在Logstash中解析JSON事件

Question

I have log in following format, it is a plain json with nested fields.

{
    "level": "info",
    "message": {
        "req": {
            "headers": {
                "host": "localhost:8080",
                "connection": "keep-alive",
                "x-forwarded-for": "192.168.1.1, 1.1.1.1",
                "x-forwarded-proto": "http"
            },
            "url": "/products?userId=493d0aec-a9a7-42a3",
            "method": "GET",
            "originalUrl": "/products?userId=493d0aec-a9a7-42a3",
            "params": {
                "0": "/products"
            },
            "query": {
                "userId": "493d0aec-a9a7-42a3"
            },
            "body": ""
        },
        "res": {
            "headers": {
                "traceid": "ac586e4e924048",
                "x-correlation-id": "57d7920d-b623-48f8",
                "content-type": "application/json;charset=UTF-8",
                "content-length": "2",
                "date": "Fri, 08 Mar 2019 09:55:45 GMT",
                "connection": "close"
            },
            "statusCode": 200,
            "body": "[]"
        },
        "gateway": "internal"
    },
    "correlationId": "57d7920d-b623-48f8",
    "timestamp": "2019-03-08T09:55:45.833Z"
}

How can I parse it correctly using Filebeat and Logstash to see all json fields in Kibana as separate (parsed) fields? I have a problem with "message" field which has nested json fields. I have no problem to parse an event which has string in "message", but not json.

My attempts:

1 . I tried to tell Filebeat that it is a json with following configuration:
(and doing nothing on LS side)

filebeat.inputs:
- type: stdin
  json.keys_under_root: true
  json.add_error_key: true

The result is strange for me, because I got "message" as a string in Kibana where all : are replaced with =>

{
    "req" => {
        "originalUrl" => "/offers", "params" => {
            "0" => "/offers"
        }, "query" => {}, "body" => "", "headers" => {
            "accept-encoding" => "gzip", "user-agent" => "okhttp/3.8.1", "x-consumer-id" => "f2a6e4cd-2224-4535

Other fields outside the "message" are parsed correctly

2 . I did nothing on Filebeat side and use filter in LS:

json {   
    source => "message"    
    target => "message_json"    
} 

Logs are not appeared in Kibana at all, I got following errors in LS:

[2019-03-08T09:55:47,084][WARN ][logstash.outputs.elasticsearch] Could not index event to Elasticsearch. {:status=>400, :action=>["index", {:_id=>nil, :_index=>"filebeat-6.5.0-2019.03.08-sdx", :_type=>"doc", :routing=>nil}, #], :response=>{"index"=>{"_index"=>"filebeat-6.5.0-2019.03.08-sdx", "_type"=>"doc", " id"=>"ERS6XGkBgE -US7A6Mvt", "status"=>400, "error"=>{"type"=>"mapper_parsing_exception", "reason"=>"failed to parse field [json.message] of type [keyword]", "caused_by"=>{"type"=>"illegal_state_exception", "reason"=>"Can't get text on a START_OBJECT at 1:461"}}}}} [2019-03-08T09:55:47,085][WARN ][logstash.outputs.elasticsearch] Could not index event to Elasticsearch. {:status=>400, :action=>["index", {:_id=>nil, :_index=>"filebeat-6.5.0-2019.03.08-sdx", :_type=>"doc", :routing=>nil}, #], :response=>{"index"=>{"_index"=>"filebeat-6.5.0-2019.03.08-sdx", "_type"=>"doc", " id"=>"EhS6XGkBgE -US7A6Mvt", "status"=>400, "error"=>{"type"=>"mapper_parsing_exception", "reason"=>"failed to parse field [json.message] of type [keyword]", "caused_by"=>{"type"=>"illegal_state_exception", "reason"=>"Can't get text on a START_OBJECT at 1:461"}}}}}

This filter works fine for me if the "message" field is a string (not a json).

Any ideas how to parse nested json in "message field"?

Answer 1

I have problems to parse json by logstash to.

I was struggling with this problem for a while. And failed to solve into logstash.

But fortunately, we have ingested node in elasticsearch itself.

I want to suggest my solution to your problem:

You make pipeline (very simple pipeline):

{
    "description": "Parse JSON log",
    "processors": [
      {
        "json": {
          "field": "message",
          "target_field": "message-json"
        }
      },
      {
        "remove": {
          "field": "message"
        }
      },
      {
        "append": {
          "field": "tags",
          "value": [
            "isjsonlog"
          ]
        }
      }
    ],
    "on_failure": [
      {
        "append": {
          "field": "tags",
          "value": [
            "nonjsonlog"
          ]
        }
      }
    ]
  }

In your output plugin, you configure:

elasticsearch {
    hosts => [ localhost ]
    index => "filebeat-%{+YYYY.MM.dd}"
    manage_template => false
    pipeline => "your_pipeline_name"
  }

And you forget problems with json parsing.

If you use filebeat you able to send json logs direct to pipeline by configuring filebeat:

 output.elasticsearch:
    ....
    pipelines:
       - pipeline: "your pipeline name"
          when:
            contains:
               message: "{"