配置架构:
a. Heka’s Agent/Aggregator架构
b:以一台agent为例进行说明,agent1配置文件如下:
[NginxLogInput]
type = "LogstreamerInput"
log_directory = "/usr/local/openresty/nginx/logs/"
file_match = 'access\.log'
decoder = "NginxLogDecoder"
hostname = "ID:XM_1_1"
[NginxLogDecoder]
type = "SandboxDecoder"
filename = "lua_decoders/nginx_access.lua"
[NginxLogDecoder.config]
log_format = '$remote_addr - $remote_user [$time_local] "$request" $status $body_bytes_sent "$http_referer" "$http_user_agent"'
type = "nginx-access"
[ProtobufEncoder]
[NgxLogOutput]
type = "HttpOutput"
message_matcher = "TRUE"
address = "http://127.0.0.1:10000"
method = "POST"
encoder = "ProtobufEncoder"
# [NgxLogOutput.headers]
# content-type = "application/octet-stream"
# [NgxLogOutput]
# type = "LogOutput"
# message_matcher = "TRUE"
# encoder = "ProtobufEncoder"
aggregator配置如下:
[LogInput]
type = "HttpListenInput"
#parser_type = "message.proto"
address = "0.0.0.0:10000"
decoder = "ProtobufDecoder"
[ProtobufDecoder]
[ESJsonEncoder]
index = "%{Type}-%{2006.01.02}"
es_index_from_timestamp = true
type_name = "%{Type}"
[PayloadEncoder]
[LogOutput]
type = "LogOutput"
message_matcher = "TRUE"
encoder = "ESJsonEncoder"
2. 问题描述:通过以上配置以后本应该可以将nginx log文件中数据发送到aggregator,并显示出来,但实际上并未显示;
3. 解决方法:
a. 修改protobuf.go中的Decode接口:
if err = proto.Unmarshal([]byte(*pack.Message.Payload), pack.Message); err == nil {
// fmt.Println("ProtobufDecoder:", string(pack.MsgBytes))
//fmt.Println("ProtobufDecoder", pack.Message.Fields)
packs = []*PipelinePack{pack}
} else {
atomic.AddInt64(&p.processMessageFailures, 1)
}
通过以上代码可以看出我们是将Unmarshal接口中的第一个参数pack.MsgBytes修改为pack.Message.Payload这样既可将agent端发送的数据在aggregator正确解析,但因pack.Message.Payload是string类型,不太建议这种修改方式;
b 修改http_listen_input.go:
pack.Message.SetSeverity(int32(6))
if hli.conf.UnescapeBody {
unEscapedBody, _ := url.QueryUnescape(string(body))
pack.Message.SetPayload(unEscapedBody)
} else {
pack.Message.SetPayload(string(body))
}
pack.MsgBytes = body //添加这个赋值
这样通过修改后,在protobuf Decode时直接将pack.MsgBytes 进行Unmarshal即可;