您现在的位置是:首页 >技术杂谈 >pyflink kafka es网站首页技术杂谈

pyflink kafka es

scan724 2024-10-13 12:01:04
简介pyflink kafka es

# -*- coding: utf-8 -*-
from pyflink.datastream import StreamExecutionEnvironment
from pyflink.datastream.functions import  MapFunction, RuntimeContext, KeyedProcessFunction
from abc import ABC, abstractmethod
from pyflink.datastream import StreamExecutionEnvironment
from pyflink.datastream.functions import  MapFunction, RuntimeContext, KeyedProcessFunction
from pyflink.datastream.state import MapStateDescriptor
from pyflink.datastream.connectors.kafka import FlinkKafkaConsumer
from pyflink.common.typeinfo import Types, TypeInformation
from pyflink.datastream.connectors.elasticsearch import Elasticsearch7SinkBuilder, ElasticsearchEmitter, FlushBackoffType
from pyflink.datastream.connectors import DeliveryGuarantee
from pyflink.common.serialization import SimpleStringSchema
import json
import re
from datetime import datetime
from elasticsearch import Elasticsearch
from pyflink.datastream.functions import RuntimeContext, FlatMapFunction
from pyflink.common.typeinfo import Types
from pyflink.datastream import StreamExecutionEnvironment
from pyflink.datastream.connectors.kafka import FlinkKafkaConsumer
from pyflink.common.serialization import SimpleStringSchema


import re
import redis


# 创建 StreamExecutionEnvironment 对象
env = StreamExecutionEnvironment.get_execution_environment()
env.set_parallelism(1)
env.add_jars("file:///root/flink-sql-connector-kafka_2.11-1.14.4.jar")

TEST_KAFKA_SERVERS = "127.0.0.1:9092"
TEST_KAFKA_TOPIC = "topic_elink"
TEST_GROUP_ID = "pyflink_group"
def get_kafka_customer_properties(kafka_servers: str, group_id: str):
    properties = {
        "bootstrap.servers": kafka_servers,
        "fetch.max.bytes": "67108864",
        "key.deserializer": "org.apache.kafka.common.serialization.StringDeserializer",
        "value.deserializer": "org.apache.kafka.common.serialization.StringDeserializer",
        "enable.auto.commit": "false",  # 关闭kafka 自动提交,此处不能传bool 类型会报错
        "group.id": group_id,
    }
    return properties
properties = get_kafka_customer_properties(TEST_KAFKA_SERVERS, TEST_GROUP_ID)


class LogEvent:
    buss_seq = None
    message = None
    index_name = None
    
    def __init__(self, bus_seq,message,index_name):
        self.bus_seq = bus_seq
        self.message = message
        self.index_name= index_name

    def to_dict(self):
        return {
            "bus_seq": self.bus_seq,
            "message": self.message,
            "index_name" : self.index_name
        }
   
class MyMapFunction(FlatMapFunction):
   def open(self, runtime_context: RuntimeContext):
       self.process_id_to_bus_seq = runtime_context.get_map_state(MapStateDescriptor('process_id', Types.STRING(), Types.STRING()))
      
   def close(self):
       pass

   def flat_map(self,line):
      bus_seq=''
      process_id=''
      message=''
      message = line.replace(" ", "")
      line = json.loads(message)['message']
      if not line.startswith("ES"):
          return 
      if '<Serial>' in line:
         try:
             pat=re.compile(r"<Serial>(d+)</Serial>")
             bus_seq=pat.search(line).group(1)
             process_id=line.split()[1]
             self.process_id_to_bus_seq.put(process_id, bus_seq)
         except:
             return 
      process_id=line.split()[1]
      if not len(process_id)==6 :
          process_id=line.split()[2]
      bus_seq=self.process_id_to_bus_seq.get(process_id)
      if not bus_seq:
          bus_seq='0'
      #self.r.delete(process_id)
      #log_event = LogEvent(bus_seq.decode('UTF-8'),line)
      #LogEvent['bus_seq']=bus_seq.decode('UTF-8')
      date_str = datetime.now().strftime("%Y-%m-%d")
      index_name='flink-test'+date_str
      try:
         log_event=LogEvent(bus_seq,line,index_name)
      except:
          return 
      yield log_event.to_dict()
    
  
data_stream = env.add_source(
        FlinkKafkaConsumer(topics=TEST_KAFKA_TOPIC,
                           properties=properties,
                           deserialization_schema=SimpleStringSchema())
            .set_commit_offsets_on_checkpoints(True)
            .set_start_from_latest()
    ).name(f"消费{TEST_KAFKA_TOPIC}主题数据")


env.add_jars("file:///root/lib/flink-sql-connector-elasticsearch7-3.0.1-1.16.jar")

es7_sink = Elasticsearch7SinkBuilder()
    .set_bulk_flush_max_actions(1)
    .set_emitter(ElasticsearchEmitter.dynamic_index('index_name'))
    .set_hosts(['127.0.0.1:9200'])
    .build()
def get_line_key(line):
          message=''
          message = line.replace(" ", "")
          line = json.loads(message)['message']
          try:
              process_id=line.split()[1]
              if not len(process_id)==6 :
                  process_id=line.split()[2]
          except:
              process_id='9999'
          return    process_id 
#data_stream.key_by(get_line_key).flat_map(MyMapFunction(),output_type=Types.MAP(Types.STRING(),Types.STRING())).sink_to(es7_sink)
data_stream.key_by(get_line_key).flat_map(MyMapFunction(),output_type=Types.MAP(Types.STRING(),Types.STRING())).sink_to(es7_sink)


# 执行任务
env.execute('Add "bus_seq" to each line')

风语者!平时喜欢研究各种技术,目前在从事后端开发工作,热爱生活、热爱工作。