本文的內(nèi)容
- 如何用filebeat kafka es做一個好用璧尸，好管理的日志收集工具
- 放棄logstash柄粹，使用elastic pipeline
- gunicron日志格式與filebeat/es配置
- flask日志格式與異常日志采集與filebeat/es配置
- 以上的配置
概況
我有一個HTTP請求，經(jīng)過的路徑為
Gateway(kong)-->WebContainer(gunicorn)-->WebApp(flask)
我準備以下流向處理我的日志
file --> filebeat --> kafka topic--> filebeat --> elastic pipeline --> elasticsearch
|
| ----------> HBase
為什么這么做
Logstash去哪里了？
- Logstash太重了，不過這不是問題藻三，也就是多個機器加點錢的問題。能把事情處理就行跪者。
- Logstash不美棵帽，Logstash雖然是集中管理配置，但是一個logstash好像總是不夠渣玲，Logstash好像可以分開配置逗概，但是你永遠不知道如何劃分哪些配置應該放在一個配置文件，哪些應該分開忘衍。
- 刪除一個配置逾苫？不可能的，我怎么知道應該刪除什么配置淑履。
- 如果用了Logstash. As a 'poor Ops guys having to understand and keep up with all the crazy input possibilities. _
Filebeat的痛處
- 看看這個Issue吧, 萬人血書讓filebeat支持
grok, 但是就是不支持隶垮，不過給了我們兩條路，比如你可以用存JSON的日志啊, 或者用pipeline - Filebeat以前是沒有一個好的kafka-input秘噪。只能自己寫kafka-es的轉(zhuǎn)發(fā)工具
簡單點
我想要的日志采集就是簡簡單單，或者說微服務(wù)的內(nèi)聚力勉耀。 一條日志采集線就不該和其他業(yè)務(wù)混合指煎。最好的就是以下這種狀態(tài)
onefile -> filebeat_config -> kafka_topic -> filebeat_config -> elastic pipepline -> es index
Gunicorn日志
gunicorn日志
gunicorn日志采集如下的信息
- time
- client_ip
- http method
- http scheme
- url
- url query string
- response status code
- client name
- rt
- trace id
- remote ips
日志格式
%(t)s [%(h)s] [%(m)s] [%(H)s] [%(U)s] [%(q)s] [%(s)s] [%(a)s] [%(D)s] [%({Kong-Request-ID}i)s] [%({X-Forwarded-For}i)s]
日志例子
[15/Nov/2019:10:23:37 +0000] [172.31.37.123] [GET] [HTTP/1.1] [/api/v1/_instance/json_schema/Team/list] [a=1] [200] [Mozilla/5.0 (Macintosh; Intel Mac OS X 10_14_3) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/78.0.3904.97 Safari/537.36] [936] [9cbf6a3b-9c3a-4835-a2ef-02e03ee826d7#16] [137.59.103.3, 172.30.17.253, 172.30.18.12]
Es processing解析
es processing是6.0之后的功能，相當于es之前自帶了一個logstash.對于復雜日志有多種processing便斥，
可以使用grok或者dissect.某些情況下dissect更加快一些.
經(jīng)過kafka至壤，再有filebeat打到ES, 需要刪除多余的信息
PUT _ingest/pipeline/gunicorn
{
"description" : "devops gunicorn pipeline",
"processors" : [
{
"remove": {"field": ["agent", "ecs", "host", "input", "kafka"]}
},
{
"json": {
"field": "message",
"add_to_root": true
}
},
{
"remove": {"field": ["@metadata", "ecs", "agent", "input"]}
},
{
"dissect" : {
"field": "message",
"pattern": "[%{@timestamp}] [%{client_ip}] [%{method}] [%{scheme}] [%{path}] [%{query_string}] [%{status}] [%{client}] [%{rt_millo}] [%{trace_id}] [%{remote_ips}]"
}
}
],
"on_failure": [
{
"set": {
"field": "_index",
"value": "failed-{{ _index }}"
}
}
]
}
Es mapping
這里比較關(guān)鍵的是ES時間格式文檔的定義， 如果某些字段我們覺得有必要分詞枢纠，就是用text像街。否則使用keyword。這樣可以更加
方便的聚合和查詢?nèi)罩緮?shù)據(jù), 開啟_source方便做一些數(shù)據(jù)統(tǒng)計
PUT _template/gunicorn
{
"index_patterns": ["*gunicorn*"],
"settings": {
"number_of_shards": 1
},
"version": 1,
"mappings": {
"_source": {
"enabled": true
},
"properties": {
"@timestamp": {
"type": "date",
"format": "dd/LLL/yyyy:HH:mm:ss Z"
},
"client_ip": {
"type": "ip"
},
"method": {
"type": "keyword"
},
"scheme": {
"type": "keyword"
},
"path": {
"type": "text"
},
"query_string": {
"type": "text"
},
"status": {
"type": "integer"
},
"client": {
"type": "text"
},
"rt_millo": {
"type": "long"
},
"trace_id": {
"type": "keyword"
},
"remote_ips": {
"type": "text"
}
}
}
}
filebeat 采集到kafka配置文件
filebeat.inputs:
- type: log
paths:
- /yourpath/gunicorn-access.log
multiline.pattern: '^\['
multiline.negate: true
multiline.match: after
tail_files: true
queue.mem:
events: 4096
flush.min_events: 512
flush.timeout: 5s
output.kafka:
hosts: ["kafka-01","kafka-02","kafka-03"]
topic: 'gunicron_access'
required_acks: 1
compression: gzip
max_message_bytes: 1000000
filebeat 從kafka消費配置文件
filebeat.inputs:
- type: kafka
hosts: ["kafka-01","kafka-02","kafka-03"]
topics: ["gunicron_access"]
group_id: "filebeat_gunicron"
output.elasticsearch:
hosts: ["es-url"]
pipeline: "gunicorn"
index: "gunicorn-%{+yyyy.MM.dd}"
setup.template.name: "gunicorn"
setup.template.pattern: "gunicorn-*"
setup.ilm.enabled: false
setup.template.enabled: false
Flask日志
Flask日志是我們程序打印的，用于查看一些異常和錯誤的日志镰绎。在上線初期脓斩，info日志是可以打開debug的日志的。這樣方便我們進行調(diào)試畴栖。
在穩(wěn)定之后應該將日志接受級別調(diào)高随静。info日志不適合做統(tǒng)計，只是除了問題我們可以快速定位問題所在吗讶。 異常應該打到info日志中
INFO日志可以使用我建議的格式燎猛。我們關(guān)心
- time
- levelname: 日志級別
- host, process, thread: 用于定位到某臺機器的某個進程下的某個線程(一些復雜的bug需要，或者開啟了異步進程)
- name, funcname, filename, lineno: 用于定位日志發(fā)生的代碼位置
- message: 日志內(nèi)容
日志格式
{
"format": "[%(asctime)s.%(msecs)03d] [%(levelname)s] [{}:%(process)d:%(thread)d] [%(name)s:%(funcName)s] [%(filename)s:%(lineno)d] %(message)s".format(HOST),
"datefmt": "%Y-%m-%d %H:%M:%S"
}
日志例子
[2019-11-18 08:47:49.424] [INFO] [cmdb-008069:5990:140482161399552] [cmdb:execute_global_worker] [standalone_scheduler.py:116] RUN_INFO: tiny_collector_ali starting at 2019-11-18 08:47:49, next run will be at approximately 2019-11-18 09:47:49
[2019-11-18 08:11:27.715] [ERROR] [cmdb-008069:5985:140184204932928] [cmdb:common_handler] [error.py:48] 404 Not Found: The requested URL was not found on the server. If you entered the URL manually please check your spelling and try again.
Traceback (most recent call last):
File "/home/server/venv3/lib/python3.6/site-packages/flask/app.py", line 1805, in full_dispatch_request
rv = self.dispatch_request()
File "/home/server/venv3/lib/python3.6/site-packages/flask/app.py", line 1783, in dispatch_request
self.raise_routing_exception(req)
File "/home/server/venv3/lib/python3.6/site-packages/flask/app.py", line 1766, in raise_routing_exception
raise request.routing_exception
File "/home/server/venv3/lib/python3.6/site-packages/flask/ctx.py", line 336, in match_request
self.url_adapter.match(return_rule=True)
File "/home/server/venv3/lib/python3.6/site-packages/werkzeug/routing.py", line 1799, in match
raise NotFound()
werkzeug.exceptions.NotFound: 404 Not Found: The requested URL was not found on the server. If you entered the URL manually please check your spelling and try again.
Es processing解析
經(jīng)過kafka照皆，再有filebeat打到ES, 需要刪除多余的信息
PUT _ingest/pipeline/info
{
"description" : "devops info pipeline",
"processors" : [
{
"remove": {"field": ["agent", "ecs", "host", "input", "kafka"]}
},
{
"json": {
"field": "message",
"add_to_root": true
}
},
{
"remove": {"field": ["@metadata", "ecs", "agent", "input"]}
},
{
"dissect" : {
"field": "message",
"pattern": "[%{@timestamp}] [%{level}] [%{host}:%{process_id}:%{thread_id}] [%{name}:%{func_name}] [%{file}:%{line_no}] %{content}"
}
}
],
"on_failure": [
{
"set": {
"field": "_index",
"value": "failed-{{ _index }}"
}
}
]
}
Es mapping
thread_id 要給一個long字段重绷， python如果獲取不到會給一個超出integer范圍的數(shù)字
PUT _template/info
{
"index_patterns": ["*info*"],
"settings": {
"number_of_shards": 1
},
"version": 1,
"mappings": {
"_source": {
"enabled": true
},
"properties": {
"@timestamp": {
"type": "date",
"format": "yyyy-MM-dd HH:mm:ss.SSS"
},
"level": {
"type": "keyword"
},
"host": {
"type": "keyword"
},
"process_id": {
"type": "integer"
},
"thread_id": {
"type": "long"
},
"name": {
"type": "keyword"
},
"func_name": {
"type": "keyword"
},
"file": {
"type": "keyword"
},
"line_no": {
"type": "integer"
},
"content": {
"type": "text"
}
}
}
}
filebeat 采集到Kafka配置文件
這里采用^\[20\d{2}來區(qū)分行首
filebeat.inputs:
- type: log
paths:
- /you_path/app.log
multiline.pattern: '^\[20\d{2}'
multiline.negate: true
multiline.match: after
tail_files: true
queue.mem:
events: 4096
flush.min_events: 512
flush.timeout: 5s
output.kafka:
hosts: ["kafka-01", "kafka-02", "kafka-03"]
topic: 'devops_app'
required_acks: 1
compression: gzip
max_message_bytes: 1000000
filebeat 從kafka消費配置文件
filebeat.inputs:
- type: kafka
hosts: ["kafka-01", "kafka-02", "kafka-03"]
topics: ["devops_app"]
group_id: "filebeat_app"
output.elasticsearch:
hosts: ["es_url"]
pipeline: "info"
index: "app-info-%{+yyyy.MM.dd}"
setup.template.name: "info"
setup.template.pattern: "app-info-*"
setup.ilm.enabled: false
setup.template.enabled: false
