基于Flink集成Aviator与AlertManager构建动态规则告警系统-开发者社区-阿里云

Flink实时告警系统设计与开发

背景

实时监控系统需要满足对多种来源的数据进行告警，为提升系统的可扩展行和灵活性，采用动态规则配置来实现多种数据源、多种告警规则的实时告警。需要实时监测和发现车端云端的信号、埋点数据是否有异常，车辆运行状况异常。

1、数据来源

2、系统架构设计

1、系统分层架构设计

本着高内聚低耦合的原则，实时告警系统采用分层设计的思想对整体的功能模块进行组合，其中：
1、Flink DataStream 层的功能是数据流在Flink内部的整体流向DAG图，如addSource、connect、process、addSink；
2、Flink Function 层的功能是对function的具体实现，如AlertManagerSinkFunction、CustomMysqlSourceFunction、RuleMatchBroadCastProcessFunction等；
3、Service 层是业务的处理过程，如负责向AlertManager传输数据的AlertManagerService、负责规则同步、更新、维护、转化、匹配的 RulesService。

2、业务模块设计

说明：业务上，需要告警的数据源目前有4中数据来源，分别是远端日志、云端微服务日志、车机端埋点、Sentry异常奔溃，其中Sentry 中的数据需要通过告警规则的筛选后发送到kafka中用于实时监控。设计上首先通过Driver中的class 路由到通用JSON告警模块或者Sentry异常奔溃业务处理模块，其次通过app.type 选择kafka中的数据源。

3、Flink DataStream 处理流程图

说明：DataStream 处理流程图展示的是数据从Kafka消费后再Flink Function 中的流向关系，Driver 负责Flink程序的启动，通过class筛选路由到通用JSON告警或者Sentry异常崩溃模块，其中内部的逻辑比较相似：
1、首先Mysql中的配置通过自定义数据源模块会被解析成配置流；
2、其次kafka topic 会被解析成数据流，通过广播连接，配置流会被广播到每个数据流的TaskManager；
3、通过规则匹配模块对数据流和规则流进行匹配；
4、匹配到数据筛选出非Sentry中的数据分别发送到AlertManager实时告警、MySQL告警统计、kafka 实时监控

4、规则引擎使用

Aviator是一个高性能、轻量级的java语言实现的表达式求值引擎，主要用于各种表达式的动态求值。Aviator是直接将表达式编译成Java字节码，交给 JVM去执行。

规则匹配模块核心使用的是Avaitor规则引擎表达式进行规则匹配，匹配的内容来源于：
1、数据流的JSON通过flattenAsMap转成map；
2、规则流中有效的Rule中获取得到的规则表达式。

5、规则设计

规则存储在MySQL中便于管理和修改，通过flink CDC可实现动态修改和同步。

CREATE TABLE  IF NOT EXISTS `flink`.`flink-alert-rule123` (
     `id` int(16) NOT NULL AUTO_INCREMENT COMMENT '主键id',
     `job_id` int(16) NOT NULL  COMMENT '需求id',
     `rule_keys`  varchar(255) DEFAULT NULL COMMENT '规则相关的key',
     `name` varchar(255) DEFAULT NULL COMMENT '规则名称',
     `exp` varchar(1020) DEFAULT NULL COMMENT '规则表达式',
     `update_time` datetime NOT NULL ON UPDATE CURRENT_TIMESTAMP COMMENT '更新规则时间',
     `create_time` datetime NOT NULL ON UPDATE CURRENT_TIMESTAMP COMMENT '创建规则时间',
     `is_valid` int(1) NOT NULL COMMENT '规则是否有效，无效不会告警',
     `app_type` varchar(255) NOT NULL COMMENT '规则适用的应用类型，值必须是AppType的枚举值，忽略大小写',
     PRIMARY KEY (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=9 DEFAULT CHARSET=utf8;

json规则的示例：

{
    "common_data":{
        "appPackage":"ltd.qisi.sotasupportapp",
        "appVersion":"3.03.01.000",
        "collectedTime":1625240289781,
        "behaviorId":"50026003",
        "qisiuiVersion":"0.2.02",
        "uid":"1924427992",
        "regionCode":"659001",
        "eventName":"mock",
        "vin":"MOCK1TELWMOMZRQAWO",
        "hardwareVersion":"3.03.01.000",
        "carseries":"E115",
        "pdsn":"47556519116431",
        "displayId":"0"
    },
    "gather_data":{
        "key1":"value5",
        "key2":"69",
        "key3":"0"
    }
}

告警规则 common_data.appPackage == 'ltd.qisi.sotasupportapp' 表示common_data.appPackage字段等于'td.qisi.sotasupportapp

6、输出业务告警数据格式设计1、车机端告警统计格式

CREATE TABLE `flink`.`flink-alert-data` (
     `app_package`          varchar(255)    comment 'app包名'
    ,`collected_time`       bigint(16)      comment '数据'
    ,`behavior_id`          varchar(255)    comment ''
    ,`qisiui_version`       varchar(255)    comment ''
    ,`uid`                  varchar(255)    comment '用户id'
    ,`region_code`          varchar(255)    comment ''
    ,`os_version`           varchar(255)    comment ''
    ,`event_name`           varchar(255)    comment ''
    ,`vin`                  varchar(255)    comment 'vin码'
    ,`hardware_version`     varchar(255)    comment ''
    ,`carseries`            varchar(255)    comment ''
    ,`pdsn`                 varchar(255)    comment ''
    ,`display_id`           varchar(255)    comment '屏幕id[ 0：主控屏；1：副驾屏；2：左后排屏；3：右后排屏；-1：未知]'
    ,`rule_name`            varchar(255)    comment ''
    ,`rule_id`              varchar(255)    comment ''
    ,`rule_exp`             varchar(255)    comment ''
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

2、云端告警统计格式

CREATE TABLE `flink`.`flink-cloud-alert-data` (
     `microservice`         varchar(255)    comment ''
    ,`reqPath`              varchar(255)    comment ''
    ,`clicnetIP`            varchar(255)    comment ''
    ,`resultCode`           varchar(255)    comment ''
    ,`createDate`           varchar(255)    comment ''
    ,`ctime`                varchar(255)    comment ''
    ,`rule_name`            varchar(255)    comment ''
    ,`rule_id`              varchar(255)    comment ''
    ,`rule_exp`             varchar(255)    comment ''
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

7、AlertManager告警模块设计对接AlertManager的模块为ISendService的实现类，通过调用void send(AlertManagerData data)方法把数据发送的出去，其其实类会调用AlertManager的post请求发送json数据，请求的基本格式为：

curl -XPOST http://localhost:9093/api/v1/alerts -d '
[
  {
    "labels": {
       "alertname": "DiskRunningFull",
       "dev": "sda1",
       "instance": "中文测试",
       "route": "WEBHOOK"
     },
     "annotations": {
        "info": "The disk sda1 is running full",
        "summary": "please check the instance example1"
      },
     "Source": {
        "link": "http://www.baidu.com"
      }
  }
]
'

8、核心代码部分

package com.qisi.driver;
import com.qisi.functions.AlertManagerSinkFunction;
import com.qisi.functions.RuleMatchBroadCastProcessFunction;
import com.qisi.functions.SinkFunctions;
import com.qisi.functions.SourceFunctions;
import com.qisi.pojo.KafkaAlertLog;
import com.qisi.pojo.Rule;
import com.qisi.utils.Config;
import org.apache.flink.api.common.restartstrategy.RestartStrategies;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.runtime.state.filesystem.FsStateBackend;
import org.apache.flink.streaming.api.datastream.ConnectedStreams;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
/**
 * 车机端埋点日志告警，用 {@link Driver}调用
 */
public class JsonAlertJobDriver {
    public static void main(String[] args) throws Exception {
        // 全局配置
        Configuration config = Config.initApplicationConfig(args, JsonAlertJobDriver.class);
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment(config);
        env.setParallelism(2);
        env.setRestartStrategy(RestartStrategies.fixedDelayRestart(4,3000L));
        //env.setStateBackend(new FsStateBackend("hdfs:///flink/checkpoint/cp/flink-real-time-alert/cp"));
        env.setStateBackend(new FsStateBackend("file:///qisi/gitlab/flink-real-time-alert/cp"));
        env.getCheckpointConfig().setTolerableCheckpointFailureNumber(3);
        env.getCheckpointConfig().disableCheckpointing();
        // 捕获 MySQL 中的规则流
        SingleOutputStreamOperator<Rule> mysqlRuleStream = env.addSource(SourceFunctions.getCustomMysqlSource(config)).name("mysqlRuleStream");
        // 获取 Kafka 日志流
        SingleOutputStreamOperator<String> kafkaSource = env.addSource(SourceFunctions.getKafkaEventSource()).name("kafkaSourceStream");
        // 连接 日志流 & 规则广播流
        ConnectedStreams<String, Rule> connectStream = kafkaSource.connect(mysqlRuleStream.broadcast());
        // 匹配出告警的 LogMessage
        SingleOutputStreamOperator<KafkaAlertLog> alertStream = connectStream.process(new RuleMatchBroadCastProcessFunction(config));
        // 发送告警信息到 AlertManager
        alertStream.addSink(new AlertManagerSinkFunction(config));
        // 告警服务发送到 Mysql 满足告警统计的需求
        alertStream.addSink(SinkFunctions.getMysqlAlertSinkFunction(config));
        // 告警结果发送到 Kafka 满足监控的需求
        alertStream.filter(KafkaAlertLog::isSentryAlertData).map(KafkaAlertLog::getJson).addSink(SinkFunctions.getFlinkKafkaProducer());
        alertStream.print();
        env.execute(JsonAlertJobDriver.class.getSimpleName());
    }
}

RulesService （管理规则的服务）

package com.qisi.services;
import com.alibaba.fastjson.JSONObject;
import com.googlecode.aviator.AviatorEvaluator;
import com.googlecode.aviator.AviatorEvaluatorInstance;
import com.googlecode.aviator.Options;
import com.qisi.pojo.AppType;
import com.qisi.pojo.Rule;
import com.qisi.utils.BeanToMap;
import com.qisi.utils.MysqlConnectionUtil;
import lombok.Data;
import lombok.extern.slf4j.Slf4j;
import org.apache.flink.configuration.ConfigOption;
import org.apache.flink.configuration.ConfigOptions;
import org.apache.flink.configuration.Configuration;
import java.io.Serializable;
import java.sql.SQLException;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
/**
 * 管理规则的服务，负责规则的更新、维护
 */
@Slf4j
@Data
public class RulesService implements Serializable{
    private AppType appType;
    private String mysqlRuleSql;
    private String mysqlUrl;
    private String url;
    private String user;
    private String pswd;
    /**
     * 保存全量的规则
     */
    private HashMap<Integer, Rule> rules = new HashMap<>();
    public RulesService(Configuration config) {
        ConfigOption<AppType> appTypeOption = ConfigOptions.key("app.type").enumType(AppType.class).noDefaultValue();
        ConfigOption<String> ruleSqlOption = ConfigOptions.key("rule.sql").stringType().noDefaultValue();
        ConfigOption<String> uelOption = ConfigOptions.key("mysql.url").stringType().noDefaultValue();
        appType =  config.get(appTypeOption);
        mysqlRuleSql = config.get(ruleSqlOption);
        mysqlUrl = config.get(uelOption);
        ConfigOption<String> urlOption = ConfigOptions.key("mysql.url").stringType().noDefaultValue();
        ConfigOption<String> userOption = ConfigOptions.key("user").stringType().noDefaultValue();
        ConfigOption<String> pswdOption = ConfigOptions.key("pswd").stringType().noDefaultValue();
        url = config.get(urlOption);
        user = config.get(userOption);
        pswd = config.get(pswdOption);
    }
    /**
     * 全量同步 mysql 的规则到{@link #rules} 适用于：
     *      1、增量同步第一次启动时
     *      2、全量同步
     */
    public void allSyncRule() {
        try {
            MysqlConnectionUtil connUtil = new MysqlConnectionUtil(url, user, pswd);
            ArrayList<Rule> rules = connUtil.onceSelectToJavaBean(mysqlRuleSql, Rule.class);
            for (Rule rule : rules) {
                if(isValid(rule)) {
                    this.rules.put(rule.getId(), rule);
                }
            }
        } catch (SQLException e) {
            log.error("all sync mysql rule error, url: {}, sql: {}", mysqlUrl, mysqlRuleSql);
            e.printStackTrace();
        }
    }
    /**
     * 增量同步规则到 this.rules
     * @param value : binlog 内容
     */
    public void incrementSyncRule(String value) {
        Rule rule = RulesService.toRule(value);
        if(isValid(rule)) {
            rules.put(rule.getId(), rule);
        }
    }
    /**
     * 增量同步规则到 this.rules
     * @param rule : Rule 对象
     */
    public void incrementSyncRule(Rule rule) {
        // 更新规则库
        if(isValid(rule)) {
            rules.put(rule.getId(), rule);
        }
    }
    private boolean isValid(Rule rule) {
        return rule.getIsValid() == 1 && appType.name().equalsIgnoreCase(rule.getAppType());
    }
    private static Rule toRule(String binlogJson){
        try {
            // 获取更新后的变化内容
            String after = JSONObject.parseObject(binlogJson).getJSONObject("after").toJSONString();
            // 解析成 Rule 类型
            return JSONObject.parseObject(after, Rule.class);
        } catch (Exception e) {
            e.printStackTrace();
            log.error("parse MySQL rules error：{}", binlogJson);
        }
        return new Rule(0);
    }
    public List<Rule> mathRules(String json) {
        List<Rule> res = new ArrayList<>();
        Map<String, Object> map = BeanToMap.jsonFlatten(json);
        for (Map.Entry<Integer, Rule> entry : this.getRules().entrySet()) {
            try {
                Rule rule = entry.getValue();
                AviatorEvaluatorInstance instance = AviatorEvaluator.getInstance();
                instance.setOption(Options.ENABLE_PROPERTY_SYNTAX_SUGAR, false);
                Object execute = AviatorEvaluator.execute(rule.getExp(), map);
                if (execute instanceof Boolean) {
                    boolean match = (boolean) execute;
                    if(match){
                        // 触发告警
                        log.info("Send Alert Data to AlertManager: | {} | <--> | {} |", rule, json);
                        res.add(rule);
                    }else{
                        log.debug("No Matched： | {} | -- | {} |", rule, json);
                    }
                }else{
                    log.error("rule set error, please modify rule:{}", rule);
                }
            } catch (Exception e) {
                log.error("AviatorEvaluator regular expression match error:| {} | -- | {} |",  this.getRules(), json);
                e.printStackTrace();
            }
        }
        return res;
    }
}

Flink CDC

private static SourceFunction<String> getBinlogSourceFunction(String table) {
        Properties properties = new Properties();
        properties.put("offset.flush.interval.ms", "10000");
        properties.put("offset.flush.timeout.ms", "40000");
        properties.put("autocommit", "false");
        return MySQLSource.<String>builder()
                .hostname(Config.MYSQL_HOST)
                .port(Config.MYSQL_PORT)
                .databaseList(Config.MYSQL_DB)
                .tableList(table)
                //.startupOptions(StartupOptions.latest())
                .username(Config.MYSQL_USER)
                .password(Config.MYSQL_PSWD)
                .deserializer(new MysqlCDCDeserializationSchema())
                .debeziumProperties(properties)
                .build();
    }

规则匹配核心类

package com.qisi.functions;
import com.qisi.pojo.KafkaAlertLog;
import com.qisi.pojo.Rule;
import com.qisi.pojo.alertManager.Route;
import com.qisi.services.RulesService;
import lombok.Data;
import lombok.extern.slf4j.Slf4j;
import org.apache.flink.configuration.ConfigOption;
import org.apache.flink.configuration.ConfigOptions;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.co.CoProcessFunction;
import org.apache.flink.util.Collector;
import java.sql.SQLException;
import java.util.List;
/**
 * 规则匹配核心类
 */
@Slf4j
@Data
public class RuleMatchBroadCastProcessFunction extends CoProcessFunction<String, Rule, KafkaAlertLog> {
    private RulesService rulesService;
    private Route route;
    private static final String alertName = "Sentry Alert";
    public RuleMatchBroadCastProcessFunction(Configuration config) throws  SQLException {
        ConfigOption<Route> routeOption = ConfigOptions.key("am.route").enumType(Route.class).defaultValue(Route.WEBHOOK);
        route = config.get(routeOption);
        rulesService = new RulesService(config);
        // rulesService.allSyncRule();
        log.info("init load Mysql rule completed, total count: {} ", rulesService.getRules().size());
    }
    @Override
    public void open(Configuration parameters) throws Exception {
        super.open(parameters);
    }
    @Override
    public void processElement1(String value, Context ctx, Collector<KafkaAlertLog> out) {
        log.debug("accept a per dada: {}", value);
        List<Rule> rules = rulesService.mathRules(value);
        for (Rule rule : rules) {
            out.collect(new KafkaAlertLog(value, rule));
        }
    }
    @Override
    public void processElement2(Rule value, Context ctx, Collector<KafkaAlertLog> out) {
        log.debug("rule refresh ：{}", value);
        rulesService.incrementSyncRule(value);
    }
}

AlertMessageFilterFunction

package com.qisi.functions;
import com.googlecode.aviator.AviatorEvaluator;
import com.qisi.pojo.LogMessage;
import com.qisi.pojo.Rule;
import com.qisi.services.RulesService;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.ProcessFunction;
import org.apache.flink.util.Collector;
import java.sql.SQLException;
import java.util.HashMap;
import java.util.Map;
/**
 * RulesService 周期性查询 MySQL 更新规则，筛选出明细 LogMessage
 */
@Deprecated
public class AlertMessageFilterFunction extends ProcessFunction<LogMessage, LogMessage> {
    private RulesService rulesService;
    private HashMap<Integer, Rule> rules;
    private long time = 0;
    private long period;
    public AlertMessageFilterFunction(long period) {
        this.period = period;
    }
    @Override
    public void open(Configuration parameters) throws SQLException {
        rulesService = new RulesService(parameters);
        rules = rulesService.getRules();
    }
    @Override
    public void processElement(LogMessage value, Context ctx, Collector<LogMessage> out) throws Exception {
        long current = ctx.timerService().currentProcessingTime();
        long period = current - time;
        if (period > this.period) {
            rulesService.incrementSyncRule(value.toString());
            time = current;
        }
        for (Map.Entry<Integer, Rule> entry : rules.entrySet()) {
            String ruleExp = entry.getValue().getExp();
            if((boolean) AviatorEvaluator.execute(ruleExp, value.toMap())){
                out.collect(value);
            }
        }
    }
    @Override
    public void close() throws Exception {
        super.close();
    }
}

发送告警

package com.qisi.services;
import com.alibaba.fastjson.JSONArray;
import com.alibaba.fastjson.JSONObject;
import com.qisi.pojo.alertManager.AlertManagerData;
import lombok.extern.slf4j.Slf4j;
import org.apache.http.HttpResponse;
import org.apache.http.client.methods.CloseableHttpResponse;
import org.apache.http.client.methods.HttpPost;
import org.apache.http.concurrent.FutureCallback;
import org.apache.http.entity.StringEntity;
import org.apache.http.impl.client.CloseableHttpClient;
import org.apache.http.impl.client.HttpClientBuilder;
import org.apache.http.impl.nio.client.CloseableHttpAsyncClient;
import org.apache.http.impl.nio.client.HttpAsyncClientBuilder;
import java.io.IOException;
import java.io.Serializable;
import java.nio.charset.StandardCharsets;
import java.util.concurrent.Future;
/**
 * AlertManager 告警服务，http 发送 AlertManagerData
 */
@Slf4j
public class AlertManagerService implements ISendService, Serializable {
    private static CloseableHttpClient httpClient = HttpClientBuilder.create().build();
    //private CloseableHttpAsyncClient asyncHttpClient = HttpAsyncClientBuilder.create().build();
    private String uri;
    public AlertManagerService(String amHost, int amPort){
        uri = String.format("http://%s:%d/api/v1/alerts", amHost, amPort);
        // asyncHttpClient.start();
    }
    public void send(AlertManagerData data) {
        JSONArray arr = new JSONArray();
        int statusCode = 0;
        try {
            arr.add(data);
            log.info("uri:{}", uri);
            HttpPost httpPost = new HttpPost(uri);
            StringEntity se = new StringEntity(JSONObject.toJSONString(arr), StandardCharsets.UTF_8);
            se.setContentEncoding("utf8");
            se.setContentType("application/json");
            httpPost.addHeader("Content-type","application/json; charset=utf-8");
            httpPost.setHeader("Accept", "application/json");
            httpPost.setEntity(se);
            log.debug("http client execute starting ...");
            CloseableHttpResponse response = httpClient.execute(httpPost);
            statusCode = response.getStatusLine().getStatusCode();
            // Future<HttpResponse> execute = asyncHttpClient.execute(httpPost, new Callbak(arr));
            // statusCode = execute.get().getStatusLine().getStatusCode();
            log.info("Code: {}，Send Alert Data Succeed: {}", statusCode, arr.toString());
        }catch (Exception e){
            log.error("Code: {}, Send Alert Data Failed: {}", statusCode, arr.toString());
            e.printStackTrace();
        }
    }
    public void close() throws IOException {
        httpClient.close();
    }
    public static class Callbak implements FutureCallback<HttpResponse> {
        private JSONArray arr;
        private Callbak(JSONArray arr) {
            this.arr = arr;
        }
        @Override
        public void completed(HttpResponse httpResponse) {
            int statusCode = httpResponse.getStatusLine().getStatusCode();
            if (statusCode == 200) {
                log.info("Code: {}，Send Alert Data Succeed: {}", statusCode, arr.toString());
            }else{
                log.error("Code: {}，Send Alert Data Succeed: {}", statusCode, arr.toString());
            }
        }
        @Override
        public void failed(Exception e) {
            log.error(e.getMessage());
            e.printStackTrace();
        }
        @Override
        public void cancelled() {
            log.info("Callbak cancelled");
        }
    }
}

Flink实时告警系统开发

热门文章

最新文章

相关课程

相关电子书

相关实验场景

探索云世界

热门

云计算

大数据

云原生

人工智能

数据库

开发与运维

活动广场

任务中心

训练营

直播

乘风者计划

下载

镜像站