Elasticsearch搜索出现NAN异常

原因分析

• Elasticsearch默认的打分，一般是不会出现异常的
• 之所以会出现NAN异常，往往是因为我们重新计算了打分，使用了function_score
• 核心原因是在function_score中，出现了计算异常，比如 0/0,比如log1p(x),x为负数等

真实案例分析

测试索引

PUT tx
POST tx/_mapping
{"properties" : {"age" : {"type" : "long"},"time" : {"type" : "date","format" : "yyyy-MM-dd"},"title" : {"type" : "text","analyzer" : "ik_smart"}}
}
POST tx/_doc/19
{"title":"中古","time" : "1969-10-12","age":12
}

搜索query示例

POST /tx/_search
{"query": {"function_score": {"query": {"match_all": {}},"field_value_factor": {"field": "time","modifier": "log1p"}}}
}
- 其实modifier.log1p和如下搜索是完全一样的
POST /tx/_search
{"query": {"function_score": {"query": {"match_all": {}},"script_score": {"script": "_score * Math.log1p(1+doc['time'].value.toInstant().toEpochMilli())"}}}
}

搜索异常返回值

{"error" : {"root_cause" : [{"type" : "exception","reason" : "function score query returned an invalid score: NaN for doc: 0"}],"type" : "search_phase_execution_exception","reason" : "all shards failed","phase" : "query","grouped" : true,"failed_shards" : [{"shard" : 0,"index" : "tx","node" : "aPOBCVYkRA-a0R9FR4eDzQ","reason" : {"type" : "exception","reason" : "function score query returned an invalid score: NaN for doc: 0"}}]},"status" : 500
}

搜索异常原因分析

• 搜索使用了function_score重新打分，并且使用了log1p函数
• 但是由于数据中的时间是"1969-10-12"，但是es计算时间是从1970年计算开始，这个值换算成毫秒就是负数，导致log1p异常
• 所以最终得分异常，出现是NAN