Spinnaker第七节—Orca代码详解
这个春节假期由于新冠弄得一直在家憋着,闲得无事把spinnaker代码翻出来把每个微服务重新review一遍,做了一些笔记,今天拿来分享一下,后续如果有spinnaker开发或者bug修复可以直接有的放矢。计划先把核心的几个微服务分几期慢慢介绍下:orca、clouddriver、deck,今天先介绍orca。
Orca名词解释
Orchestration
与pipeline平行,我个人理解是orca这个微服务主要是为了支持pipeline工作流的,为了对接临时性的任务包扩展出了ORCHESTRATION
Execution
一次执行。分为两种类型:PIPELINE和ORCHESTRATION,核心属性数一个stage的集合
Stage
orca最复杂的类型,关联一个execution和多个task,还维护有parentStage、downStreamStage、output等重要信息
StageContext
驱动这个阶段的环境,提供组件步骤所需的输入
Task
Stage的组成部分,orca中最小的操作单元
ExecutionRunner Execution的执行者
ExecutionRepository Execution的存储者,我们的是redis,也承载着检索查询
ExecutionLauncher Execution的启动者,接收json封装成Execution,进行存储和启动
1 Json封装Execution也会分为PIPELINE和ORCHESTRATION两种,本质没区别
2 最终启动执行:
public Execution start(Execution execution) throws Exception {
executionRunner.start(execution);
return execution;
}
orca-web子项目(入口)
controller
OperationsController: 与Execution相关的controller
/orchestrate*
最终startPipeline,以PIPELINE形式启动execution
/plan*
不会真正执行,经过一系列校验后通过ExecutionRepository存储executor
/ops*
最终startTask,以ORCHESTRATION形式启动executor
TaskController:
/pipelines*
对于pipeline的停、重、查,最终映射的对象类型还是execution
/task*
对于task的停、重、查,最终映射的对象类型还是execution
orca-web总结:OperationsController和TaskController看似有重叠的部分都会涉及到execution,但是侧重点不同。OperationsController侧重于创建和启动,TaskController侧重于后续的暂停、撤销和查询。
orca-core子项目(抽象层)
根目录:
定义一系列的task:retryableTask、cancelableTask等
annotations 自定义sync,用于标注同步,可以作为自定义标注的学习参考
clouddriver.utils 根据stage获取各种clouddriver的信息
commands 意义不大,不需要看
config:orca配置
exceptions:除开定义了异常,还定义了处理异常的Handler,借鉴下
/*
* Copyright 2014 Netflix, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.netflix.spinnaker.orca.exceptions;
import java.util.Collections;
import java.util.Map;
import com.google.common.base.Throwables;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.core.annotation.Order;
import static java.lang.String.format;
import static org.springframework.core.Ordered.LOWEST_PRECEDENCE;
@Order(LOWEST_PRECEDENCE)
public class DefaultExceptionHandler implements ExceptionHandler {
private final Logger log = LoggerFactory.getLogger(getClass());
@Override public boolean handles(Exception e) {
return true;
}
@Override
public ExceptionHandler.Response handle(String taskName, Exception e) {
Map<String, Object> exceptionDetails = ExceptionHandler.responseDetails("Unexpected Task Failure", Collections.singletonList(e.getMessage()));
exceptionDetails.put("stackTrace", Throwables.getStackTraceAsString(e));
log.warn(format("Error occurred during task %s", taskName), e);
return new ExceptionHandler.Response(e.getClass().getSimpleName(), taskName, exceptionDetails, false);
}
}
events:
定义了execution、stage、task的各种事件,
也定了以上事件的监听基类,本质上没有自己做处理,都是丢给delegate在处理,代理模式
private void onTaskStarted(TaskStarted event) {
Execution execution = retrieve(event);
List<Stage> stages = execution.getStages();
stages
.stream()
.filter(it -> it.getId().equals(event.getStageId()))
.findFirst()
.ifPresent(stage ->
delegate.beforeTask(
persister,
stage,
stage.getTasks().stream().filter(it -> it.getId().equals(event.getTaskId())).findFirst().get()
)
);
}
listeners:
定义了ExecutionListener和StageListener两个接口,也就是events中的delegate,定义了一些阶段和要重写的方法。像echo通知等就是这种实现机制。
/*
* Copyright 2016 Netflix, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License")
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.netflix.spinnaker.orca.listeners;
import com.netflix.spinnaker.orca.ExecutionStatus;
import com.netflix.spinnaker.orca.pipeline.model.Stage;
import com.netflix.spinnaker.orca.pipeline.model.Task;
public interface StageListener {
default void beforeTask(Persister persister,
Stage stage,
Task task) {
// do nothing
}
default void beforeStage(Persister persister,
Stage stage) {
// do nothing
}
default void afterTask(Persister persister,
Stage stage,
Task task,
ExecutionStatus executionStatus,
boolean wasSuccessful) {
// do nothing
}
default void afterStage(Persister persister,
Stage stage) {
// do nothing
}
}
locks
看了下是为了多节点部署而设计的,目前阶段可以不关心
pipeline
最核心的一个类StageDefinitionBuilder,所有的stage都要继承它
AcquireLockStage、CheckPreconditionsStage、EvaluateVariablesStage、ReleaseLockStage、WaitStage
orca-clouddriver子项目(核心实现)
先搞明白3个服务:oort、mort和kato,这三个都是指向了clouddriver的restful
oort是大头:对接cluster、image、instance、loadbalance等云平台资源
mort是小头:对接vpc、安全组、账号鉴权
kkato是辅助:对接其它辅助的资源
pipeline:
stage对于task的组装,注意其中的provider包是留给厂商自己私有化的,所以关于orca云厂商差异化的部分也都集中在这里。
tasks:
具体的task实现,也是未来集中关心的地方,provider中各个厂商都有自己的私有化
上一篇: Weblogic9扩展JAAS应用注意点