Azkaban的Web Server源码探究系列20:resolve&buildFlow

上节，我们加载了job的描述属性 + 正确性验证 ！

==============================================================

stop in azkaban.project.DirectoryFlowLoader.resolveDependencies

==============================================================

这个也很简单，比如有个Hello.job,World.job

指定了world依赖Hello

则最后在

private HashMap<String, Map<String, Edge>> nodeDependencies;

这种结构里存的是什么呢？

HashMap<

String, ---World

Map<

String, ---Hello

Edge ---<Hello,World>

>

>

是不是很简单，没啥可说的！

接下来是

// Create the flows.

buildFlowsFromDependencies();

到目前为止，我们装载了点+校验正确性+边。有了点和边，其实就有了图，有了图，我们就可以构建Flow了

这也正是我们下面要做的事情！

======================================================================================

细节略

======================================================================================

最后就是生成报告，只要写的没问题，都通过！

然后就是数据库的操作，这是我们需要关心的！

---

synchronized (project) {

int newVersion = projectLoader.getLatestProjectVersion(project) + 1;

Map<String, Flow> flows = loader.getFlowMap();

for (Flow flow : flows.values()) {

flow.setProjectId(project.getId());

flow.setVersion(newVersion);

}

logger.info("Uploading file to db " + archive.getName());

projectLoader.uploadProjectFile(project, newVersion, fileType, archive.getName(), archive,

uploader.getUserId());

logger.info("Uploading flow to db " + archive.getName());

projectLoader.uploadFlows(project, newVersion, flows.values());

logger.info("Changing project versions " + archive.getName());

projectLoader.changeProjectVersion(project, newVersion, uploader.getUserId());

project.setFlows(flows);

logger.info("Uploading Job properties");

projectLoader.uploadProjectProperties(project, new ArrayList<Props>(jobProps.values()));

logger.info("Uploading Props properties");

projectLoader.uploadProjectProperties(project, propProps);

}

logger.info("Uploaded project files. Cleaning up temp files.");

projectLoader.postEvent(project, EventType.UPLOADED, uploader.getUserId(),

"Uploaded project files zip " + archive.getName());

try {

FileUtils.deleteDirectory(file);

} catch (IOException e) {

file.deleteOnExit();

e.printStackTrace();

}

logger.info("Cleaning up old install files older than " + (project.getVersion() - projectVersionRetention));

projectLoader.cleanOlderProjectVersion(project.getId(), project.getVersion() - projectVersionRetention);

return reports;

}

首先是查询出最大版本号，+1最为最新值。

---

1）上传文件，分段，10M为1个单位，相关表---Project-files

2) project_versions

3)project_flows :记录根据file分析出来的拓扑图信息

比如一个可能的拓扑图如下：

json = "{"metadata":{},"project.id":9,"nodes":[{"layout":{"level":0},"propSource":null,"jobSource":"Hello.job","expectedRuntime":1,"id":"Hello","jobType":"command"},{"layout":{"level":1},"propSource":null,"jobSource":"World.job","expectedRuntime":1,"id":"World","jobType":"command"}],"edges":[{"source":"Hello","target":"World"}],"failure.email":[],"success.email":[],"id":"World","type":"flow","version":11,"mailCreator":"default","props":[],"layedout":false}"