Docs related to gh-33280 (log correlation) and
gh-35593 (application name in each log line) need some polishing:
- Fix project names
- Show how to avoid having the application name duplicated in logs
- Call out that a trailing space is needed in the correlation pattern
Closes gh-36035
See gh-33280
See gh-35593
Update Logback and Log4J2 so that they include the application name on
each log line. If `spring.application.name` had not been set, or if
`logging.include-application-name` is `false` then the name is not
logged.
Closes gh-35593
Add support for logging correlation IDs with Logback or Log4J2 whenever
Micrometer tracing is being used.
The `LoggingSystemProperties` class now accepts a defualt value resolver
which will be used whenever a value isn't in the environment. The
`AbstractLoggingSystem` provides a resolver that supports the
`logging.pattern.correlation` property and will return a value whenever
`LoggingSystem.EXPECT_CORRELATION_ID_PROPERTY` is set.
Using `LoggingSystem.EXPECT_CORRELATION_ID_PROPERTY` allows us to
provide a consistent width for the correlation ID, even when it's
missing from the MDC.
The exact correlation pattern returned will depend on the `LoggingSytem`
implementation. Currently Logback and Log4J2 are supported and both
make use of a custom converter which delegates to a new
`CorrelationIdFormatter` class.
Closes gh-33280
Update `ReactorClientHttpConnectorFactory` to that SSL configuration
is applied in addition to any configured mappers.
Prior to this commit, SSL configuration would prevent configured
mappers from being applied.
See gh-35914
In Spring Framework 5.x with Jetty 9, the reactive
JettyRequestUpgradeStrategy was able to initialize Jetty's WebSocket
infrastructure itself. With Jetty 10 this is no longer possible and
Boot must perform the initialization as part of preparing the
reactive JettyWebServer.
This commit updates the reactive WebSocket auto-configuration to
initialize Jetty's WebSocket infrastructure as part of creating the
reactive JettyWebServer.
Fixes gh-33347
Gradle 7.3.3 does not support Java 20 so the test that verifies that
the plugin fails fast when running on a version of Gradle that it
does not support needs to be disabled on Java 20 and later.
See gh-35931
Using `@EnableBatchProcessing` is no longer necessary to enable
Batch's auto-configuration so it's no longer a good example when
discussing sliced tests and user configuration.
This commit replaces `@EnableBatchProcessing` with an alternative
annotation, `@EnableMongoAuditing` and updates the accompanying
text.
Closes gh-33435
Previously, we tried to prevent spring-boot-docker-compose from
causing problems with AOT and native images by excluding it from
the AOT processing classpath. This allowed AOT processing to
succeed. We cannot apply the same exclusion to the native image
classpath so spring-boot-docker-compose was still included in the
native image. This results in a failure at runtime due to missing
reflection hints.
This commit reverts that changes that excluded
spring-boot-docker-compose from the AOT processing classpath. This
allows AOT processing to generate reflection hints but reintroduces
the failure caused by the connection details bean definitions using
an instance supplier callback. To overcome this problem we disable
DockerComposeLifecycleManager during AOT processing and in a native
image. This ensures that no attempt is made to call docker compose up
and no connection details beans are defined.
Fixes gh-35676
When excluded from the AOT processing classpath, a reflection hint
for DevTools' RestartScopeInitializer that's registered in
spring.factories is not generated. When the native image is compiled
we have no control over its classpath so DevTools is included. This
causes a failure at runtime as RestartScopeInitializer cannot be
loaded due to the missing reflection hint.
Until we have control over the native image classpath, we need to
include DevTools on the AOT processing classpath and then rely on
DevTools disabling itself in a native image which it already does.
Fixes gh-35853
@ConditionalOnEnabledTracing is now applied to the minimal amount of
beans. The beans which are annotated with it are beans that will lead
to span sending to backends.
This leaves the majority of the Micrometer Tracing, Brave and
OpenTelemetry infrastructure untouched in tests.
Closes gh-35354
Previously, when using SpringApplication.from() any additional
sources configured using with() would be applied to every
SpringApplication that was created within the scope of the call to
run(). This caused problems with Spring Cloud's bootstrap context
where the additional sources would be applied to both the user's
application and to the boostrap context's application.
This commit updates the hook that's used to apply the additional
sources so that it's only applied once. This results in the
additional sources only being added to the first SpringApplication
that is run.
Closes gh-35873
Support custom Log4J2 log levels by changing `LoggerConfiguration` so
that it can now report levels using a `LevelConfiguration` object
rather than the limited `LogLevel` enum.
The `Log4J2LoggingSystem` class now uses `LevelConfiguration.ofCustom`
for custom logging levels, rather than throwing an exception.
The `LoggersEndpoint` has also been updated so that it can return the
custom logger name.
Fixes gh-35227
When Hazelcast is started on a fixed port and that port is already in
use, it does not fail to start which makes the problem hard to
detect. A symptom of the problem is that clients will not be able to
connect and will either retry indefinitely or will timeout depending
on their configuration.
This commit updates the Hazelcast client tests to start the Hazelcast
instance on an ephemeral port and to customize the client
configuration with the instance's address before use. This should
allow the client tests to work reliably in an environment where
Hazelcast's default port is already in use.
Closes gh-35903
Most notably, this commit splits the tests that use Spring REST Docs
out into a separate task for which predictive test selection is
disabled. This allows it to be cached and use Gradle's built-in
up-to-date checking, thereby avoiding the generation of new snippets
and the need to then run the asciidoctor and asciidoctorPdf tasks.
It also updates spring-boot-smoke-test-junit-vintage to disable
predictive test selection so that we can continue to assert that
the some tests were executed.
See gh-35869
Johnny Lim
Scott Frederick
James Stansell
Andy Wilkinson
ThomazPassarelli
Moritz Halbritter
Jonatan Ivanov
Phillip Webb
Fernando Cappi
Stephane Nicoll
Sébastien Deleuze