在发送数据之前,我使用的是JSON.stringify数据,它看起来像这样
{"data": [{"key1": value1, "key2": value2}, {"key1": value1, "key2": value2}]}
但是一旦它通过AWS API Gateway并且Kinesis Firehose将它放到S3就看起来像这样
{
"key1": value1,
"key2": value2
}{
"key1": value1,
"key2": value2
}
JSON对象之间的分隔符逗号消失了,但我需要它来正确处理数据。
API网关中的模板:
#set($root = $input.path('$'))
{
"DeliveryStreamName": "some-delivery-stream",
"Records": [
#foreach($r in $root.data)
#set($data = "{
""key1"": ""$r.value1"",
""key2"": ""$r.value2""
}")
{
"Data": "$util.base64Encode($data)"
}#if($foreach.hasNext),#end
#end
]
}
答案 0 :(得分:1)
我最近遇到了同样的问题,我能找到的唯一答案基本上只是在每个JSON消息的末尾添加换行符(" \ n"),只要你将它们发布到Kinesis流,或使用某种原始JSON解码器方法,可以处理没有分隔符的连接JSON对象。
我发布了一个python代码解决方案,可以在相关的Stack Overflow帖子上找到: https://stackoverflow.com/a/49417680/1546785
答案 1 :(得分:0)
一旦AWS Firehose将JSON对象转储到s3,就完全有可能从文件中读取单个JSON对象。
使用 Python ,您可以使用raw_decode
包中的json
函数
from json import JSONDecoder, JSONDecodeError
import re
import json
import boto3
NOT_WHITESPACE = re.compile(r'[^\s]')
def decode_stacked(document, pos=0, decoder=JSONDecoder()):
while True:
match = NOT_WHITESPACE.search(document, pos)
if not match:
return
pos = match.start()
try:
obj, pos = decoder.raw_decode(document, pos)
except JSONDecodeError:
# do something sensible if there's some error
raise
yield obj
s3 = boto3.resource('s3')
obj = s3.Object("my-bukcet", "my-firehose-json-key.json")
file_content = obj.get()['Body'].read()
for obj in decode_stacked(file_content):
print(json.dumps(obj))
# { "key1":value1,"key2":value2}
# { "key1":value1,"key2":value2}
来源:https://stackoverflow.com/a/50384432/1771155
使用胶水/ Pyspark ,您可以使用
import json
rdd = sc.textFile("s3a://my-bucket/my-firehose-file-containing-json-objects")
df = rdd.map(lambda x: json.loads(x)).toDF()
df.show()
答案 2 :(得分:0)
您可以考虑的一种方法是,通过添加Lambda函数作为其数据处理器,为Kinesis Firehose传递流配置数据处理,该函数将在最终将数据传递到S3存储桶之前执行。
DeliveryStream:
...
Type: AWS::KinesisFirehose::DeliveryStream
Properties:
DeliveryStreamType: DirectPut
ExtendedS3DestinationConfiguration:
...
BucketARN: !GetAtt MyDeliveryBucket.Arn
ProcessingConfiguration:
Enabled: true
Processors:
- Parameters:
- ParameterName: LambdaArn
ParameterValue: !GetAtt MyTransformDataLambdaFunction.Arn
Type: Lambda
...
然后在Lambda函数中,确保将'\n'
附加到记录的JSON字符串中,请参见Node.js中Lambda函数myTransformData.ts
的下方:
import {
FirehoseTransformationEvent,
FirehoseTransformationEventRecord,
FirehoseTransformationHandler,
FirehoseTransformationResult,
FirehoseTransformationResultRecord,
} from 'aws-lambda';
const createDroppedRecord = (
recordId: string
): FirehoseTransformationResultRecord => {
return {
recordId,
result: 'Dropped',
data: Buffer.from('').toString('base64'),
};
};
const processData = (
payloadStr: string,
record: FirehoseTransformationEventRecord
) => {
let jsonRecord;
// ...
// Process the orginal payload,
// And create the record in JSON
return jsonRecord;
};
const transformRecord = (
record: FirehoseTransformationEventRecord
): FirehoseTransformationResultRecord => {
try {
const payloadStr = Buffer.from(record.data, 'base64').toString();
const jsonRecord = processData(payloadStr, record);
if (!jsonRecord) {
console.error('Error creating json record');
return createDroppedRecord(record.recordId);
}
return {
recordId: record.recordId,
result: 'Ok',
// Ensure that '\n' is appended to the record's JSON string.
data: Buffer.from(JSON.stringify(jsonRecord) + '\n').toString('base64'),
};
} catch (error) {
console.error('Error processing record ${record.recordId}: ', error);
return createDroppedRecord(record.recordId);
}
};
const transformRecords = (
event: FirehoseTransformationEvent
): FirehoseTransformationResult => {
let records: FirehoseTransformationResultRecord[] = [];
for (const record of event.records) {
const transformed = transformRecord(record);
records.push(transformed);
}
return { records };
};
export const handler: FirehoseTransformationHandler = async (
event,
_context
) => {
const transformed = transformRecords(event);
return transformed;
};
一旦使用了换行符分隔符,Athena之类的AWS服务将能够正确处理S3存储桶(而非just seeing the first JSON record only)中的JSON记录数据。