Skip to main content

10 posts tagged with "spring-boot"

View All Tags

解决 Spring Boot 3.5.0 后 Jasypt 无法解析环境变量中的加密字符串问题

· 6 min read
orange
orange
programmer on jvm platform

当 Spring Boot 项目升级到 3.5.0 及更高版本后，使用 jasypt-spring-boot-starter 会遇到一个问题：应用程序无法正确解析环境变量中配置的加密字符串。具体表现为，在应用程序运行期间，读取到的配置值仍然是加密后的字符串（例如 ENC(加密字符串)），而非解密后的原始值。

通过关闭OpenEntityManagerInViewInterceptor确保当下游服务响应缓慢时不会导致数据库连接被占用完

· 3 min read
orange
orange
programmer on jvm platform

OpenEntityManagerInViewInterceptor是spring中的一个拦截器，它的作用是在整个请求上下文复用同一个EntityManager ，从而避免在一个请求中多次创建一个EntityManager, 从而提高性能.
但是如果在一个请求的处理代码中请求了下游服务，而下游服务响应缓慢，那么在下游服务响应之前，EntityManager 会一直被占用，直到整个请求处理完成，这样会导致当服务并发请求量较大时，数据库连接被占用完.
为了避免这种情况，可以通过关闭OpenEntityManagerInViewInterceptor来确保当下游服务响应缓慢时不会导致数据库连接被占用完. 下面将介绍如何关闭OpenEntityManagerInViewInterceptor.

Spring Boot项目启用native构建后由于序列化对象缺少ReflectionHints导致json数据为'{}'

· One min read
orange
orange
programmer on jvm platform

基于graalvm的spring boot项目打包好后调用/scheduling/api/v1/taskDefinitions接口返回如下数据, 其中triggerStrategy字段是空json对象

[
	{
		"id": "BILLING_SYNC_FOR_DEPLOYMENT_127",
		"triggerStrategy": {},
		"tags": {
			"EXECUTOR": "com.fastonetech.billing.sync.scheduling.BillingSyncTaskExecutor"
		},
		"variables": {
			"SOURCE_RCLONE_CONFIG": "TENCENT_ap-beijing",
			"SOURCE_BUCKET": "cheng1201-1310454728",
			"SOURCE_PATH": "",
			"TARGET_RCLONE_CONFIG": "aggregation",
			"TARGET_BUCKET": "billing-aggregation",
			"TARGET_PATH": "TENCENT/ap-beijing/cheng1201-1310454728"
		}
	}
]

spring-boot项目中yml配置文件对duration配置在不给定单位的情况下默认为毫秒

· One min read
orange
orange
programmer on jvm platform

如果在application.yml中配置duration值时不给定单位的话, 那么解析出的单位为毫秒, 如果需要自定义单位的话需要提供@DurationUnit,

建议配置Duration单位, 强制指定单位确保行为符合期望

内部调用栈​

  • BindConverter
  • List<ConversionService> delegates
  • List<Converter> converters
  • StringToDurationConverter (SpringBoot)
  • NumberToDurationConverter (SpringBoot) 此处会根据注解去判断单位
  • DurationStyle.SIMPLE.parse

JavaDelegate里调用Service的方法并且Service方法上有@Transactional注解但DB修改未生效

· 2 min read
orange
orange
programmer on jvm platform

原因​

Camunda调用Delegate时会开启事务(为了保证失败时可以回滚数据), Spring的@Transaction propagation = Propagation.REQUIRED , 然后就用了Camunda的那个事务， Camunda的事务中修改是对DB不生效的所以你看不到事务过程中修改的数据提交的结果

解决方案​


@Component
class StartClusterDelegate(
	private val clusterOperator: ClusterOperator
) : JavaDelegate, LogCapability {

	override fun execute(execution: DelegateExecution) {
		clusterOperator.start(execution.getVariableAs(VARNAME_CLUSTER_ID))
	}
}

@Component
class ClusterOperator(
	private val clusterService: ClusterService,
	private val csClusterOperator: CsClusterOperator,
	private val csStorageOperator: CsStorageOperator,
	private val clusterBastionUserOperator: ClusterBastionUserOperator,
	private val netEnvOperator: NetEnvOperator
) : LongTimeOperationCapability {

	override fun start(id: Long) {
		clusterService.get(id).orNull()!!.let { cluster ->
			clusterService.patch(id) {
				state = ClusterState.STARTING
			}
			logger.info("Begin start cluster $id")
			csStorageOperator.start(cluster.csStorageId)
			csClusterOperator.start(cluster.csClusterId!!)
			clusterBastionUserOperator.start(cluster.id)
			logger.info("Successful to start cluster $id")
			clusterService.patch(id) {
				state = ClusterState.RUNNING
			}
		}

	}

}

@Service
class ClusterService(
	private val engine: ProcessEngine,
	private val clusterRepo: ClusterRepository,
	private val client: DeployService
) : LogCapability {

	@Transactional(propagation = Propagation.REQUIRES_NEW)
	fun patch(id: Long, patchFun: PatchClusterDto.() -> Unit): Either<ClusterOperationError, ClusterDetail> =
		clusterRepo.findById(id).map { originCluster ->
			val patch = PatchClusterDto(patchFun).apply {
				logger.info("patch request {}", this)
			}
			val patchedCluster = originCluster.copy(
				cloudAccount = patch.cloudAccount ?: originCluster.cloudAccount,
				csNetEnvId = patch.csNetEnvId ?: originCluster.csNetEnvId,
				csStorageId = patch.csStoreId ?: originCluster.csStorageId,
				csClusterId = patch.csClusterId ?: originCluster.csClusterId,
				isDelivered = when (patch.deliveryCluster) {
					DeliveryCluster.YES -> true
					DeliveryCluster.NO -> false
					DeliveryCluster.NONE -> originCluster.isDelivered
				},
				state = patch.state ?: originCluster.state
			)
			patchedCluster.apply {
				logger.info("patched cluster {}", this)
			}
		}.map {
			clusterRepo.save(it)
			get(it.id)
		}.orElseGet {
			Either.left(ClusterOperationError.NoSuchCluster(id))
		}
}