fluent-plugin-elasticsearch："Could not push log to Elasticsearch" 错误 "error"=>{"type"=>"mapper_parsing_exception"}

Question

当我使用 fluent-plugin-elasticsearch 将 Fluentd 收集的数据注入 Elasticsearch 时，一些数据导致了以下错误：

2017-04-09 23:47:37 +0900 [error]: Could not push log to Elasticsearch: {"took"=>3, "errors"=>true, "items"=>[{"index"=>{"_index"=>"logstash-201704", "_type"=>"ruby", "_id"=>"AVtTLz_cUzkwT9CQCxrH", "status"=>400, "error"=>{"type"=>"mapper_parsing_exception", "reason"=>"failed to parse [message]", "caused_by"=>{"type"=>"illegal_state_exception", "reason"=>"Can't get text on a START_OBJECT at 1:27"}}}}, .....]}

elasticsearch 似乎禁止了错误 failed to parse [message] 和 Can't get text on a START_OBJECT at 1:27 的数据。但是我看不到什么数据被发送到 Elasticsearch 以及出了什么问题。

有什么想法吗？

Answer 1

fluent-plugin-elasticsearch 使用 _bulk API 发送数据。我将请求转储代码放在 /opt/td-agent/embedded/lib/ruby/gems/2.1.0/gems/elasticsearch-api-5.0.4/lib/elasticsearch/api/actions/bulk.rb 上，如下所示：

  def bulk(arguments={})
    ...
      payload = body
    end
    $log.info([method, path, params, payload].inspect)  # <=== here ($log is global logger of fluentd)
    perform_request(method, path, params, payload).body

我发现发送到 Elasticsearch 的请求如下：

POST /_bulk
{"index":{"_index":"logstash-201704","_type":"ruby"}}
{"level":"INFO","message":{"status":200,"time":{"total":46.26,"db":33.88,"view":12.38},"method":"PUT","path":"filtered","params":{"time":3815.904,"chapter_index":0},"response":[{}]},"node":"main","time":"2017-04-09T14:39:06UTC","tag":"filtered.console","@timestamp":"2017-04-09T23:39:06+09:00"}

问题是 message 字段包含 JSON 对象，尽管此字段在 Elasticsearch 上被映射为已分析的字符串。

fluent-plugin-elasticsearch："Could not push log to Elasticsearch" 错误 "error"=>{"type"=>"mapper_parsing_exception"}

fluent-plugin-elasticsearch: "Could not push log to Elasticsearch" error with "error"=>{"type"=>"mapper_parsing_exception"}

elasticsearch

fluent