varnish-4.x VCL之directors

varnish-4.x VCL之directors

man vcl

man vmod_directors

一.引用语法

示例:

import directors;

sub vcl_init {

    new vdir = directors.random();

    vdir.add_backend(backend1, 10); 

    vdir.add_backend(backend2, 5);

}

二.4种调度算法

       · Object fallback

       · VOID fallback.add_backend(BACKEND)

       · BACKEND fallback.backend()

A fallback director will try each of the added backends in turn, and return the first one that is healthy

fallback算法,将request按顺序下发给第一个healthy的后端server

       · Object hash

       · VOID hash.add_backend(BACKEND, REAL)

       · BACKEND hash.backend(STRING_LIST)

The director chooses the backend server by computing a hash/digest of the string given to .backend().

 Commonly used with client.identity or a session cookie to get sticky sessions.

根据hash策略下发request, 可保证相同客户端的request下发到相同的后端server上

       · Object random

       · VOID random.add_backend(BACKEND, REAL)

       · BACKEND random.backend()

The random director distributes load over the backends using a weighted random probability distribution

按权重比下发请求到后端server

权重比计算方式:100 * (weight / (sum(all_added_weights))) ,如:

    vdir.add_backend(backend1, 10); 

    vdir.add_backend(backend2, 5); 

# 2/3 to backend1, 1/3 to backend2.

       · Object round_robin

       · VOID round_robin.add_backend(BACKEND)

       · BACKEND round_robin.backend()

说明: hash和random都支持权重

如下是本人测试的示例文件,包含healthcheck申明,director引用, purge acl

vcl 4.0;

import std;

import directors;

probe healthcheck_nginx {

    .url = "/healthy.jpg";

    .timeout = 3s;

    .interval = 5s;

    .window = 5;

    .threshold = 3;

}

probe healthcheck_apache {

    .url = "/healthy.jpg";

    .timeout = 3s;

    .interval = 5s;

    .window = 5;

    .threshold = 3;

}

backend nginx {

    .host = "192.168.192.10";

    .port = "8080";

    .connect_timeout = 3s;

    .first_byte_timeout = 5s;

    .between_bytes_timeout = 5s;

    .max_connections = 30000;

    .probe = healthcheck_nginx;

}

backend apache {

    .host = "192.168.192.10";

    .port = "8081";

    .connect_timeout = 3s;

    .first_byte_timeout = 5s;

    .between_bytes_timeout = 5s;

    .max_connections = 3000;

    .probe = healthcheck_apache;

}

sub vcl_init {

    new round_robin_director = directors.round_robin();

    round_robin_director.add_backend(apache);

    round_robin_director.add_backend(nginx);

    new random_director = directors.random();

    random_director.add_backend(nginx, 10);

    random_director.add_backend(apache, 5);

    new hash_director = directors.hash();

    hash_director.add_backend(nginx, 10);

    hash_director.add_backend(apache, 5);

}

acl purgers {

    "127.0.0.1";

    "192.168.0.0"/24;

}

sub vcl_recv {

    # allow PURGE from localhost and 192.168.0...

    if (req.restarts == 0) {

        unset req.http.X-Purger;

    }

    if (req.method == "PURGE") {

        if (!client.ip ~ purgers) {

            return (synth(405, "Purging not allowed for " + client.ip));

        }

        return (purge);

    }

    #set req.backend_hint = round_robin_director.backend();

    set req.backend_hint = hash_director.backend(req.http.cookie);

}

sub vcl_purge {

    set req.method = "GET";

    set req.http.X-Purger = "Purged";

    return (restart);

}

sub vcl_deliver {

    if (req.http.X-Purger) {

        set resp.http.X-Purger = req.http.X-Purger;

    }

}