an application at development time with bootRun in Gradle or spring-boot:run in Maven, the JVM will be configured with flags (-Xverify:none and -XX:TieredStopAtLevel=1) to optimise it for reduced launch time.
When running on JDK 13 -Xverify:none is not specified as it has been deprecated.
Several other performance improvements have also been made in this release:
-
Time taken to bind large numbers of configuration properties has been significantly reduced
-
As Spring Boot fully prepares a
PersistenceUnitby scanning JPA entities, Hibernate’s own entity scanning has been disabled as it is redundant -
Injection points in auto-configurations have been refined to only apply when a bean has to be created
-
Beans related to Actuator endpoints are now only created if the endpoint is both enabled and exposed (via JMX or HTTP)
-
Conditions on codec auto-configuration have been improved so that the codecs are no longer configured when they won’t be used
-
Tomcat’s MBean Registry has been disabled by default, reducing Tomcat’s memory footprint by approximately 2MB
It is now possible to enable global lazy initialization to reduce startup time via the spring.main.lazy-initialization property.
Please note that using that feature comes at a cost:
-
Handling of HTTP requests may take longer while any deferred initialisation occurs
-
Failures that would normally occur at startup will now not occur until later
Individual beans can opt out of lazy initialization by annotating their definition with @Lazy(false).
Where it is not possible to use @Lazy(false) to opt out of lazy initialization, a LazyInitializationExcludeFilter bean can be used instead.
For example, to never set IntegrationFlow beans to lazy, you can use the following code:
@Bean
static LazyInitializationExcludeFilter integrationLazyInitExcludeFilter() {
return LazyInitializationExcludeFilter.forBeanTypes(IntegrationFlow.class);
}Spring Boot 2.2 ships with Spring Data Moore. Please see the What’s new in Spring Data Moore blog post to learn more.
The behaviour on shutdown for both the auto-configured TaskExecutor and TaskScheduler is now configurable, see the spring.task.execution.shutdown and spring.task.scheduling.shutdown namespaces for more details.
ConditionalOnCloudPlatform now detects if the application is running on Kubernetes.
SpringBootTest allows to specify application arguments that will trigger the creation of an ApplicationArguments bean.
Classes annotated with @ConfigurationProperties can now be found via classpath scanning as an alternative to using @EnableConfigurationProperties or @Component.
If you use @SpringBootApplication, scanning is enabled by default for the package that contains the @SpringBootApplication-annotated class.
Scanning can be enabled manually, and the scanned packages can be customized, using @ConfigurationPropertiesScan.
Configuration properties now support constructor-based binding, which allows a @ConfigurationProperties-annotated class to be immutable.
Constructor-based binding can be enabled by annotating a @ConfigurationProperties class or one of its constructors with @ConstructorBinding.
Annotations such as @DefaultValue and @DateTimeFormat can now be used on constructor parameters that are provided by configuration property binding.
Please see the relevant section of the reference documentation for further details.
A new Spring Boot starter has been added in this release, spring-boot-starter-rsocket.
This starter brings the required dependencies for building an application that uses RSocket support.
RSocket strategies are auto-configured to provide the required infrastructure for encoding and decoding RSocket payloads using CBOR and JSON.
An RSocketRequester.Builder is auto-configured to allow you to create an RSocketRequester to send requests to remote RSocket services.
For more on that, check out the relevant section if the reference documentation.
Auto-configuration for Spring Security’s RSocket integration is now available when its spring-security-rsocket module is on the classpath.
The auto-configuration will enable RSocket security and configure the server RSocket factory with Spring Security’s interceptor.
When testing an RSocket server application, the local.rsocket.server.port property is set to the port that the RSocket server is listening on.
It can be injected into test classes using @LocalRSocketServerPort.
The ApplicationContextRunner test utility now allows to register bean inline, see withBean for more details.
Methods have been added to RestTemplateBuilder to add default headers to all requests and to enable general request customization.
Auto-configuration has been added for the reactive Elasticsearch components introduced in Spring Data Moore.
A ReactiveElasticSearchTemplate is auto-configured with the spring.data.elasticsearch.client.reactive.* properties.
Auto-configured support for reactive Elasticsearch repositories is also provided.
The threaddump actuator endpoint can now return a thread dump in plain text that is compatible with both Thread Dump Analyzer and https://fastthread.io.
Input parameters of an Actuator @Endpoint can be managed using custom @EndpointConverter-annotated converters.
The RedisCacheManager configuration has been improved recently and a RedisCacheManagerBuilderCustomizer bean allows you to get a handle to the builder before it is used to configure an immutable RedisCacheManager.
In an environment with multiple data sources, a DataSource bean can be qualified with @BatchDataSource to indicate it is the one to be used by Spring Batch.
The build info goal has an extra time property that allows to configure how build.time is handled.
It can be disabled completely or set to a fixed time to make the output of build.properties repeatable.
It is now possible to track the total size of a connection pool by tracking the size and idle metrics.
If you have a custom DataSourcePoolMetadata implementation, consider implementing the getIdle method.
It is now possible to organize health indicators into groups. A typical example if you deploy your application to Kubernetes, you may want one different sets of health indicators for your “liveness” and “readiness” probes.
Groups can be configured via configuration properties.
The following creates a custom group with only the DataSource indicator:
management.endpoint.health.group.custom.include=dbThe custom group can be invoked by hitting localhost:8080/actuator/health/custom.
Check the updated reference documentation for more details.
Component details returned from /actuator/health can now be configured to be shown independently of the details.
The management.endpoint.health.show-components property works in a similar way to show-details and can be set to never, when-authorized or always.
For example, if you want to always show the individal health indicators composed to create the system health, but you don’t want to show their details you can use:
management.endpoint.health.show-components=always
management.endpoint.health.show-details=neverThe same property can also be set on a health indicator group.
Flyway will be auto-configured to use any JavaMigration beans that are found in the application context.
The sanitization performed by the configprops and env endpoints has been improved to include URI properties.
Any password found in the URI’s user information will now be sanitized.
ASCII banner files can now make use of ANSI 256 color escape codes by using {AnsiColor.NNN} (where NNN is the color code).
You can also use them with image banners by setting the spring.banner.image.bitdepth property to 8.
We’ve also added a spring.banner.image.pixelmode property that you can set to block to use ASCII block chars.
The result looks like this:
Auto-configuration for Spring Security’s SAML 2.0 Relying Party is now available when its spring-security-saml2-server-provider module is on the classpath.
Relying parties can be registered using the spring.security.saml2.relyingparty.registration.* properties.
Spring Boot 2.2 moves to new versions of several Spring projects:
-
Reactor Dysprosium
-
Spring AMQP 2.2
-
Spring Batch 4.2
-
Spring Data Moore
-
Spring Framework 5.2
-
Spring HATEOAS 1.0
-
Spring Integration 5.2
-
Spring Kafka 2.3
-
Spring Security 5.2
-
Spring Session Corn
Numerous third-party dependencies have also been updated, some of the more noteworthy of which are the following:
-
Artemis 2.9
-
Elasticsearch 6.7
-
Flyway 6.0
-
Git Commit ID Plugin 3.0
-
Hazelcast 3.12
-
HSQLDB 2.5
-
Jackson 2.10
-
Jedis 3.1
-
Jersey 2.29
-
Kafka 2.2
-
Lettuce 5.2
-
Micrometer 1.3
-
Mockito 3.1
-
Solr 8.0
Apart from the changes listed above, there have also been lots of minor tweaks and improvements including:
-
loadOnStartupcan be configured if you’re wrapping aServletas an actuator endpoint. -
Flyway bootstrap failures are covered by a dedicated
FailureAnalyzer -
The
MessageConverterfor Kafka batch listeners is properly auto-configured. -
Add support for Jaybird 4 (
jdbc:firebirdprefix). -
Add opt-in support for Neo4j-OGM native types.
-
Client-side HTTP-metrics have an
outcometag. -
Numerous upgrades of third-party dependencies
-
Prometheus push gateway can now be configured using
https -
PrinterandParserbeans are automatically registered with the application conversion service. -
When using the H2 console auto-configuration, JDBC connection URL is now logged at startup for ease of connection.
-
Couchbase role-based access is now supported via
spring.couchbase.usernameandspring.couchbase.password. -
Support for OAuth2 resource server opaque token authentication has been added and can be configured using the
spring.security.oauth2.resourceserver.opaquetokenconfiguration properties. -
The application will fail fast if
spring.config.locationspecifies a file with an extension that noPropertySourceLoadercan read. -
The layout to use to repackage the application can be set on the command line with Maven using the
spring-boot.repackage.layoutproperty. -
RestTemplateBuilder#defaultHeaderaccepts more than one value. -
Custom resource handlers can serve a
favicon. -
Kafka’s
RecordInterceptoris detected and associated to the auto-configured listener container. -
YAML configuration can now use
onorofffor boolean types. -
Actuator discovery page is now available on
/when using a separate management port. -
@WebMvcTestnow scansHandlerInterceptorbeans. -
@WebFluxTestscansWebFilterbeans. -
Configurable timeout for JMS listener container’s receive.
-
Support for Spring Session’s flush mode.
-
Support for Spring AMQP’s
confirm-type. -
Application properties for configuring Jetty’s thread pool.
-
Support for parallel test execution when using @AutoConfigureMockMvc with JUnit 5.
-
Zip64 files are now supported inside "Far Jars".
-
The
ROLLING_FILE_LOG_PATTERNis now exposed for use in logback configurations. -
You can now inject and use a customized
WebTestClientin any@SpringBootTest. -
Dependency management for Oracle’s JDBC driver has been added.
-
Dependency management for Awaitility has been added.
-
The
logging.fileproperty has been renamed tologging.file.name. -
The
logging.pathproperty has been renamed tologging.file.path. -
ReactiveWebServerApplicationContext#getWebServerFactory. -
The
agentproperty of the Maven plugin has been renamed toagents. -
Joda time support is deprecated in favour of
java.time. -
ApplicationHealthIndicatorin favour ofPingHealthIndicatorthat is always contributed. -
ConfigurationBeanFactoryMetadatain favour ofConfigurationPropertiesBean. -
ConfigurationPropertiesBindingPostProcessorconstructors in favor of@EnableConfigurationPropertiesor theregistermethod. -
ConfigurationPropertiesBindingPostProcessor.VALIDATOR_BEAN_NAMEhas moved toEnableConfigurationProperties.VALIDATOR_BEAN_NAME. -
ConfigurationPropertiesBindingPostProcessorRegistrarin favor of@EnableConfigurationProperties. -
WebTestClientBuilderCustomizerhas been relocated toorg.springframework.boot.test.web.reactive.server.