bloom-server 基于 rust 编写的 rest api cache 中间件,他位于lb 与api worker 之间，使用redis 作为缓存内容存储，
我们需要做的就是配置proxy，同时他使用基于share 的概念，进行cache 的分布存储，包含了请求端口（proxy，访问数据）
以及cache 控制端口（api 方便cache 策略的控制）

测试环境使用openresty+ docker + docker-compose 运行

一张参考图

环境准备

• docker-compose 文件

 

version: "3"

services:

  redis:

    image: redis

    ports:

    - "6379:6379"

  lb: 

    image: openresty/openresty:alpine-fat

    volumes:

    - "./nginx/nginx-lb.conf:/usr/local/openresty/nginx/conf/nginx.conf"

    ports:

    - "9000:80"

  webapi:

    image: openresty/openresty:alpine-fat

    volumes:

    - "./nginx/nginx.conf:/usr/local/openresty/nginx/conf/nginx.conf"

    ports:

    - "8090:80"

  webapi2:

    image: openresty/openresty:alpine-fat

    volumes:

    - "./nginx/nginx2.conf:/usr/local/openresty/nginx/conf/nginx.conf"

    ports:

    - "8091:80"

  bloom:

    image: valeriansaliou/bloom:v1.25.0

    environment:

    - "RUST_BACKTRACE=1"

    volumes:

    - "./bloom/config.cfg:/etc/bloom.cfg"

    ports:

    - "8080:8080"

    - "8811:8811"

  bloom2:

    image: valeriansaliou/bloom:v1.25.0

    environment:

    - "RUST_BACKTRACE=1"

    volumes:

    - "./bloom/config2.cfg:/etc/bloom.cfg"

    ports:

    - "8081:8080"

    - "8812:8811"

​

 

• webapi 配置
  主要是基于openresty 的 lua 脚本编写
  webapi1 nginx/nginx.conf

 

worker_processes 1;

user root;  

events {

    worker_connections 1024;

}

http {

    include mime.types;

    default_type application/octet-stream;

    sendfile on;

    lua_code_cache off;

    lua_need_request_body on;

    gzip on;

    resolver 127.0.0.11 ipv6=off;          

    real_ip_header X-Forwarded-For;

    real_ip_recursive on;

    server {

        listen 80;

        server_name app;

        charset utf-8;

        default_type text/html;

        location / {

           default_type text/plain;

           index index.html index.htm;

        }

        location /userinfo {

            default_type application/json;

            content_by_lua_block {

             local cjson = require("cjson");

             local user = {

                 name ="dalongdempo",

                 age =333,

             }

             ngx.say(cjson.encode(user))

            }

        }

        error_page 500 502 503 504 /50x.html;

        location = /50x.html {

            root html;

        }

​

    }

}

   

webapi2 nginx/nginx2.conf

worker_processes 1;

user root;  

events {

    worker_connections 1024;

}

http {

    include mime.types;

    default_type application/octet-stream;

    sendfile on;

    lua_code_cache off;

    lua_need_request_body on;

    gzip on;

    resolver 127.0.0.11 ipv6=off;          

    real_ip_header X-Forwarded-For;

    real_ip_recursive on;

    server {

        listen 80;

        server_name app;

        charset utf-8;

        default_type text/html;

        location / {

           default_type text/plain;

           index index.html index.htm;

        }

        location /userinfo2 {

            default_type application/json;

            content_by_lua_block {

             local cjson = require("cjson");

             local user = {

                 name ="dalongdempo222",

                 age =333,

             }

             ngx.say(cjson.encode(user))

            }

        }

        error_page 500 502 503 504 /50x.html;

        location = /50x.html {

            root html;

        }

​

    }

}

   

• bloom 配置
  bloom 类似一个sidecar，官方建议的方式是部署在每个服务的节点，但是实际上我们可以部署在
  其他地方，进行分离部署
  bloom1 配置 bloom/config.cfg 注意我将log级别修改为了debug，方便测试

 

# Bloom

# HTTP REST API caching middleware

# Configuration file

# Example: https://github.com/valeriansaliou/bloom/blob/master/config.cfg

​​

[server]

​

log_level = "debug"

inet = "0.0.0.0:8080"

​

​

[control]

​

inet = "0.0.0.0:8811"

tcp_timeout = 600

​

​

[proxy]

​

[[proxy.shard]]

​

shard = 0

host = "webapi"

port = 80

​

​

[cache]

​

ttl_default = 600

executor_pool = 64

​

disable_read = false

disable_write = false

​

compress_body = true

​

​

[redis]

​

host = "redis"

port = 6379

​

database = 0

​

pool_size = 80

max_lifetime_seconds = 60

idle_timeout_seconds = 600

connection_timeout_seconds = 1

​

max_key_size = 256000

max_key_expiration = 2592000

   

bloom2 配置 bloom/config2.cfg

# Bloom

# HTTP REST API caching middleware

# Configuration file

# Example: https://github.com/valeriansaliou/bloom/blob/master/config.cfg

​

​

[server]

​

log_level = "debug"

inet = "0.0.0.0:8080"

​

​

[control]

​

inet = "0.0.0.0:8811"

tcp_timeout = 600

​

​

[proxy]

​

[[proxy.shard]]

​

shard = 1

host = "webapi2"

port = 80

​

​

[cache]

​

ttl_default = 600

executor_pool = 64

​

disable_read = false

disable_write = false

​

compress_body = true

​

​

[redis]

​

host = "redis"

port = 6379

​

database = 0

​

pool_size = 80

max_lifetime_seconds = 60

idle_timeout_seconds = 600

connection_timeout_seconds = 1

​

max_key_size = 256000

max_key_expiration = 2592000

   

启动&&测试

• 启动

docker-compose up -d

 

• 测试效果
  从bloom 访问,注意需要Bloom-Request-Shard

 

curl -X GET \

  http://localhost:8081/userinfo2 \

  -H 'Bloom-Request-Shard: 1' \

  -H 'Content-Type: application/json' \

  -H 'Postman-Token: d13543ca-a031-47e3-b47a-996a6faaad53' \

  -H 'cache-control: no-cache'

{"age":333,"name":"dalongdempo222"}

curl -X GET \

  http://localhost:8080/userinfo \

  -H 'Bloom-Request-Shard: 0' \

  -H 'Content-Type: application/json' \

  -H 'Postman-Token: bc168c7c-3b8b-471d-aa01-6eb6cb0d421c' \

  -H 'cache-control: no-cache'

   

redis key

KEYS *

1) "bloom:0:a:dc56d17a"

2) "bloom:1:a:dc56d17a"

   

openresty(lb) 端访问： 不用添加header，因为nginx端已经配置了

curl http://localhost:9000/userinfo

{"age":333,"name":"dalongdempo"}

 

说明

bloom-server 还是一个不错的rest cache proxy，同时openresty 端也是可以做的，作者在项目中也介绍为什么不那么
干的原因，同时bloom-server 基于rust 编写，具有类型安全，以及很不错的性能，同时也包含了好多可选的控制，但是
如果你在运行的时候可能会发现部分提示并不是很友好，尤其对于在排查问题的时候，但是总的来说设计还是挺好的

参考资料

https://github.com/valeriansaliou/bloom
https://crates.io/crates/bloom-server
https://github.com/rongfengliang/bloom-server-docker-compose