I have Spring Data REST custom controller that returns ResponseEntity<StreamingResponseBody>
In my application.yaml file, I have defined a custom task executor to be used for the StreamingResponseBody.
spring:
task:
execution:
pool:
max-size: 16
queue-capacity: 100
However, mvc is still using the SimpleAsyncTaskExecutor, instead of the one defined above.
An Executor is required to handle java.util.concurrent.Callable return values.
Please, configure a TaskExecutor in the MVC config under "async support".
The SimpleAsyncTaskExecutor currently in use is not suitable under load.
After a little debugging, I found out that the StreamingResponseBodyReturnValueHandler does not set the applicationTaskExecutor on the WebAsyncTask type i.e. StreamingResponseBodyTask and as a result the SimpleAsyncTaskExecutor is used.
Callable<Void> callable = new StreamingResponseBodyTask(outputMessage.getBody(), streamingBody);
WebAsyncUtils.getAsyncManager(webRequest).startCallableProcessing(callable, mavContainer);
The WebMvcAutoConfiguration is picking up the applicationTaskExecutor and setting it correctly, when I debugged this method in WebMvcAutoConfiguration
#Override
public void configureAsyncSupport(AsyncSupportConfigurer configurer) {
if (this.beanFactory.containsBean(TaskExecutionAutoConfiguration.APPLICATION_TASK_EXECUTOR_BEAN_NAME)) {
Object taskExecutor = this.beanFactory
.getBean(TaskExecutionAutoConfiguration.APPLICATION_TASK_EXECUTOR_BEAN_NAME);
if (taskExecutor instanceof AsyncTaskExecutor) {
configurer.setTaskExecutor(((AsyncTaskExecutor) taskExecutor));
}
}
Duration timeout = this.mvcProperties.getAsync().getRequestTimeout();
if (timeout != null) {
configurer.setDefaultTimeout(timeout.toMillis());
}
}
Am I missing anything? How can I apply the ThreadPoolTaskExecutor for a StreamingResponseBody?
spring-data-rest Custom Repositories (#RepositoryRestResource or #BasePathAwareController) Ignore AsyncSupportConfigurer
see RepositoryRestMvcConfiguration#repositoryExporterHandlerAdapter here.
But spring-webmvc applies AsyncSupportConfigurer see WebMvcConfigurationSupport#requestMappingHandlerAdapter here
AsyncSupportConfigurer configurer = getAsyncSupportConfigurer();
if (configurer.getTaskExecutor() != null) {
adapter.setTaskExecutor(configurer.getTaskExecutor());
}
if (configurer.getTimeout() != null) {
adapter.setAsyncRequestTimeout(configurer.getTimeout());
}
adapter.setCallableInterceptors(configurer.getCallableInterceptors());
adapter.setDeferredResultInterceptors(configurer.getDeferredResultInterceptors());
If it's possible, use a simple #RestController or #Controller,
If not, you can create an issue here to add support for that
Related
Hy
I am using springboot 3 with the new micrometer observation. Is there a way to prevent generating a trace_id/span_id for some paths like /actuator/prometheus? Observation add a trace id for each call to /actuator/*.
Thank you
You need to give more information about a problem, but i think you have set this line in application.propeties like:
management.endpoints.web.exposure.include=/actuator/*
But exists option like:
management.endpoints.web.exposure.exclude=/actuator/prometheus
I managed to find a half solution to the problem, by defining the ObservationRegistry this way:
#Bean
#ConditionalOnMissingBean
ObservationRegistry observationRegistry() {
PathMatcher pathMatcher = new AntPathMatcher("/");
ObservationRegistry observationRegistry = ObservationRegistry.create();
observationRegistry.observationConfig().observationPredicate((name, context) -> {
if(context instanceof ServerRequestObservationContext) {
return !pathMatcher.match("/actuator/**", ((ServerRequestObservationContext) context).getCarrier().getRequestURI());
} else {
return true;
}
});
return observationRegistry;
}
This doesn't completely ignore the actuator requests, only the first span. So if you have for example Spring Security on your classpath, those spans are left intact.
EDIT: Looks like you don't need to redefine the entire observation registry, you can use a bean like the one show here: https://docs.spring.io/spring-security/reference/servlet/integrations/observability.html
EDIT2: From what I can tell, you need to include these 2 beans, and no actuator call will be traced (completely disables tracing for spring security too):
#Bean
ObservationRegistryCustomizer<ObservationRegistry> skipActuatorEndpointsFromObservation() {
PathMatcher pathMatcher = new AntPathMatcher("/");
return (registry) -> registry.observationConfig().observationPredicate((name, context) -> {
if (context instanceof ServerRequestObservationContext observationContext) {
return !pathMatcher.match("/actuator/**", observationContext.getCarrier().getRequestURI());
} else {
return true;
}
});
}
#Bean
ObservationRegistryCustomizer<ObservationRegistry> skipSecuritySpansFromObservation() {
return (registry) -> registry.observationConfig().observationPredicate((name, context) ->
!name.startsWith("spring.security"));
}
Also, you might want to keep an eye out on this issue: https://github.com/spring-projects/spring-framework/issues/29210
I want to increase sampling of only those spans which contains error or exception.
If that's possible then how to do that?
I'm using Spring boot(2.2+), sleuth(2.2+) and zipkin(2.2+).
If you're using Brave (which is the default) you can register a bean of type brave.handler.SpanHandler and over there in the method end return false if the MutableSpan had any errors.
#Component
class ExceptionCheckingSpanHandler extends brave.handler.SpanHandler {
#Override
public boolean end(TraceContext context, MutableSpan span, Cause cause) {
return span.error() != null;
}
}
I use Spring Boot 2.1.2.RELEASE, and I try to use Micrometer with CompositeMeterRegistry. My goal is to publish some selected meters to ElasticSearch. The code below shows my sample config. The problem is, that the filter is completely ignored (so all metrics are sent to ElasticSearch), although I can see in the logs that it was processed ("filter reply of meter ..." lines).
Strangely, if I define the MeterFilter as a Spring bean, then it's applied to ALL registries (however, I want it to be applied only on "elasticMeterRegistry").
Here is a sample configuration class:
#Configuration
public class AppConfiguration {
#Bean
public ElasticConfig elasticConfig() {
return new ElasticConfig() {
#Override
#Nullable
public String get(final String k) {
return null;
}
};
}
#Bean
public MeterRegistry meterRegistry(final ElasticConfig elasticConfig) {
final CompositeMeterRegistry registry = new CompositeMeterRegistry();
registry.add(new SimpleMeterRegistry());
registry.add(new JmxMeterRegistry(new JmxConfig() {
#Override
public Duration step() {
return Duration.ofSeconds(10);
}
#Override
#Nullable
public String get(String k) {
return null;
}
}, Clock.SYSTEM));
final ElasticMeterRegistry elasticMeterRegistry = new ElasticMeterRegistry(elasticConfig, Clock.SYSTEM);
elasticMeterRegistry.config().meterFilter(new MeterFilter() {
#Override
public MeterFilterReply accept(Meter.Id id) {
final MeterFilterReply reply =
id.getName().startsWith("logback")
? MeterFilterReply.NEUTRAL
: MeterFilterReply.DENY;
log.info("filter reply of meter {}: {}", id.getName(), reply);
return reply;
}
});
registry.add(elasticMeterRegistry);
return registry;
}
}
So, I expect ElasticSearch to receive only "logback" metrics, and JMX to receive all metrics.
UPDATE:
I have played with filters and found a "solution", but I don't really understand why the code above doesn't work.
This works:
elasticMeterRegistry.config().meterFilter(new MeterFilter() {
#Override
public MeterFilterReply accept(Meter.Id id) {
final MeterFilterReply reply =
id.getName().startsWith("logback")
? MeterFilterReply.ACCEPT
: MeterFilterReply.DENY;
log.info("filter reply of meter {}: {}", id.getName(), reply);
return reply;
}
});
The difference is: I return ACCEPT instead of NEUTRAL.
Strangely, the following code does not work (ES gets all metrics):
elasticMeterRegistry.config().meterFilter(
MeterFilter.accept(id -> id.getName().startsWith("logback")));
But this works:
elasticMeterRegistry.config().meterFilter(
MeterFilter.accept(id -> id.getName().startsWith("logback")));
elasticMeterRegistry.config().meterFilter(
MeterFilter.deny());
CONCLUSION:
So, it seems that instead of NEUTRAL, the filter should return ACCEPT. But for meters not starting with "logback", my original filter (with NEUTRAL) returns DENY. Then why are those metrics published to ElasticSearch registry?
Can someone explain this?
This is really a composite of questions. I'll just point out a few points.
For the MeterRegistry bean you defined, Spring Boot will auto-configure an ElasticMeterRegistry bean as there's no ElasticMeterRegistry bean. Instead of creating a CompositeMeterRegistry bean on your own, just define a custom ElasticMeterRegistry bean which is applied the MeterFilter you want and let Spring Boot create one (CompositeMeterRegistry bean) for you.
For MeterFilterReply, ACCEPT will accept the meter immediately, DENY will deny the meter immediately, and NEUTRAL will postpone the decision to next filter(s). Basically meters will be accepted unless there's any DENY.
im developing a Project using Spring Batch and JavaConfig (no XML).
I am creating Jobs using an Autowired jobBuilderFactory.
is it somehow possible to set the Incrementer for the Factory globally ?
return jobBuilderFactory.get("jobName").incrementer(new RunIdIncrementer()).start(stepOne()).next(lastStep()).build();
sorry if this is a dump question but i am new to Spring Batch and did not find a working solution.
With XML config you would use bean definition inheritance, but you said you don't use XML.
Since there is no equivalent of XML bean definition inheritance with Java config (See details here: https://stackoverflow.com/a/23266686/5019386), you can create the RunIdIncrementer globally in your config and use it in job definitions:
public JobParametersIncrementer jobParametersIncrementer() {
return new RunIdIncrementer();
}
public JobBuilder getJobBuilder(String jobName) {
return jobBuilderFactory.get(jobName)
.incrementer(jobParametersIncrementer());
}
#Bean
public Job job1() {
return getJobBuilder("job1")
.start(step())
.build();
}
#Bean
public Job job2() {
return getJobBuilder("job2")
.start(step())
.build();
}
But again and as said in comments, you will end up having run.id values that are not consecutive for each job.
#ConfigurationProperties locations is deprecated in Spring Boot 1.4.x and option is now removed in 1.5.x
I was using it like this: BucketTestConfig.java
For now with deprecation, I'm trying to set the system property spring.config.location for both production code and test code as an alternative.
./gradlew clean test is still failing although I set the system property.
What is the best alternative for deprecated #ConfigurationProperties locations in this case?
UPDATE:
Using SpringApplicationBuilder.properties() doesn't work in the test (BucketTestRepositoryTests).
Using SpringApplicationBuilder.listeners() doesn't work in the test (BucketTestRepositoryTests), either.
UPDATE (2nd):
There was no reason to depend on #ConfigurationProperties in my case, so I went with Yaml instead as follows: https://github.com/izeye/spring-boot-throwaway-branches/commit/a1290672dceea98706b1a258f8a17e2628ea01ee
So this question's title is invalid and this question can be deleted.
Follow this thread for more information.
Basically, this thread suggests two options
First option is to set spring.config.name to a list of the files you want to load:
new SpringApplicationBuilder(Application.class)
.properties("spring.config.name=application,mine")
.run(args);
Second options is to add listeners
new SpringApplicationBuilder(SanityCheckApplication.class)
.listeners(new LoadAdditionalProperties())
.run(args);
Content of LoadAdditionalProperties
#Component
public class LoadAdditionalProperties implements ApplicationListener<ApplicationEnvironmentPreparedEvent> {
private ResourceLoader loader = new DefaultResourceLoader();
#Override
public void onApplicationEvent(ApplicationEnvironmentPreparedEvent event) {
try {
Resource resource = loader.getResource("classpath:mine.properties");
PropertySource<?> propertySource = new PropertySourcesLoader().load(resource);
event.getEnvironment().getPropertySources().addLast(propertySource);
} catch (IOException ex) {
throw new IllegalStateException(ex);
}
}
}