-
Notifications
You must be signed in to change notification settings - Fork 40.8k
Spring Boot 2.6 Release Notes
Classes, methods and properties that were deprecated in Spring Boot 2.4 have been removed in this release. Please ensure that you aren’t calling deprecated methods before upgrading.
Circular references between beans are now prohibited by default.
If your application fails to start due to a BeanCurrentlyInCreationException
you are strongly encouraged to update your configuration to break the dependency cycle.
If you are unable to do so, circular references can be allowed again by setting spring.main.allow-circular-references
to true
, or using the new setter methods on SpringApplication
and SpringApplicationBuilder
This will restore 2.5’s behaviour and automatically attempt to break the dependency cycle.
The default strategy for matching request paths against registered Spring MVC handler mappings has changed from AntPathMatcher
to PathPatternParser
.
A custom dispatcher servlet path (spring.mvc.servlet.path
) is not supported when using PathPatternParser
.
If you are using Spring Security, you should review your use of mvcMatchers
to ensure that your matchers continue to meet your needs.
With AntPathMatcher
, authorizeRequests.mvcMatchers("hello").permitAll()
would grant access to /hello
.
The more precise matching of PathPatternParser
requires the use of authorizeRequests.mvcMatchers("/hello").permitAll()
(note the leading /
) instead.
If you need to switch the default back to AntPathMatcher
, you can set spring.mvc.pathmatch.matching-strategy
to ant-path-matcher
.
The actuator endpoints now also use PathPattern
based URL matching.
Note that the path matching strategy for actuator endpoints is not configurable via a configuration property.
If you are using Actuator and Springfox, this may result in your application failing to start. Please see this Springfox issue for further details.
The env
info contributor is now disabled by default.
This contributor exposes any property from the Spring Environment whose name starts with info
.
If you have any properties starting with info
and would like them to appear them in the InfoEndpoint
, set management.info.env.enabled
to true
.
The spring.boot.application.running
startup step logged to ApplicationStartup
has been renamed to spring.boot.application.ready
.
If you are processing files generated from FlightRecorderApplicationStartup
or BufferingApplicationStartup
you will need to use the new name.
Injecting Resources
directly no longer works as this configuration has been harmonized in WebProperties
. If you need to access this information, you need to inject WebProperties
instead.
Dependency management for org.jboss:jboss-transaction-spi
has been removed.
If you are using org.jboss:jboss-transaction-spi
you should define your own dependency management that meets your application’s needs.
Dependency management for com.nimbusds:oauth2-oidc-sdk
and com.nimbusds:nimbus-jose-jwt
has been removed.
If you are using Spring Security you should rely on the versions that it will pull in as transitive dependencies.
If you are not using Spring Security you should define your own dependency management that meets your application’s needs.
The property to control the version of Prometheus has changed from prometheus-pushgateway.version
to prometheus-client.version
. This is to reflect the fact that the property manages the version of every module in Prometheus’s client, not just the pushgateway.
To use embedded mongo, the spring.mongodb.embedded.version
property must now be set.
This helps to ensure that the MongoDB version that is used by the embedded support matches the MongoDB version that your application will use in production.
The dependency management for the Oracle database driver has been streamlined.
If you are still relying on the old com.oracle.ojdbc
groupId, you need to upgrade to the com.oracle.database.jdbc
group as we have removed dependency management for the former.
The 7.12 release of Flyway moved the Vault-related settings to a closed-source extension. Unfortunately, this prevents Spring Boot from configuring them. As a result, the corresponding spring.flyway.vault-secrets
, spring.flyway.vault-token
, and spring.flyway.vault-url
properties have been removed. If you are a Flyway Teams user, the settings can be configured via a FlywayConfigurationCustomizer
bean, FluentConfigiguration.getExtensionConfiguration
and the Vault ApiExtension
The spring.webflux.session
property group has been deprecated and relocated to server.reactive.session
.
The old properties will continue to work, but you should migrate to the new ones if possible.
The configuration properties used to configure an Elasticsearch client have been consolidated.
Previously, a number of common properties for configuring the blocking high-level REST client and the reactive REST client were duplicated across spring.elasticsearch.rest
and spring.data.elasticsearch.clients.reactive
.
If you are using the blocking REST client, the following table lists the old properties and their replacements:
Deprecated Property | Replacement |
---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
If you are using the reactive client, the following table lists the old properties and their replacements: |
Deprecated Property | Replacement |
---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
To align the default behavior with Spring Data Couchbase, @Persistent
-annotated types are no longer considered.
If you are relying on that annotation, @Document
can be used instead.
The Maven plugin’s build info support now uses the value of the project.build.outputTimestamp
property as the default build time.
If the property is not set, the the build session’s start time is used as it was previously.
As before, the time can be disabled entirely by setting it to off
.
If you are using @ConfigurationProperties
with a Java 16 record and the record has a single constructor, it no longer needs to be annotated with @ConstructorBinding
.
If your record has multiple constructors, @ConstructorBinding
should still be used to identify the constructor to use for property binding.
For a resource server application using spring-security-oauth2-resource-server
which is configured with an OpenID connect issuer-uri
, Spring Boot now auto-configures a SupplierJwtDecoder
instead of a NimbusJwtDecoder
.
This enables Spring Security’s lazy OIDC discovery support which improves startup time.
Similarly, for reactive applications, a ReactiveSupplierJwtDecoder
is auto-configured.
Spring Boot 2.6 upgrades to Apache Kafka 3.0 which intended to enable the strongest delivery guarantees by default. Due to a Kafka bug, these changes were not fully in place until Kafka 3.0.1 which is used by Spring Boot 2.6.5.
Enabling idempotence can cause problems, for example a ClusterAuthorizationException
, when you upgrade. If you experience problems and want to disable idempotence (the default prior to Kafka 3.0.1), set the following property:
spring.kafka.producer.properties.enable.idempotence=false
Tip
|
Check the configuration changelog for a complete overview of the changes in configuration. |
You can now configure SameSite
attributes on session cookies for servlet applications using the server.servlet.session.cookie.same-site
property.
This works with auto-configured Tomcat, Jetty and Undertow servers.
In addition, the CookieSameSiteSupplier
interface can be used if you want to apply a SameSite
attribute to other cookies.
See the updated documentation for more details and some example code.
The session properties supported for reactive servers (previously under spring.webflux.session
) have been expanded with this release.
The new properties are available under server.reactive.session
and now offer parity with the servlet versions.
Spring Boot sanitizes sensitive values present in the /env
and /configprops
endpoints.
While it was possible to configure which properties get sanitized via configuration properties, users might want to apply sanitization rules based on which PropertySource
the property originated from.
For example, Spring Cloud Vault uses vault to store encrypted values and load them into the Spring environment.
Since all values are encrypted, it would make sense to blank the values of every key in an entire property source.
Such sanitization customizations can be configured by adding a @Bean
of type SanitizingFunction
.
The info
endpoint can now expose Java Runtime information under the java
key, as shown in the following example:
{
"java": {
"vendor": "BellSoft",
"version": "17",
"runtime": {
"name": "OpenJDK Runtime Environment",
"version": "17+35-LTS"
},
"jvm": {
"name": "OpenJDK 64-Bit Server VM",
"vendor": "BellSoft",
"version": "17+35-LTS"
}
}
}
To expose this information in the info
endpoint’s response, set the management.info.java.enabled
property to true
.
It’s now possible to exclude specific properties from being added to the build-info.properties
file generated by the Spring Boot Maven or Gradle plugin.
Maven users can exclude the standard group
, artifact
, name
, version
or time
properties using the <excludeInfoProperties>
tag.
For example, to exclude the version
property the following configuration can be used:
<configuration>
<excludeInfoProperties>
<excludeInfoProperty>version</excludeInfoProperty>
</excludeInfoProperties>
</configuration>
Gradle users can use null
values with the properties
block of the DSL:
task buildInfo(type: org.springframework.boot.gradle.tasks.buildinfo.BuildInfo) {
destinationDir project.buildDir
properties {
version = null
}
}
Health groups can be made available at an additional path on either the main or management port. This is useful in cloud environments such as Kubernetes, where it is quite common to use a separate management port for the actuator endpoints for security purposes. Having a separate port could lead to unreliable health checks because the main application might not work properly even if the health check is successful. A typical configuration would have all the actuator endpoints on a separate port and health groups that are used for liveness and readiness available at an additional path on the main port.
Health groups can be configured to include/exclude parts of a CompositeHealthContributor
.
This can be done by specifying the fully qualified path to the component.
For example, a component spring
which is nested inside a composite named test
, can be included using test/spring
.
Auto-configuration exposes two metrics related to application startup:
-
application.started.time
: time taken to start the application. -
application.ready.time
: time taken for the application to be ready to service requests.
Micrometer’s DiskSpaceMetrics
are now auto-configured.
disk.free
and disk.total
metrics are provided for the partition identified by the current working directory.
To change the path that is used, define your own DiskSpaceMetrics
bean.
Micrometer’s ExecutorServiceMetrics
are now auto-configured for all ThreadPoolTaskExecutor
and ThreadPoolTaskScheduler
beans, as long as the underlying ThreadPoolExecutor
is available.
Metrics are tagged with the name of the executor derived from its bean name.
Micrometer’s JettyConnectionMetrics
are now auto-configured.
Additionally, when server.ssl.enabled
is set to true
, Micrometer’s JettySslHandshakeMetrics
are also auto-configured.
Support for exporting metrics to the Dynatrace v2 API has been added.
With a local OneAgent running on the host, a dependency on io.micrometer:micrometer-registry-dynatrace
is all that is required.
Without a local OneAgent, the management.metrics.export.dynatrace.uri
and management.metrics.export.dynatrace.api-token
properties must be configured.
Other settings that are specific to the v2 API can be configured using the management.metrics.export.dynatrace.v2
properties.
Please refer to the updated reference documentation for further details.
The Maven and Gradle plugins now support applying additional tags to a generated image after it is built using a tags
configuration parameter.
A network
configuration parameter has been added to the Maven plugin spring-boot:build-image
goal and the Gradle bootBuildImage
task. This parameter can be used to configure the network driver used by the container that runs the Cloud Native Buildpacks builder process.
Auto-configuration for Spring Data Envers is now provided.
To use it, add a dependency on org.springframework.data:spring-data-envers
and update your JPA repositories to extend from RevisionRepository
.
Redis (both Jedis and Lettuce) will now automatically enable pooling when commons-pool2
is on the classpath.
Set spring.redis.jedis.pool.enabled
or spring.redis.lettuce.pool.enabled
to false
to disable pooling if required.
Auto-configuration for Spring AMQP’s new spring-rabbit-stream
module has been added.
A StreamListenerContainer
is auto-configured when the spring.rabbitmq.listener.type
property is set to stream
.
The spring.rabbitmq.stream.*
properties can be used to configure access to the broker and spring.rabbitmq.listener.stream.native-listener
can be used to enable native listener support.
Previously, Kafka only supported file-based key and trust stores for SSL. With KIP-651, it is now possible to use the PEM format. Spring Boot has added the following properties that allow configuring SSL certificates and private keys using the PEM format:
-
spring.kafka.ssl.key-store-key
-
spring.kafka.ssl.key-store-certificate-chain
-
spring.kafka.ssl.trust-store-certificates
The Maven Plugin’s start
goal has been made more configurable from the command line.
Its wait
and maxAttempts
properties can be specified using spring-boot.start.wait
and spring-boot.start.maxAttempts
respectively.
A new annotation, @WebServiceServerTest
, that can be used to test Web Service @Endpoint
beans has been introduced.
The annotation creates a test slice containing @Endpoint
beans and auto-configures a MockWebServiceClient
bean that can be used to test your web service endpoints.
The application’s MessageSource
is now used when resolving {parameters}
in constraint messages.
This allows you to use your application’s messages.properties
files for Bean Validation messages.
Once the parameters have been resolved, message interpolation is completed using Bean Validation’s default interpolator.
Developers could use WebTestClient
to test WebFlux apps in mock environments, or any Spring web app against live servers. This change also enables WebTestClient
for Spring MVC in mock environments: classes annotated with @AutoConfigureMockMvc
can get injected a WebTestClient
. This makes our support complete, you can now use a single API to drive all your web tests!
Spring Integration PollerMetadata
(poll unbounded number of messages every second) can now be customized with spring.integration.poller.*
configuration properties.
Log4j2’s composite configuration is now supported.
To use it, configure the logging.log4j2.config.override
property with a comma-separated list of secondary configuration files that will override the main configuration. The main configuration is sourced from Spring Boot’s defaults, a well-known standard location such as log4j2.xml
, or the location specified by the logging.config
property as before.
Spring Boot 2.6 moves to new versions of several Spring projects:
Numerous third-party dependencies have also been updated, some of the more noteworthy of which are the following:
-
Commons DBCP 2.9
-
Commons Pool 2.11
-
SnakeYAML 1.29
Apart from the changes listed above, there have also been lots of minor tweaks and improvements including:
-
The failure analysis for a
NoSuchMethodError
now includes information about the location from which the calling class was loaded. -
A
ClientResourcesBuilderCustomizer
bean can now be defined to customize Lettuce’sClientResources
will retaining the default auto-configuration. -
Configuration properties for Flyway’s
detectEncoding
,failOnMissingLocations
, andignoreMigrationPatterns
configuration settings have been added. -
A custom
ResourceLoader
to be used by the application can be provided when creating aSpringApplicationBulder
. -
A
WebSessionIdResolver
can now be defined to customize the resolver that will be used by the auto-configuredWebSessionManager
. -
Any
RSocketConnectorConfigurer
beans are now automatically applied to the auto-configuredRSocketRequester.Builder
. -
spring-boot-configuration-processor
can now generate metadata for classes annotated with Lombok’s@Value
. -
A new
server.tomcat.reject-illegal-header
configuration property can be set to true to configure Tomcat to accept illegal headers. -
When using Stackdriver, labels can now be configured on the monitor resource by setting
management.metrics.export.stackdriver.resource-labels.*
configuration properties. -
@EntityScan
now supports comma-separated values in itsbasePackages
attribute. -
A new configuration property,
server.netty.idle-timeout
, that can be used to control Reactor Netty’s idle timeout has been added. -
The location from which Devtools loads its global settings can now be configured using the
spring.devtools.home
system property or theSPRING_DEVTOOLS_HOME
environment variable. -
The setter methods on
RabbitTemplateConfigurer
are nowpublic
-
The
heapdump
endpoint is now supported on OpenJ9 where it will produce a heap dump in PHD format. -
New configuration properties are now supported for multipart support in WebFlux, under
spring.webflux.multipart.*
-
Any
ContainerCustomizer
bean is now called to configure the auto-configured Spring AMQPMessageListenerContainer
-
Jackson’s default leniency can be configured using the
spring.jackson.default-leniency
property. -
Distribution statistic’s expiry and buffer length are now configurable.
-
Command Latency metrics for Lettuce are now auto-configured.
-
Disk space metrics can be configured with one or more paths using the
management.metrics.system.diskspace.paths
property. -
Users can take control over the Redis auto-configuration by providing a
RedisStandaloneConfiguration
bean. -
The URLs of all available data sources is now logged when the H2 console is auto-configured.
-
A new configuration property,
spring.integration.management.default-logging-enabled
, can be used to disable Spring Integration’s default logging by setting its value tofalse
. -
Auto-configuration for
UserDetailsService
will now back off in the presence of anAuthenticationManagerProvider
bean.
-
AbstractDataSourceInitializer
has been deprecated in favor ofDataSourceScriptDatabaseInitializer
. Similarly, subclasses ofAbstractDataSourceInitializer
have been deprecated in favour of newDataSourceScriptDatabaseInitializer
-based equivalents. -
SpringPhysicalNamingStrategy
has been deprecated in favor of Hibernate 5.5’sCamelCaseToUnderscoresNamingStrategy
-
Three methods on
AbstractApplicationContextRunner
have been deprecated in favor of newRunnerConfiguration
-based equivalents -
The
started
andrunning
methods inSpringApplicationRunListener
have been replaced with versions that accept aDuration
. -
Constructors in
ApplicationStartedEvent
andApplicationReadyEvent
have been replaced with versions that accept aDuration
. -
The
EnvironmentEndpoint.sanitize
has been deprecated for removal.