mirror of https://github.com/easzlab/kubeasz.git
124 lines
5.4 KiB
Markdown
124 lines
5.4 KiB
Markdown
|
## EX-LB 负载均衡部署
|
|||
|
|
|||
|
根据[HA 2x架构](00-planning_and_overall_intro.md),k8s集群自身高可用已经不依赖于外部 lb 服务;但是有时我们要从外部访问 apiserver(比如 CI 流程),就需要 ex-lb 来请求多个 apiserver;
|
|||
|
|
|||
|
还有一种情况是需要[负载转发到ingress服务](../op/loadballance_ingress_nodeport.md),也需要部署ex-lb;
|
|||
|
|
|||
|
当遇到公有云环境无法自建 ex-lb 服务时,可以配置对应的云负载均衡服务。
|
|||
|
|
|||
|
### ex-lb 服务组件
|
|||
|
|
|||
|
ex-lb 服务由 keepalived 和 haproxy 组成:
|
|||
|
- haproxy:高效代理(四层模式)转发到多个 apiserver
|
|||
|
- keepalived:利用主备节点vrrp协议通信和虚拟地址,消除haproxy的单点故障
|
|||
|
|
|||
|
``` bash
|
|||
|
roles/ex-lb/
|
|||
|
├── clean-ex-lb.yml
|
|||
|
├── defaults
|
|||
|
│ └── main.yml
|
|||
|
├── ex-lb.yml
|
|||
|
├── tasks
|
|||
|
│ └── main.yml
|
|||
|
└── templates
|
|||
|
├── haproxy.cfg.j2
|
|||
|
├── haproxy.service.j2
|
|||
|
├── keepalived-backup.conf.j2
|
|||
|
└── keepalived-master.conf.j2
|
|||
|
```
|
|||
|
|
|||
|
Haproxy支持四层和七层负载,稳定性好,根据官方文档,HAProxy可以跑满10Gbps-New benchmark of HAProxy at 10 Gbps using Myricom's 10GbE NICs (Myri-10G PCI-Express);另外,openstack高可用也有用haproxy的。
|
|||
|
|
|||
|
keepalived观其名可知,保持存活,它是基于VRRP协议保证所谓的高可用或热备的,这里用来预防haproxy的单点故障。
|
|||
|
|
|||
|
keepalived与haproxy配合,实现master的高可用过程如下:
|
|||
|
|
|||
|
+ 1.keepalived利用vrrp协议生成一个虚拟地址(VIP),正常情况下VIP存活在keepalive的主节点,当主节点故障时,VIP能够漂移到keepalived的备节点,保障VIP地址高可用性。
|
|||
|
+ 2.在keepalived的主备节点都配置相同haproxy负载配置,并且监听客户端请求在VIP的地址上,保障随时都有一个haproxy负载均衡在正常工作。并且keepalived启用对haproxy进程的存活检测,一旦主节点haproxy进程故障,VIP也能切换到备节点,从而让备节点的haproxy进行负载工作。
|
|||
|
+ 3.在haproxy的配置中配置多个后端真实kube-apiserver的endpoints,并启用存活监测后端kube-apiserver,如果一个kube-apiserver故障,haproxy会将其剔除负载池。
|
|||
|
|
|||
|
#### 安装haproxy
|
|||
|
|
|||
|
+ 使用apt源安装
|
|||
|
|
|||
|
#### 配置haproxy (roles/ex-lb/templates/haproxy.cfg.j2)
|
|||
|
|
|||
|
配置由全局配置和三个listen配置组成:
|
|||
|
- listen kube-master 用于转发至多个apiserver
|
|||
|
- listen ingress-node 用于转发至node节点的ingress http服务,[参阅](../op/loadballance_ingress_nodeport.md)
|
|||
|
- listen ingress-node-tls 用于转发至node节点的ingress https服务
|
|||
|
|
|||
|
如果用apt安装的话,可以在/usr/share/doc/haproxy目录下找到配置指南configuration.txt.gz,全局和默认配置这里不展开,关注`listen` 代理设置模块,各项配置说明:
|
|||
|
+ 名称 kube-master
|
|||
|
+ bind 监听客户端请求的地址/端口,保证监听master的VIP地址和端口
|
|||
|
+ mode 选择四层负载模式 (当然你也可以选择七层负载,请查阅指南,适当调整)
|
|||
|
+ balance 选择负载算法 (负载算法也有很多供选择)
|
|||
|
|
|||
|
#### 安装keepalived
|
|||
|
|
|||
|
+ 使用apt源安装
|
|||
|
|
|||
|
#### 配置keepalived主节点 [keepalived-master.conf.j2](../../roles/ex-lb/templates/keepalived-master.conf.j2)
|
|||
|
|
|||
|
``` bash
|
|||
|
global_defs {
|
|||
|
router_id lb-master-{{ inventory_hostname }}
|
|||
|
}
|
|||
|
|
|||
|
vrrp_script check-haproxy {
|
|||
|
script "killall -0 haproxy"
|
|||
|
interval 5
|
|||
|
weight -60
|
|||
|
}
|
|||
|
|
|||
|
vrrp_instance VI-kube-master {
|
|||
|
state MASTER
|
|||
|
priority 120
|
|||
|
unicast_src_ip {{ inventory_hostname }}
|
|||
|
unicast_peer {
|
|||
|
{% for h in groups['ex-lb'] %}{% if h != inventory_hostname %}
|
|||
|
{{ h }}
|
|||
|
{% endif %}{% endfor %}
|
|||
|
}
|
|||
|
dont_track_primary
|
|||
|
interface {{ LB_IF }}
|
|||
|
virtual_router_id {{ ROUTER_ID }}
|
|||
|
advert_int 3
|
|||
|
track_script {
|
|||
|
check-haproxy
|
|||
|
}
|
|||
|
virtual_ipaddress {
|
|||
|
{{ EX_APISERVER_VIP }}
|
|||
|
}
|
|||
|
}
|
|||
|
```
|
|||
|
+ vrrp_script 定义了监测haproxy进程的脚本,利用shell 脚本`killall -0 haproxy` 进行检测进程是否存活,如果进程不存在,根据`weight -30`设置将主节点优先级降低30,这样原先备节点将变成主节点。
|
|||
|
+ vrrp_instance 定义了vrrp组,包括优先级、使用端口、router_id、心跳频率、检测脚本、虚拟地址VIP等
|
|||
|
+ 特别注意 `virtual_router_id` 标识了一个 VRRP组,在同网段下必须唯一,否则出现 `Keepalived_vrrp: bogus VRRP packet received on eth0 !!!`类似报错
|
|||
|
+ 配置 vrrp 协议通过单播发送
|
|||
|
|
|||
|
#### 配置keepalived备节点 [keepalived-backup.conf.j2](../../roles/ex-lb/templates/keepalived-backup.conf.j2)
|
|||
|
|
|||
|
+ 备节点的配置类似主节点,除了优先级和检测脚本,其他如 `virtual_router_id` `advert_int` `virtual_ipaddress`必须与主节点一致
|
|||
|
|
|||
|
### 启动 keepalived 和 haproxy 后验证
|
|||
|
|
|||
|
+ lb 节点验证
|
|||
|
|
|||
|
``` bash
|
|||
|
systemctl status haproxy # 检查进程状态
|
|||
|
journalctl -u haproxy # 检查进程日志是否有报错信息
|
|||
|
systemctl status keepalived # 检查进程状态
|
|||
|
journalctl -u keepalived # 检查进程日志是否有报错信息
|
|||
|
```
|
|||
|
+ 在 keepalived 主节点
|
|||
|
|
|||
|
``` bash
|
|||
|
ip a # 检查 master的 VIP地址是否存在
|
|||
|
```
|
|||
|
### keepalived 主备切换演练
|
|||
|
|
|||
|
1. 尝试关闭 keepalived主节点上的 haproxy进程,然后在keepalived 备节点上查看 master的 VIP地址是否能够漂移过来,并依次检查上一步中的验证项。
|
|||
|
1. 尝试直接关闭 keepalived 主节点系统,检查各验证项。
|
|||
|
|