阿里云日志服务对接Grafana

cd /var/lib/grafana/plugins/
git clone https://github.com/aliyun/aliyun-log-grafana-datasource-plugin
systemctl restart grafana-server

? | select approx_distinct(client_ip) as uv, count(1) as pv, __time__ - __time__% 60 as time group by time order by time limit 1000

Y - column输入：pv,uv
X - column [time]输入：time

这块日志服务SQL语句，先确保能在阿里日志查询分析控制台能查询通过，有数据输出，无语法错误，否则在Grafana无数据图形显示。
说明：
approx_distinct(client_ip)：估算 client_ip 列的唯一值的个数。
__time__ ：日志服务每条日志中内置的时间类型。灵活的时间维度统计，数学取模方法分组。__time_ - __time_ % 60 表示按1分钟进行取模，即1分钟跨度显示。
d. 在 Visualization 中，Grafana可以支持多重类型的视图。对于PV，UV数据，在标签栏中单击Graph，创建一个Graph视图。

e. General中Title修改该Panel的名称，这里统计PV、UV，相应的名称修改为PV、UV。

f. 最后一项Alert，暂时不设置报警通知，如果要设置报警通知，在该Panel中单击Create Alert。在Notifications配置报警通知。

4.2 配置出入网带宽

参考4.1 配置PV、UV，使用同样的方法添加出入网带宽的流量。

? |  select sum(size) as net_out, sum(request_length) as net_in, __time__ -__time__% 60 as time group by time order by time limit 1000

Y - column输入：net_in,net_out
X - column [time]输入：time

4.3 不同HTTP方法的占比

? |  select count(1) as pv, split_part(request,‘ ‘,1) as request group by request

Y - column输入：request,pv
X - column [time]输入：pie
4.4 不同HTTP状态码占比

? |  select count(1) as pv, status group by status

Y - column输入：status,pv
X - column [time]输入：pie
4.5 热门来源页面

?  | select count(1) as pv, referer group by referer order by pv desc limit 10

Y - column输入：referer,pv
X - column [time]输入：pie
4.6 延时最高页面

?  | select split_part(request,‘ ‘,2) as top_latency_url, request_time group by top_latency_url,request_time order by request_time desc limit 20

Y - column输入：top_latency_url, request_time
X - column [time]保持为空。
4.7 热门页面

?  | select count(1) as pv, split_part(request,‘ ‘,2) as path  group by path order by pv desc limit 10

Y - column输入：path,pv
X - column [time]保持为空。
4.8 前后端平均延时

?  |  select avg(request_time) as response_time, avg(upstream_time) as upstream_response_time,  __time__ -__time__% 60 as time group by time order by time limit 1000

Y - column输入：upstream_response_time,response_time
X - column [time]输入：time
4.9 客户端统计

?  | select count(1) as pv, case when  regexp_like(agent , ‘okhttp‘) then ‘okhttp‘ when  regexp_like(agent ,  ‘iPhone‘) then ‘iPhone‘ when regexp_like(agent ,  ‘Android‘)  then ‘Android‘ else ‘unKnown‘ end as agent group by  agent order by pv desc limit 10

Y - column输入：agent,pv
X - column [time]输入：pie
4.10 各省份占比统计

?  | select ip_to_province(client_ip) as province, count(1) as pv group by province HAVING ip_to_province(client_ip) <> ‘-1‘ order by pv desc limit 10

日志IP地理函数 ip_to_province(client_ip) 对无效IP，会返回”-1“。针对这块用 HAVING 过滤掉。
Y - column输入：province,pv
X - column [time]输入：pie
4.11 各城市占比统计

?  | select ip_to_city(client_ip) as city, count(1) as pv group by city HAVING ip_to_city(client_ip) <> ‘-1‘ and ip_to_city(client_ip) <> ‘内网IP‘ order by pv desc limit 10

Y - column输入：city,pv
X - column [time]输入：pie
4.12 保存和发布Dashboard
单击页面上方的保存按钮，发布Dashboard。