[QUESTION]:ilogtail的压测结果和阿里云官网文档提供的数据差距很大 #1143
Replies: 7 comments
-
#809 可以参考下 |
Beta Was this translation helpful? Give feedback.
-
我的投递目标是SLS的logstroe,并且我的数据是随机的,这篇文档的测试数据应该都是相同的。 |
Beta Was this translation helpful? Give feedback.
-
@xiongjunkun 1. 测试的网络环境是怎么样的呢,到sls使用的是公网还是阿里云内网? 2. logstore的shard数量是多少?单shard写入能力是有上限的,具体可以看下sls 官方文档 |
Beta Was this translation helpful? Give feedback.
-
1、走的公网,我自己机器上面的k8s。 |
Beta Was this translation helpful? Give feedback.
-
比较大的可能还是网络瓶颈。
|
Beta Was this translation helpful? Give feedback.
-
ilogtail默认采集动态数据,所以“提前在节点中构造20G的日志文件”的方法是需要调整下。 |
Beta Was this translation helpful? Give feedback.
-
可以使用https://github.com/iLogtail/mock-log 构造动态数据 |
Beta Was this translation helpful? Give feedback.
-
目前正在调研各种日志采集客户端的性能,所以针对ilogtail进行了一系列压测,发现压测结果和阿里云官网提供的数据差距很大,有人知道原因吗?
我的压测步骤:
1、提前在节点中构造20G的日志文件:5个日志文件,每个日志文件的大小是4G,并且日志内容是随机的。
2、在K8s中部署ilogtail,并且在ilogtail容器的环境变量中配置ilogtail的资源规格:
- name: cpu_usage_limit
value: "1"
- name: mem_usage_limit
value: "1024"
- name: max_bytes_per_sec
value: "209715200"
- name: process_thread_count
value: "8"
- name: send_request_concurrency
value: "20"
最后压测的结果,极简模式的采集速度是 38.94 MB/s 。
然后我将cpu_usage_limit和mem_usage_limit调整为2c2g,发现极简模式的采集速度还是38M左右,远远低于阿里元官网说明的100M/s
Beta Was this translation helpful? Give feedback.
All reactions