用fastjson解析streamsets接口返回的复杂json

水深无声 2021-09-29 01:14 318阅读 0赞

Streamsets提供了丰富的restful接口，通过swagger方式提供，使用很方便，功能很强大。

地址： [http://localhost:18630/collector/restapi][http_localhost_18630_collector_restapi]

这里我们在对Streamsets做二次开发的时候，需要解析返回结果并保存到数据库中，这里把几个有代表性的json解析方式讲解一下。

### 1、查询所有任务的运行状态 ###

在manager分组下， 不需要参数，直接通过GET  /rest/v1/pipelines/status

返回结果如下：

{
      "addNewPipeline": {
        "pipelineId": "addNewPipeline",
        "rev": "0",
        "user": "admin",
        "status": "EDITED",
        "message": "Pipeline edited",
        "timeStamp": 1556437391339,
        "attributes": {
          "IS_REMOTE_PIPELINE": false
        },
        "executionMode": "STANDALONE",
        "metrics": null,
        "retryAttempt": 0,
        "nextRetryTimeStamp": 0,
        "name": "addNewPipeline"
      },
      "Names2PgStage1wholecopy8396290d-163e-4ec6-9371-5f09baf29a23": {
        "pipelineId": "Names2PgStage1wholecopy8396290d-163e-4ec6-9371-5f09baf29a23",
        "rev": "0",
        "user": "admin",
        "status": "STOPPED",
        "message": "Stopped while the pipeline was in RETRY state",
        "timeStamp": 1560759731119,
        "attributes": {
          "IS_REMOTE_PIPELINE": false,
          "RUNTIME_PARAMETERS": null,
          "INTERCEPTOR_CONFIGS": []
        },
        "executionMode": "STANDALONE",
        "metrics": null,
        "retryAttempt": 0,
        "nextRetryTimeStamp": 0,
        "name": "Names2PgStage1wholecopy8396290d-163e-4ec6-9371-5f09baf29a23"
      }
    }

通过观察可以发现，这个json是有规律的，格式是固定的，所以解析并不难。

JSONObject jsonObject = JSONObject.parseObject(result);
    
            Collection<Object> list=jsonObject.values();
            Iterator<Object> it= list.iterator();
            while(it.hasNext()){
                Object obj = it.next();
                Map m=JSONObject.parseObject(obj.toString(),Map.class);
                String pipelineId=m.get("pipelineId").toString();
                Tasks tasks=new Tasks();
                tasks.setPipeline_id(pipelineId);
                List<Tasks> lists=tasksService.selectByTask(tasks);
                if(lists!=null && lists.size()>0){
                    Tasks task=lists.get(0);
                    task.setTask_status(m.get("status").toString());
                    task.setCommitted_offsets(m.get("message")==null?""
                    :m.get("message").toString());
                    tasksService.editTask(task);
                }
            }

### 2、查询所有日志 ###

在system分组下，参数可选，直接GET  /rest/v1/system/logs?endingOffset=-1

返回结果如下：

[
      {
        "timestamp": "2019-06-20 09:04:48,875",
        "s-user": "*admin",
        "s-entity": "hdp/hdpd32e5f16-4d10-439f-b5ad-4fbc29c78211",
        "s-runner": "",
        "thread": "Spool Directory Runner - 0",
        "s-stage": "",
        "severity": "DEBUG",
        "category": "ProductionSourceOffsetTracker",
        "message": "Saving offset {$com.streamsets.pipeline.stage.origin.spooldir.SpoolDirSource.offset.version$=1, SSTableExport-1.0-SNAPSHOT.jar={\"POS\":\"-1\"}} for pipeline hdpd32e5f16-4d10-439f-b5ad-4fbc29c78211"
      },
      {
        "timestamp": "2019-06-20 09:04:48,877",
        "s-user": "*admin",
        "s-entity": "hdp/hdpd32e5f16-4d10-439f-b5ad-4fbc29c78211",
        "s-runner": "",
        "thread": "Spool Directory Runner - 0",
        "s-stage": "",
        "severity": "DEBUG",
        "category": "DirectorySpooler",
        "message": "Polling for file, waiting '60000' ms"
      }
    ]

通过观察发现这个json也是有固定格式的，遍历解析就可以了

JSONArray jsonArray = JSONArray.parseArray(result);
                for (Object obj : jsonArray) {
                    JSONObject jsonObject = (JSONObject) obj;
    
                    AlertLogs alertLogs=new AlertLogs();
    
                    alertLogs.setSdc_id(sdc.getSdc_id());
                    alertLogs.setSdc_name(sdc.getSdc_name());
                    alertLogs.setSdc_ip(sdc.getSdc_ip());
                    alertLogs.setSdc_port(sdc.getSdc_port());
                    alertLogs.setLog_time(jsonObject.getDate("timestamp"));
                    alertLogs.setMessage(jsonObject.getString("message"));
                    alertLogs.setLog_level(jsonObject.getString("severity"));
                    String entity=jsonObject.getString("s-entity");
                    if(entity.contains("/")){
                        String[] arr=entity.split("/");
                        StringBuffer sb=new StringBuffer();
                        for(int i=0;i<arr.length-1;i++){
                            sb.append(arr[i]);
                        }
                        alertLogs.setPipeline_title(sb.toString());
                        alertLogs.setPipeline_id(arr[arr.length-1]);
                    }
                    alertLogsService.insert(alertLogs);
                }

### 3、查询streamsets的jmx信息 ###

这个接口是在页面中调用的，swagger页面没有列出，地址 GET  /rest/v1/system/jmx

返回结果如下，结果特别长，这里列举一部分：

{
        "beans": [
            {
                "name": "metrics:name=sdc.pipeline.testcopya2cc2ed4-bd24-441a-9b09-6a2fdbb4b082.0.pipeline.batchInputRecords.meter",
                "modelerType": "com.codahale.metrics.jmx.JmxReporter$JmxMeter",
                "RateUnit": "events/second",
                "OneMinuteRate": 1.5702124778961105e-86,
                "FiveMinuteRate": 1.5921483231199352e-14,
                "FifteenMinuteRate": 0.015958339755563485,
                "MeanRate": 6.420727951968242,
                "Count": 79884
            },
            {
                "name": "metrics:name=jvm.threads.new.count",
                "modelerType": "com.codahale.metrics.jmx.JmxReporter$JmxGauge",
                "Value": 0
            },
            {
                "name": "metrics:name=sdc.pipeline.loalToTrash.0.stage.Directory_01.errorRecords.meter",
                "modelerType": "com.codahale.metrics.jmx.JmxReporter$JmxMeter",
                "RateUnit": "events/second",
                "OneMinuteRate": 0,
                "FiveMinuteRate": 0,
                "FifteenMinuteRate": 0,
                "MeanRate": 0,
                "Count": 0
            },
            {
                "name": "metrics:name=jvm.memory.pools.Metaspace.usage",
                "modelerType": "com.codahale.metrics.jmx.JmxReporter$JmxGauge",
                "Value": 0.9130996604846318
            },
            {
                "name": "metrics:name=sdc.pipeline.loalToTrash.0.stage.Directory_01.errorRecords.histogramM5",
                "modelerType": "com.codahale.metrics.jmx.JmxReporter$JmxHistogram",
                "50thPercentile": 0,
                "SnapshotSize": 1028,
                "Max": 0,
                "Mean": 0,
                "Min": 0,
                "75thPercentile": 0,
                "95thPercentile": 0,
                "98thPercentile": 0,
                "99thPercentile": 0,
                "999thPercentile": 0,
                "StdDev": 0,
                "Count": 2749
            }
    ]

通过观察发现这个json格式不统一，没有规律，所以只能分别遍历，通过name判断是哪些指标，然后对应解析，这里列举了一部分代码：

if(name.equals(JmxEnum.OPERATING_SYSTEM)){
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.ProcessCpuLoad);
                    metricSdcJmx.setUsage(jsonObject.getDouble("ProcessCpuLoad"));
                    metricSdcJmxService.insert(metricSdcJmx);
                }else if(name.equals(JmxEnum.THREADING)){
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.ThreadCount);
                    metricSdcJmx.setVcount(jsonObject.getInteger("ThreadCount"));
                    metricSdcJmxService.insert(metricSdcJmx);
    
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.PeakThreadCount);
                    metricSdcJmx.setVcount(jsonObject.getInteger("PeakThreadCount"));
                    metricSdcJmxService.insert(metricSdcJmx);
    
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.TotalStartedThreadCount);
                    metricSdcJmx.setVcount(jsonObject.getInteger("TotalStartedThreadCount"));
                    metricSdcJmxService.insert(metricSdcJmx);
                }else if(name.equals(JmxEnum.CLASS_LOADING)){
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.LoadedClassCount);
                    metricSdcJmx.setVcount(jsonObject.getInteger("LoadedClassCount"));
                    metricSdcJmxService.insert(metricSdcJmx);
    
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.UnloadedClassCount);
                    metricSdcJmx.setVcount(jsonObject.getInteger("UnloadedClassCount"));
                    metricSdcJmxService.insert(metricSdcJmx);
    
                    metricSdcJmx.setMetric_name(JmxMetricNameEnum.TotalLoadedClassCount);
                    metricSdcJmx.setVcount(jsonObject.getInteger("TotalLoadedClassCount"));
                    metricSdcJmxService.insert(metricSdcJmx);
                }

### 4、查询指定任务的运行metric ###

在manager分组下，需要pipelineId参数，GET  /rest/v1/pipeline/\{pipelineId\}/metrics

返回结果如下：

//主要的内容
    {
      "version": "string",
      "gauges": {},
      "counters": {},
      "histograms": {},
      "meters": {},
      "timers": {}
    }
    
    //详细内容，部分
    {
      "version": "4.0.0",
      "gauges": {
        "RuntimeStatsGauge.gauge": {
          "value": {
            "batchCount": 95,
            "idleBatchCount": 0,
            "timeOfLastReceivedRecord": 1557296520561,
            "lastBatchInputRecordsCount": 0,
            "lastBatchOutputRecordsCount": 0,
            "lastBatchErrorRecordsCount": 0,
            "lastBatchErrorMessagesCount": 0,
            "totalRunners": 1,
            "availableRunners": 1
          }
        },
    。。。。
     "counters": {
        "custom.Directory_01.pending.files.0.counter": {
          "count": 0
        },
        。。
     "histograms": {
        "pipeline.errorRecordsPerBatch.histogramM5": {
          "count": 95,
          "max": 0,
          "mean": 0,
          "min": 0,
          "p50": 0,
          "p75": 0,
          "p95": 0,
          "p98": 0,
          "p99": 0,
          "p999": 0,
          "stddev": 0
        },
    。。。。
    },
      "meters": {
        "custom.Directory_01.spoolQueue.0.meter": {
          "count": 90,
          "m1_rate": 11.866331343607989,
          "m5_rate": 16.560799463327818,
          "m15_rate": 17.50688058809427,
          "m30_rate": 17.751728101390487,
          "h1_rate": 17.875433024825686,
          "h6_rate": 17.97917871834381,
          "h12_rate": 17.989586346833782,
          "h24_rate": 17.994792420114447,
          "mean_rate": 2.9818143805481996,
          "units": "events/second"
        },。。。
     "timers": {
        "pipeline.batchProcessing.timer": {
          "count": 95,
          "max": 5.009,
          "mean": 0.2708105263157895,
          "min": 0.003,
          "p50": 0.006,
          "p75": 0.009000000000000001,
          "p95": 5.005,
          "p98": 5.007160000000001,
          "p99": 5.009,
          "p999": 5.009,
          "stddev": 1.1221307457361267,
          "m15_rate": 17.512359692670998,
          "m1_rate": 11.9344832175679,
          "m5_rate": 16.576790580401955,
          "mean_rate": 3.1389838868870585,
          "duration_units": "seconds",
          "rate_units": "calls/second"
        },
    。。。
    }

通过观察发现，这个json也是不统一，没有规律，但是整体上分了几个部分，所以通过解析后，分别解析不同的部分

JSONObject gauges= jsonObject.getJSONObject("gauges");
            Iterator gauges_it= gauges.entrySet().iterator();
            while(gauges_it.hasNext()) {
                Map.Entry<String, JSONObject> entry = (Map.Entry<String, JSONObject>) gauges_it.next();
                String key = entry.getKey();
                JSONObject json = entry.getValue();
                if (key.startsWith("RuntimeStatsGauge")) {
                    JSONObject gauge=json.getJSONObject("value");
                    MetricsPipeline metricsPipeline=new MetricsPipeline();
                    metricsPipeline.setCollect_time(collect_time);
                    metricsPipeline.setPipeline_id(pipelineId);
                    metricsPipeline.setMetric_name(key+".lastBatchErrorMessagesCount");
                    metricsPipeline.setVcount(gauge.getInteger("lastBatchErrorMessagesCount"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".timeOfLastReceivedRecord");
                    metricsPipeline.setVcount(gauge.getInteger("timeOfLastReceivedRecord"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".idleBatchCount");
                    metricsPipeline.setVcount(gauge.getInteger("idleBatchCount"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".lastBatchErrorRecordsCount");
                    metricsPipeline.setVcount(gauge.getInteger("lastBatchErrorRecordsCount"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".lastBatchOutputRecordsCount");
                    metricsPipeline.setVcount(gauge.getInteger("lastBatchOutputRecordsCount"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".availableRunners");
                    metricsPipeline.setVcount(gauge.getInteger("availableRunners"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".lastBatchInputRecordsCount");
                    metricsPipeline.setVcount(gauge.getInteger("lastBatchInputRecordsCount"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".totalRunners");
                    metricsPipeline.setVcount(gauge.getInteger("totalRunners"));
                    metricsPipelineService.insert(metricsPipeline);
    
                    metricsPipeline.setMetric_name(key+".batchCount");
                    metricsPipeline.setVcount(gauge.getInteger("batchCount"));
                    metricsPipelineService.insert(metricsPipeline);
                }else if(key.startsWith("custom")){
                    JSONObject gauge=json.getJSONObject("value");
                    MetricsPipeline metricsPipeline=new MetricsPipeline();
                    metricsPipeline.setCollect_time(collect_time);
                    metricsPipeline.setPipeline_id(pipelineId);
                    metricsPipeline.setMetric_name(key);
                    metricsPipeline.setStatus(gauge.getString("Status"));
                    metricsPipeline.setCurrent_stage(gauge.getString("Current Offset"));
                    metricsPipeline.setThread_name(gauge.getString("Thread Name"));
                    metricsPipeline.setCurrent_file(gauge.getString("Current File"));
                    metricsPipelineService.insert(metricsPipeline);
                }

[http_localhost_18630_collector_restapi]: http://localhost:18631/collector/restapi