flink读取hive的数据

flink1.8 对hive 的支持不够好,造成300W的数据,居然读了2个小时,打算将程序迁移至spark。 先把代码贴上。 后发现sql不应该有where条件,去掉后速度还行。

专注于为中小企业提供成都网站设计、网站建设、外贸网站建设服务,电脑端+手机端+微信端的三站合一,更高效的管理,为中小企业南城免费做网站提供优质的服务。我们立足成都,凝聚了一批互联网行业人才,有力地推动了成百上千企业的稳健成长,帮助中小企业通过网站建设实现规模扩充和转变。

maven



            org.apache.hive
            hive-jdbc
            1.1.0
        

        
            org.apache.hadoop
            hadoop-common
            3.1.2
        

         
              jdk.tools 
              jdk.tools 
              1.8 
              system 
              ${JAVA_HOME}/lib/tools.jar 
         

java


private final static String driverName = "org.apache.hive.jdbc.HiveDriver";// jdbc驱动路径
    private final static String url = ";";// hive库地址+库名
    private final static String user = "";// 用户名
    private final static String password = "!";// 密码
    private final static String table="";
    private final static String sql = " ";

    public static void main(String[] arg) throws Exception {

        long time=System.currentTimeMillis();   
        HttpClientUtil.sendDingMessage("开始同步hive-"+table+";"+Utils.getTimeString());        
        /**
         * 初始化环境
         */
        final ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
        env.setParallelism(4);

        try {
            TypeInformation[] types = new TypeInformation[]{BasicTypeInfo.STRING_TYPE_INFO, BasicTypeInfo.STRING_TYPE_INFO};
            String[] colName = new String[]{"user","name"};     
            RowTypeInfo rowTypeInfo = new RowTypeInfo(types, colName);
            JDBCInputFormatBuilder builder = JDBCInputFormat.buildJDBCInputFormat().setDrivername(driverName)
                    .setDBUrl(url)
                    .setUsername(user).setPassword(password);

            Calendar calendar = Calendar.getInstance();
            calendar.setTime(new Date());
            calendar.add(Calendar.DATE, -1); //用昨天产出的数据
            SimpleDateFormat sj = new SimpleDateFormat("yyyyMMdd");
            String d=sj.format(calendar.getTime());

            JDBCInputFormat jdbcInputFormat = builder.setQuery(sql+" and dt='"+d+"' limit 100000000").setRowTypeInfo(rowTypeInfo).finish();
            DataSource rowlist = env.createInput(jdbcInputFormat);

            DataSet temp= rowlist.filter(new FilterFunction(){

                @Override
                public boolean filter(Row row) throws Exception {
                    String key=row.getField(0).toString();
                    String value=row.getField(1).toString();
                    if(key.length()<5 || key.startsWith("-") || key.startsWith("$") || value.length()<5 || value.startsWith("-") || value.startsWith("$")) {
                        return false;
                    }else {
                        return true;
                    }
                }

            }).map(new MapFunction(){

                @Override
                public RedisDataModel map(Row value) throws Exception {
                    RedisDataModel m=new RedisDataModel();
                    m.setExpire(-1); 
                    m.setKey(JobConstants.REDIS_FLINK_IMEI_USER+value.getField(0).toString());      
                    m.setGlobal(true);
                    m.setValue(value.getField(1).toString());
                    return m;
                } 

            });

            HttpClientUtil.sendDingMessage("同步hive-"+table+"完成;开始推送模型,共有"+temp.count()+"条;"+Utils.getTimeString()); 

            RedisOutputFormat redisOutput = RedisOutputFormat.buildRedisOutputFormat()
                    .setHostMaster(AppConfig.getProperty(JobConstants.REDIS_HOST_MASTER))
                    .setHostSentinel(AppConfig.getProperty(JobConstants.REDIS_HOST_SENTINELS))
                    .setMaxIdle(Integer.parseInt(AppConfig.getProperty(JobConstants.REDIS_MAXIDLE)))
                    .setMaxTotal(Integer.parseInt(AppConfig.getProperty(JobConstants.REDIS_MAXTOTAL))) 
                    .setMaxWaitMillis(Integer.parseInt(AppConfig.getProperty(JobConstants.REDIS_MAXWAITMILLIS)))
                    .setTestOnBorrow(Boolean.parseBoolean(AppConfig.getProperty(JobConstants.REDIS_TESTONBORROW)))
                    .finish();   
            temp.output(redisOutput);               
            env.execute("hive-"+table+" sync");

            HttpClientUtil.sendDingMessage("同步hive-"+table+"完成,耗时:"+(System.currentTimeMillis()-time)/1000+"s"); 
        } catch (Exception e) {
            logger.error("",e); 
            HttpClientUtil.sendDingMessage("同步hive-"+table+"失败,时间戳:"+time+",原因:"+e.toString());
        } 

文章标题:flink读取hive的数据
文章地址:http://ybzwz.com/article/pdsgjh.html