presto 转换静态catlog为动态catlog

 

 

近年来,基于hadoop的sql框架层出不穷,presto也是其中的一员.从2012年发展至今,依然保持年轻的活力(版本迭代依然很快),presto的相关介绍,我们就不赘述了,相信看官多对presto有或多或少的了解,详细的一些说明可以看官网(https://prestodb.io)的说明.

　　presto自身功能和思想富有先进性,虽然由于是内存计算,稳定性方面还有很大提升空间,但整体依然在adhoc方面有很好的竞争力.其中在catalog加载的方式上来说比较的固化,官方并没有做出动态的方案出来,导致在添加catalog后必须重启整个集群才可以将新添加的catalog数据源添加到presto中,这无疑在实际的生产环境中很不友好.尤其是在一些中台项目中,需要动态规划的东西非常多.这种模式的catalog添加方式显然不能满足我们的开发需要.

　　因此,在环境的加持下,对presto的加载catlog的方式的源码进行了改造,使其具有热动态添加的功能.我们采用了外部数据库作为他的catlog资源库,对其进行热加载

　(1)添加restful API请求接口.

　　为了使框架本身具有添加catalog的功能,需要使其本身具有Api访问接口的方式来来对catalog的资源进行调整的功能

1.新增CatalogResource类来实现api的请求接口
　2.新增TimiCatalogStoreConfig类来实现与数据库交互的持久层
　　3.新增TimiCatalogStore类来替换原本的catlog加载类
　　　4.新增CatalogInfo类来实现对catalog Model信息的解析

#1 CatalogResource

1 2	@Path("/presto/catalog") public class CatalogResource{<br><br>

@GET
@Path("test")
public Response test()
{
    return Response.ok("Hello world").build();
}

1

}

　　在ServerMainModule类中setup方法,最后一行添加jaxrsBinder(binder).bind(CatalogResource.class);将添加的请求类添加进来,然后启动主服务,并确认所开启的presto的请求接口地址,默认端口是:8080请求http://localhost:8080/presto/catalog/test

　返回 "Hello world" 则表示restful API 接口添加成功.

#2 TimiCatalogStoreConfig 类中主要实现了读取数据库连接配置,以及具体执行的catalog执行动作,并使用jaxrsBinder(binder).bind(TimiCatalogStoreConfig.class);注入到项目启动的容器中.并将Announcer,disabledCatalogs,ConnectorManager注入到类中.具体实现

1 2 3 4 5 6 7 8 9 10 11 12

public class TimiCatalogStoreConfig {       private final Announcer announcer;     private static final Logger log = Logger.get(TimiCatalogStoreConfig.class);     private final Set<String> disabledCatalogs;     private final ConnectorManager connectorManager;<br> public TimiCatalogStoreConfig(Announcer announcer,Set<String> disabledCatalogs,ConnectorManager connectorManager ) {         this.announcer = announcer;         this.disabledCatalogs = ImmutableSet.copyOf(disabledCatalogs);         this.connectorManager = connectorManager;     } }

　　然后就是实现对catlog增删查改动作,并将操作的结构实现到ConnectorManager中,

首先将Server中的CatalogStore替换成我们自定义实现的TimiCatalogStore并注入相关类

1 2 3 4 5 6 7 8 9

@Inject    public TimiCatalogStore(ConnectorManager connectorManager, Announcer announcer, StaticCatalogStoreConfig config,TimiCatalogStoreConfig catalogStoreConfig) {        this(connectorManager,                announcer,                config.getCatalogConfigurationDir(),                firstNonNull(config.getDisabledCatalogs(), ImmutableList.of()),                catalogStoreConfig        );    }

　　然后实现loadCatalogs方法,首次调用的时候使用load();方法加载mysql中存储的所有catlog,然后使用ScheduledExecutorService定时的方式从mysql中提取有变化的catlog加载到presto的ConnectorManager中.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

public static void updateConnectorIdAnnouncement(Announcer announcer, CatalogName connectorId)     {         //         // This code was copied from PrestoServer, and is a hack that should be removed when the connectorId property is removed         //         // get existing announcement         ServiceAnnouncement announcement = getPrestoAnnouncement(announcer.getServiceAnnouncements());         // update connectorIds property         Map<String, String> properties = new LinkedHashMap<>(announcement.getProperties());         String property = nullToEmpty(properties.get("connectorIds"));         Set<String> connectorIds = new LinkedHashSet<>(Splitter.on(',').trimResults().omitEmptyStrings().splitToList(property));         connectorIds.add(connectorId.toString());         properties.put("connectorIds", Joiner.on(',').join(connectorIds));         // update announcement         announcer.removeServiceAnnouncement(announcement.getId());         announcer.addServiceAnnouncement(serviceAnnouncement(announcement.getType()).addProperties(properties).build());         announcer.forceAnnounce();     }

　　在这里我们设定的1分钟从mysql库充更新一次catalog列表

1 2 3 4 5 6

scheduledExecutorService.scheduleWithFixedDelay(new Runnable() {     @Override     public void run() {         reload();     } }, 60, 60, TimeUnit.SECONDS);

　　调用reload方法定时读取

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

public void reload() {           try{             //获取最新的catalogs             Map<String, CatalogInfo> catalogInfos = catalogStoreConfig.load();               catalogInfos.forEach(                     (key, catalogInfo) -> {                         if (!catalogInfoMap.containsKey(key)) {                             //相同--catlog                             try {                                 System.out.println("添加数据源"+JSON.toJSONString(catalogInfo)); //                                log.info("添加数据源:{}",JSON.toJSONString(catalogInfos.get(key)));                                 CatalogName catalogName = loadCatalog(catalogInfo);                                 updateConnectorIdAnnouncement(announcer,catalogName);                             } catch (Exception e) {                                 e.printStackTrace();                             }                           } else {                             //不同catlog                             if (!JSON.toJSONString(catalogInfoMap.get(key)).equals(JSON.toJSONString(catalogInfo))){                                 connectorManager.dropConnection(catalogInfo.getCatalogName());                                 try {                                     System.out.println("添加数据源"+JSON.toJSONString(catalogInfo));                                     CatalogName catalogName  = loadCatalog(catalogInfo);                                     updateConnectorIdAnnouncement(announcer,catalogName);                                 } catch (Exception e) {                                     e.printStackTrace();                                 }                             }                           }                     }             );             catalogInfoMap.putAll(catalogInfos);           }catch (Exception e){             e.printStackTrace();         }     }

　　从mysql库中取出来的catlog信息和对现有的catlog进行对比,如果是不同的catlog就添加到pres