-
Notifications
You must be signed in to change notification settings - Fork 40.8k
Spring Boot 2.0 Release Notes
Since this is a major release of Spring Boot, upgrading existing applications can be a little more involved that usual. We’ve put together a dedicated migration guide to help you upgrade your existing Spring Boot 1.5 applications.
If you’re currently running with an earlier version of Spring Boot, we strongly recommend that you upgrade to Spring Boot 1.5 before migrating to Spring Boot 2.0.
Tip
|
Check the configuration changelog for a complete overview of the changes in configuration. |
Spring Boot 2.0 requires Java 8 as a minimum version.
Many existing APIs have been updated to take advantage of Java 8 features such as: default methods on interfaces, functional callbacks, and new APIs such as javax.time
.
If you are currently using Java 7 or earlier, you’ll need to upgrade your JDK before you can develop Spring Boot 2.0 applications.
Spring Boot 2.0 also works well, and has been tested with JDK 9. All our jars ship with automatic module name entries in the manifests for module system compatibility.
Spring Boot 2.0 builds on and requires Spring Framework 5. You might like to read about the new features available in Spring Framework 5.0, and check out their upgrade guide before continuing.
We’ve upgraded to the latest stable releases of other third-party jars wherever possible. Some notable dependency upgrades in this release include:
-
Tomcat 8.5
-
Flyway 5
-
Hibernate 5.2
-
Thymeleaf 3
Many projects in the Spring portfolio are now offering first-class support for developing reactive applications. Reactive applications are fully asynchronous and non-blocking. They’re intended for use in an event-loop execution model (instead of the more traditional one thread-per-request execution model). The "Web on Reactive Stack" section of the Spring Framework reference documentation provides an excellent primer to the subject.
Spring Boot 2.0 fully supports reactive applications via auto-configuration and starter-POMs. The internals of Spring Boot itself have also been updated where necessary to offer reactive alernatives (the most noticeable being our embedded server support).
Spring WebFlux is a fully non-blocking reactive alternative to Spring MVC. Spring Boot provides auto-configuration for both annotation based Spring WebFlux applications, as well as WebFlux.fn which offers a more functional style API.
To get started, use the spring-boot-starter-webflux
starter POM which will provide Spring WebFlux backed by an embedded Netty server.
See the Spring Boot reference documentation for details.
Where the underlying technology enables it, Spring Data also provides support for reactive applications. Currently Cassandra, MongoDB, Couchbase and Redis all have reactive API support.
Spring Boot includes special starter-POMs for these technologies that provide everything you need to get started.
For example, spring-boot-starter-data-mongodb-reactive
includes dependencies to the reactive mongo driver and project reactor.
Spring Boot 2.0 can make use of Spring Security 5.0 to secure your reactive applications. Auto-configuration is provided for WebFlux applications whenever Spring Security is on the classpath.
Access rules for Spring Security with WebFlux can be configured via a SecurityWebFilterChain
.
If you’ve used Spring Security with Spring MVC before, this should feel quite familiar.
See the Spring Boot reference documentation and Spring Security documentation for more details.
Since WebFlux does not rely on Servlet APIs, we’re now able to offer support for Netty as an embedded server for the first time.
The spring-boot-starter-webflux
starter POM will pull-in Netty 4.1 and Ractor Netty.
Note
|
You can only use Netty as a reactive server. Blocking servlet API support is not provided. |
HTTP/2 support is provided for Tomcat, Undertow and Jetty. Support depends on the chosen web server and the application environment (since the protocol is not supported out-of-the-box by JDK 8).
See this "how to" section for details.
The mechanism use to bind Environment
properties to @ConfigurationProperties
has been completely overhauled in Spring Boot 2.0.
We’ve taken the opportunity to tighten the rules that govern relaxed binding and we’ve fixed many inconsistencies from Spring Boot 1.x.
The new Binder
API can also be used outside of @ConfigurationProperties
directly in your own code.
For example, the following will bind to a List
of PersonName
objects:
List<PersonName> people = Binder.get(environment)
.bind("my.property", Bindable.listOf(PersonName.class))
.orElseThrow(IllegalStateException::new);
The configuration source could be represented in YAML like this:
my:
property:
- first-name: Jane
last-name: Doe
- first-name: John
last-name: Doe
For more information on the updated binding rules see this wiki page.
YAML files and Properties files loaded by Spring Boot now include Origin
information which can help you track where an item was loaded from.
Several Spring Boot features take advantage of this information and show it when appropriate.
For example, the BindException
class thrown when binding fails is an OriginProvider
.
This means origin information can be displayed nicely from a failure analyzer.
Another example is the env
actuator endpoint which includes origin information when it’s available.
The snippet below shows that the spring.security.user.name
property came from line 1
, column 27
of the application.properties
file packaged in the jar:
{
"name": "applicationConfig: [classpath:/application.properties]",
"properties": {
"spring.security.user.name": {
"value": "user",
"origin": "class path resource [application.properties]:1:27"
}
}
}
Binding makes use of a new ApplicationConversionService
class which offers some additional converters which are especially useful for property binding.
Most noticeable are converters for Duration
types and delimited strings.
The Duration
converter allows durations to be specified in either ISO-8601 form, or as a simple string (for example 10m
for 10 minutes with support of other units).
Existing properties have been changed to always use Duration
. For instance, the session timeout can be configured to 180 seconds in application.properties
as follows:
server.servlet.session.timeout=180s
The @DurationUnit
annotation ensures back-compatibility by setting the unit that is used if not is specified.
For example, a property that expected seconds in Spring Boot 1.5 now has @DurationUnit(ChronoUnit.SECONDS)
to ensure a simple value such as 10
actually uses 10s
.
Delimited string conversion allows you to bind a simple String
to a Collection
or Array
without necessarily splitting on commas.
For example, LDAP base-dn
properties use @Delimiter(Delimiter.NONE)
so that LDAP DNs (which typically include commas) are not misinterpreted.
Spring Boot’s Gradle plugin has been largely rewritten to enable a number of significant improvements. You can read more about the plugin’s capabilities in its reference and api documentation.
Spring Boot now requires Gradle 4.x. Please check the migration guide if you’re upgrading a project that uses Gradle.
Spring Boot 2.0 now includes support for Kotlin 1.2.x and offers a runApplication
function which provides a way to run a Spring Boot application using idiomatic Kotlin.
We also expose and leverage the Kotlin support that other Spring projects such as Spring Framework, Spring Data, and Reactor have added to their recent releases.
For more information, refer to the Kotlin support section of the reference documentation.
There have been many improvements and refinements to the actuator endpoints with Spring Boot 2.0.
All HTTP actuator endpoints are now exposed under the /actuator
path and resulting JSON payloads have been improved.
We now also don’t expose so many endpoints by default.
If you’re upgrading an existing Spring Boot 1.5 application, be sure to check the migration guide and pay particular attention to the management.endpoints.web.exposure.include
property.
The JSON payloads returned from many endpoints have been improved with Spring Boot 2.0.
Many endpoints now have JSON that more accurately reflects the underlying data.
For example, the /actuator/conditions
endpoint (/autoconfig
in Spring Boot 1.5) now has a top level contexts
key to group results by ApplicationContext
.
Extensive REST API documentation is now also generated using Spring REST Docs and published with each release.
In addition to Spring MVC and JMX support, you can now access actuator endpoints when developing pure Jersey or WebFlux applications.
Jersey support is provided via a custom Jersey Resource
and WebFlux uses a custom HandlerMapping
.
The /actuator
endpoint now provides a HAL formatted response providing links to all active endpoints (even if you don’t have Spring HATEOAS on your classpath).
In order to support Spring MVC, JMX, WebFlux and Jersey, we’ve developed a new programming model for actuator endpoints.
The @Endpoint
annotation can be used in combination with @ReadOperation
, @WriteOperation
and @DeleteOperation
to develop endpoints in a technology agnostic way.
You can also use @EndpointWebExtension
or @EndpointJmxExtension
to write technology specific enhancements to endpoints.
See the updated reference documentation for details.
Spring Boot 2.0 no longer ships with its own metrics APIs. Instead we rely on micrometer.io for all application monitoring needs.
Micrometer includes support for dimensional metrics which, when paired with a dimensional monitoring system, allows for efficient access to a particular named metric with the ability to drill down across its dimensions.
Metrics can be exported to a wide range of systems and out-of-the box Spring Boot 2.0 provides support for Atlas, Datadog, Ganglia, Graphite, Influx, JMX, New Relic, Prometheus, SignalFx, StatsD and Wavefront. In additional Simple in-memory metrics can also be used.
Integration is provided with JVM metrics (including CPU, memory, threads and GC), Logback, Tomcat, Spring MVC & RestTemplate
.
See the updated "Metrics" section of the reference documentation for more details.
In addition the "Reactive Spring Data" support mentioned above, several other updates and improvements have been made in the area of Data.
The default database pooling technology in Spring Boot 2.0 has been switched from Tomcat Pool to HikariCP. We’ve found that Hakari offers superior performance, and many of our users prefer it over Tomcat Pool.
Database initialization logic has been rationalized in Spring Boot 2.0.
Initialization for Spring Batch, Spring Integration, Spring Session and Quartz now occurs by default only when using and embedded database.
The enabled
property has been replaced with a more expressive enum.
For example, if you want to always perform Spring Batch initialization you can set spring.batch.initialize-schema=always
.
If Flyway or Liquibase is managing the schema of your DataSource and you’re using an embedded database, Spring Boot will now automatically switch off Hibernate’s automatic DDL feature.
Spring Boot 2.0 now detects the jOOQ dialect automatically based on the DataSource (similarly to what is done for the JPA dialect).
A new @JooqTest
annotation has also been introduced to ease testing where only jOOQ has to be used.
The JdbcTemplate
that Spring Boot auto-configures can now be customized via spring.jdbc.template
properties.
Also, the NamedParameterJdbcTemplate
that is auto-configured reuses the JdbcTemplate
behind the scenes.
Spring Boot exposes a new spring.data.web
configuration namespace that allows to easily configure paging and sorting.
Spring Boot now auto-configures the open-source time series database InfluxDB.
To enable InfluxDB support you need to set a spring.influx.url
property, and include influxdb-java
on your classpath.
If only a custom url
or user
property is provided, the auto-configuration for Flyway and Liquibase now reuses the standard datasource properties rather than ignoring them.
This allows you to create a custom DataSource
for the purpose of the migration with only the required information.
Support is now offered for custom Hibernate naming strategies.
For advanced scenarios, you can now define ImplicitNamingStrategy
or PhysicalNamingStrategy
to use as regular beans in the context.
It is now also possible to customize the properties Hibernate uses in a more fine-grained way by exposing a HibernatePropertiesCustomizer
bean.
When using the reactive driver, it is now possible to apply advanced customizations to the Mongo client that Spring Boot auto-configures by defining a bean of type MongoClientSettingsBuilderCustomizer
.
In addition to the WebFlux and WebFlux.fn support mentioned above, the following refinements have also been made to help when developing web applications.
When using an embedded container, the context path is logged alongside the HTTP port when your application starts. For a example, embedded Tomcat now looks something like this:
Tomcat started on port(s): 8080 (http) with context path '/foo'
The Thymeleaf starter now includes thymeleaf-extras-java8time
which provides support for javax.time
types.
A new spring-boot-starter-json
starter gathers the necessary bits to read and write JSON.
It provides not only jackson-databind
, but also useful modules when working with Java8: jackson-datatype-jdk8
, jackson-datatype-jsr310
and jackson-module-parameter-names
.
This new starter is now used where jackson-databind
was previously defined.
If you prefer something other than Jackson, our support for GSON has been greatly improved in Spring Boot 2.0. We’ve also introduced support for JSON-B (including JSON-B testing support).
Auto-configuration support is now include for the Quartz Scheduler.
We’ve also added a new spring-boot-starter-quartz
starter POM.
You can use in-memory JobStores
, or a full JDBC-based store.
All JobDetail
, Calendar
and Trigger
beans from your Spring application context will be automatically registered with the Scheduler
.
For more details read the new "Quartz Scheduler" section of the reference documentation.
There have been a few additions and tweaks to the testing support provided in Spring Boot 2.0:
-
A new
@WebFluxTest
annotation has been added to support “slice” testing of WebFlux applications. -
Converter
andGenericConverter
beans are now automatically scanned with@WebMvcTest
and@WebFluxTest
. -
An
@AutoConfigureWebTestClient
annotation had been added to provide aWebTestClient
bean for tests to use. The annotation is automatically applied to@WebFluxTest
tests. -
A new
ApplicationContextRunner
test utility has been added which makes it very easy to test your auto-configurations. We’ve moved most of our internal test suite to this new model. See the updated documentation for details.
As well as the changes listed above, there have also been lots of minor tweaks and improvements including:
-
@ConditionalOnBean
now uses a logicalAND
rather than a logicalOR
when determining whether or not the condition has been met. -
Unconditional classes are now included in the auto-configuration report.
-
The
spring
CLI application now includes anencodepassword
command which can be used to create Spring Security compatible hashed passwords. -
Scheduled tasks (i.e.
@EnableScheduling
) can be be reviewed using thescheduledtasks
actuator endpoint. -
The
loggers
actuator endpoint now allows you to reset a logger level to its default. -
Spring Session users can now find and delete sessions via the
sessions
actuator endpoint. -
Maven-based applications using
spring-boot-starter-parent
now use the-parameters
flag by default. -
Our build now uses concourse for CI and our project POM files have been restructured so that they are simpler (these changes should be transparent to most users, but if you find any issues with the published POMs please report them).
Finally, and just for fun, Spring Boot 2.0 now supports animated GIF banners. See this project for an example.