openstack octavia 简介以及手工安装过程

谁借莪１个温暖的怀抱￠ 2022-03-20 01:28 529阅读 0赞

openstack octavia 是 openstack lbaas的支持的一种后台程序，提供为虚拟机流量的负载均衡。实质是类似于trove，调用 nove 以及neutron的api生成一台安装好haproxy和keepalived软件的虚拟机，并连接到目标网路。octavia共有4个组件 housekeeping,worker,api,health-manager，octavia agent。api作用就不详细说了。worker：主要作用是和nova，neutron等组件通信，用于虚拟机调度以及把对于虚拟机操作的指令下发给octavia agent。housekeeping：查看octavia/controller/housekeeping/house\_keeping.py得知其三个功能点：SpareAmphora,DatabaseCleanup,CertRotation。依次是清理虚拟机的池子，清理过期数据库，更新证书。health-manager：检查虚拟机状态，和虚拟机中的octavia agent通信，来更新各个组件的状态。octavia agent 位于虚拟机内部：对下是接受指令操作下层的haproxy软件，对上是和health-manager通信汇报各种情况。可以参考博文[http://lingxiankong.github.io/blog/2016/03/30/octavia/?utm\_source=tuicool&utm\_medium=referral][http_lingxiankong.github.io_blog_2016_03_30_octavia_utm_source_tuicool_utm_medium_referral]

写的比我更详细一点

目前官方不提供安装文档。谷歌了下似乎也没人写过具体的安装步骤，只推荐用devstack来进行安装。本人尝试根据devstack的安装脚本总结了下安装octavia的步骤，验证是成功的，不当之处请各位指正。

一 安装

1、创建数据库

mysql> CREATE DATABASE octavia;
    mysql> GRANT ALL PRIVILEGES ON octavia.* TO 'octavia'@'localhost'  IDENTIFIED BY 'OCTAVIA_DBPASS';mysql> GRANT ALL PRIVILEGES ON octavia.* TO 'octavia'@'%' \  IDENTIFIED BY 'OCTAVIA_DBPASS';

2 创建用户 角色 endpoint

openstack user create --domain default --password-prompt octavia
    openstack role add --project service --user cinder admin
    openstack endpoint create octavia public http://10.1.65.58:9876/ --region RegionOne 
    openstack endpoint create octavia admin http://10.1.65.58:9876/ --region RegionOne 
    openstack endpoint create octavia internal http://10.1.65.58:9876/ --region RegionOne

3 安装软件包

yum install openstack-octavia-worker openstack-octavia-api python-octavia openstack-octavia openstack-octavia openstack-octavia

4 导入镜像 镜像是从devstack 生成的系统中导出来的

openstack  image create amphora-x64-haproxy --public --container-format=bare --disk-format qcow2

5 创建管理网络，并在主机创建ovs端口，使octavia-worker，octavia-housekeeping，octavia-health-manager能和生成的虚拟机实例通讯

5.1 生成管理网络，网段

openstack network create lb-mgmt-net
    openstack subnet create --subnet-range 192.168.0.0/24 --allocation-pool       start=192.168.0.2,end=192.168.0.200 --network lb-mgmt-net lb-mgmt-subnet

5.2 生成管理端口防火墙规则

5555端口是管理网络，考虑到octavia组件尚不成熟，开启了22端口，镜像本身也是开启了22端口，这点吐槽下trove，同样是不成熟的模块，默认不开启22端口，还得去改源码。

openstack security group create lb-mgmt-sec-grp
    openstack security group rule create --protocol udp --dst-port 5555 lb-health-mgr-sec-grp
    openstack security group rule create --protocol tcp --dst-port 22 lb-mgmt-sec-grp

5.3 在管理网络创建一个端口用于连接宿主机中的octavia health\_manager

neutron port-create --name octavia-health-manager-standalone-listen-port --security-group lb-health-mgr-sec-grp --device-owner Octavia:health-mgr --binding:host_id=controller lb-mgmt-net

5.4 创建宿主机的ovs端口 并连接至5.1生成的网络

ovs-vsctl  --may-exist add-port br-int o-hm0 -- set Interface o-hm0 type=internal -- set Interface o-hm0 external-ids:iface-status=active -- set Interface o-hm0 external-ids:attached-mac=fa:16:3e:6f:9f:9a -- set Interface o-hm0 external-ids:iface-id=457e4953-b2d6-49ee-908b-2991506602b2

其中iface-id 和attached-mac 为 5.3生成的port的 属性

ip link set dev o-hm0 address fa:16:3e:6f:9f:9a

5.5 在宿主机上创建dhcp （为啥不用传统的dnsmasq呢？）

dhclient -v o-hm0 -cf /etc/octavia/dhcp/dhclient.conf

6 配置修改，和其他openstack组件设置差不多

6.1 设置数据库

[database]
    connection = mysql+pymysql://octavia:OCTAVIA_DBPASS@controller/octavia

6.2 设置消息队列

[oslo_messaging_rabbit]
    rabbit_host = controller
    rabbit_userid = openstack
    rabbit_password = RABBIT_PASS

6.3 设置 keystone的认证信息

[keystone_authtoken]
    auth_version = 2
    admin_password = OCTAVIA_PASS
    admin_tenant_name = octavia
    admin_user = octavia
    auth_uri = http://controller:5000/v2.0

6.4 设置health\_manager组件监听地址，此ip地址等于5.3中创建的io地址

[health_manager]
    bind_port = 5555
    bind_ip = 192.168.0.7
    controller_ip_port_list = 192.168.0.7:5555

6.5 设置和虚拟机通信的 公钥私钥

[haproxy_amphora]
    server_ca = /etc/octavia/certs/ca_01.pem
    client_cert = /etc/octavia/certs/client.pem
    key_path = /etc/octavia/.ssh/octavia_ssh_key
    base_path = /var/lib/octavia
    base_cert_dir = /var/lib/octavia/certs
    connection_max_retries = 1500
    connection_retry_interval = 1

6.6 设置 用于开启虚拟机实例的信息

[controller_worker]
    amp_boot_network_list = 826be4f4-a23d-4c5c-bff5-7739936fac76 # 步骤5.1中生成的id
    amp_image_tag = amphora # 步骤4 中已经定义这个metadata
    amp_secgroup_list = d949202b-ba09-4003-962f-746ae75809f7 # 步骤5.2 生成的安全组id
    amp_flavor_id = dd49b3d5-4693-4407-a76e-2ca95e00a9ec
    amp_image_id = b23dda5f-210f-40e6-9c2c-c40e9daa661a #步骤4中生成的镜像id
    amp_ssh_key_name = 155 #
    amp_active_wait_sec = 1
    amp_active_retries = 100
    network_driver = allowed_address_pairs_driver
    compute_driver = compute_nova_driver
    amphora_driver = amphora_haproxy_rest_driver

7 修改neutron配置

7.1 修改 /etc/neutron/neutron.conf 增加lbaas服务

service_plugins = [existing service plugins],neutron_lbaas.services.loadbalancer.plugin.LoadBalancerPluginv2

7.2 在\[service\_providers\] 章节 设置lbaas 的服务提供者为octavia

service_provider = LOADBALANCERV2:Octavia:neutron_lbaas.drivers.octavia.driver.OctaviaDriver:default

8 启动服务

如果之前 开启了 LBaaS v2 with an agent 服务 请关闭，并清理下neutron数据库下 lbaas\_loadbalancers lbaas\_loadbalancer\_statistics 不然会报错

同步数据库

octavia-db-manage   upgrade head

重启neutron

systemctl restart neutron-server

启动octavia

systemctl restart  octavia-housekeeping  octavia-worker octavia-api octavia-health-manager

二 验证

9.1 创建loadbalancer

9.2 查看虚拟机，值得注意的地方，loadbalancer的地址是vip，和虚拟机的地址是不相同的

9.3 创建linstener

neutron lbaas-listener-create --name test-lb-tcp --loadbalancer test-lb-1 --protocol TCP  --protocol-port 22

9.4 设置安全组

neutron port-update  --security-group default d163b73c-258a-4e03-90ad-5db31cfe23ac

9.5 创建pool,新建三台虚拟机，并加入pool

openstack server create  --flavor m1.small --nic net-id=22525640-297e-40eb-bd77-0a9afd861f8c --image "cirros for kvm raw"  --min 3 --max 3 test
    
    [root@controller ~]# openstack server list |grep test-
    | d8dc22d4-e657-4c54-96f9-3a53ca67533d | test-3                                       | ACTIVE  | lbtest=192.168.123.8                                 | cirros for kvm raw                  |
    | c7926665-84c5-48a5-9de5-5e15e71baa5d | test-2                                       | ACTIVE  | lbtest=192.168.123.13                                | cirros for kvm raw                  |
    | fcf60c23-b799-4d08-a5a7-2b0fc9f1905e | test-1                                       | ACTIVE  | lbtest=192.168.123.11                                | cirros for kvm raw                  |
    
    neutron lbaas-pool-create   --name test-lb-pool-tcp  --lb-algorithm ROUND_ROBIN --listener test-lb-tcp --protocol TCP
     
    for i in {8,13,11}
    do
    neutron lbaas-member-create --subnet lbtest  --address 192.168.123.${i}  --protocol-port 22  test-lb-pool-tcp
    done

9.6 验证

[root@controller ~]# >/root/.ssh/known_hosts;ip netns exec qrouter-4718cc34-68cc-47a7-9201-405d1fc09213 ssh cirros@192.168.123.10 "hostname"
    The authenticity of host '192.168.123.10 (192.168.123.10)' can't be established.
    RSA key fingerprint is 72:c4:11:41:53:51:f2:1b:b5:e6:1b:69:a8:c2:5b:d4.
    Are you sure you want to continue connecting (yes/no)? yes
    Warning: Permanently added '192.168.123.10' (RSA) to the list of known hosts.
    cirros@192.168.123.10's password: 
    test-3
    [root@controller ~]# >/root/.ssh/known_hosts;ip netns exec qrouter-4718cc34-68cc-47a7-9201-405d1fc09213 ssh cirros@192.168.123.10 "hostname"
    The authenticity of host '192.168.123.10 (192.168.123.10)' can't be established.
    RSA key fingerprint is 3d:88:0f:4a:b1:77:c9:6a:fd:82:4d:31:0c:ca:82:d8.
    Are you sure you want to continue connecting (yes/no)? yes
    Warning: Permanently added '192.168.123.10' (RSA) to the list of known hosts.
    cirros@192.168.123.10's password: 
    test-1
    [root@controller ~]# >/root/.ssh/known_hosts;ip netns exec qrouter-4718cc34-68cc-47a7-9201-405d1fc09213 ssh cirros@192.168.123.10 "hostname"
    The authenticity of host '192.168.123.10 (192.168.123.10)' can't be established.
    RSA key fingerprint is 1c:03:f0:f9:92:a7:0f:5d:9d:09:22:14:94:62:e4:c4.
    Are you sure you want to continue connecting (yes/no)? yes
    Warning: Permanently added '192.168.123.10' (RSA) to the list of known hosts.
    cirros@192.168.123.10's password: 
    test-2

三 过程分析

10.1 worker的相关操作

创建 云主机实例，关联到管理网络：

REQ: curl -g -i -X POST http://controller:8774/v2.1/9a4b2de78c2d45cfbf6880dd34877f7b/servers -H "User-Agent: python-novaclient" -H "Content-Type: application/json" -H "Accept: application/json" -H "X-OpenStack-Nova-API-Version: 2.1" -H "X-Auth-Token: {SHA1}0f810ab0fdd5b92489f73a7f0988adfc9da4e517" -d '{"server": {"name": "amphora-4f22d55b-0680-4111-aef6-da98c9ccd1d4", "imageRef": "b23dda5f-210f-40e6-9c2c-c40e9daa661a", "key_name": "155", "flavorRef": "dd49b3d5-4693-4407-a76e-2ca95e00a9ec", "max_count": 1, "min_count": 1, "personality": [{"path": "/etc/octavia/amphora-agent.conf", "contents": ""}, {"path": "/etc/octavia/certs/client_ca.pem", "contents": "="}, {"path": "/etc/octavia/certs/server.pem", "contents": ""}], "networks": [{"uuid": "826be4f4-a23d-4c5c-bff5-7739936fac76"}], "security_groups": [{"name": "d949202b-ba09-4003-962f-746ae75809f7"}], "config_drive": true}}' _http_log_request /usr/lib/python2.7/site-packages/keystoneauth1/session.py:337

当检测到目标云主机的管理网络状态变为active后进行下一步

REQ: curl -g -i -X GET http://controller:8774/v2.1/9a4b2de78c2d45cfbf6880dd34877f7b/servers/d3c97360-56b2-4f75-b905-2ef83870a342/os-interface -H "User-Agent: python-novaclient" -H "Accept: application/json" -H "X-OpenStack-Nova-API-Version: 2.1" -H "X-Auth-Token: {SHA1}3f6ccac4cb8b70b06fb5e62b9db2272702d8ec67" _http_log_request /usr/lib/python2.7/site-packages/keystoneauth1/session.py:337
    2016-10-17 12:06:30.041 29993 DEBUG novaclient.v2.client [-] RESP: [200] Content-Length: 286 Content-Type: application/json Openstack-Api-Version: compute 2.1 X-Openstack-Nova-Api-Version: 2.1 Vary: OpenStack-API-Version, X-OpenStack-Nova-API-Version X-Compute-Request-Id: req-ccc07b37-e942-4a5b-a87a-b0e8d3887ba3 Date: Mon, 17 Oct 2016 04:06:30 GMT Connection: keep-alive
    RESP BODY: {"interfaceAttachments": [{"port_state": "ACTIVE", "fixed_ips": [{"subnet_id": "4e3409e5-4e9a-4599-9b2e-f760b2fab380", "ip_address": "192.168.0.11"}], "port_id": "bbf99a69-0fb2-42a6-b7de-b7969bda9d73", "net_id": "826be4f4-a23d-4c5c-bff5-7739936fac76", "mac_addr": "fa:16:3e:01:04:2c"}]}

2016-10-17 12:06:30.078 29993 DEBUG octavia.controller.worker.tasks.amphora_driver_tasks [-] Finalized the amphora. execute /usr/lib/python2.7/site-packages/octavia/controller/worker/tasks/amphora_driver_tasks.py:164

创建对外服务的vip的端口

2016-10-17 12:06:30.226 29993 DEBUG octavia.controller.worker.controller_worker [-] Task 'octavia.controller.worker.tasks.network_tasks.AllocateVIP' (af8ea5a0-42c8-4d30-9ffa-016668811fc8) transitioned into state 'RUNNING' from state 'PENDING' _task_receiver /usr/lib/python2.7/site-packages/taskflow/listeners/logging.py:189
    2016-10-17 12:06:30.227 29993 DEBUG octavia.controller.worker.tasks.network_tasks [-] Allocate_vip port_id c7d7b552-83ac-4e0c-84bf-0b9cae661eab, subnet_id 74aea53a-014a-4f9c-86f9-805a2a772a27,ip_address 192.168.123.31 execute /usr/lib/python2.7/site-packages/octavia/controller/worker/tasks/network_tasks.py:328

在该vip下面创建一个实际的port 并把该port   attach至 云主机

2016-10-17 12:06:32.662 29993 DEBUG octavia.network.drivers.neutron.allowed_address_pairs [-] Created vip port: 1627d28d-bf54-46eb-9d78-410c5d647bf4 for amphora: 3f6e22a1-e0b0-4098-ba20-daf47cfdae19 _plug_amphora_vip /usr/lib/python2.7/site-packages/octavia/network/drivers/neutron/allowed_address_pairs.py:97
    2016-10-17 12:06:32.663 29993 DEBUG novaclient.v2.client [-] REQ: curl -g -i -X POST http://controller:8774/v2.1/9a4b2de78c2d45cfbf6880dd34877f7b/servers/d3c97360-56b2-4f75-b905-2ef83870a342/os-interface -H "User-Agent: python-novaclient" -H "Content-Type: application/json" -H "Accept: application/json" -H "X-OpenStack-Nova-API-Version: 2.1" -H "X-Auth-Token: {SHA1}3f6ccac4cb8b70b06fb5e62b9db2272702d8ec67" -d '{"interfaceAttachment": {"port_id": "1627d28d-bf54-46eb-9d78-410c5d647bf4"}}' _http_log_request /usr/lib/python2.7/site-packages/keystoneauth1/session.py:337

创建listener

2016-10-17 19:01:09.384 29993 DEBUG octavia.amphorae.drivers.haproxy.rest_api_driver [-] request url https://192.168.0.9:9443/0.5/listeners/c3a1867c-b2e5-49a7-819b-7a7d39063dda/reload request /usr/lib/python2.7/site-packages/octavia/amphorae/drivers/haproxy/rest_api_driver.py:248
    2016-10-17 19:01:09.412 29993 DEBUG octavia.amphorae.drivers.haproxy.rest_api_driver [-] Connected to amphora. Response: <Response [202]> request /usr/lib/python2.7/site-packages/octavia/amphorae/drivers/haproxy/rest_api_driver.py:270
    2016-10-17 19:01:09.414 29993 DEBUG octavia.controller.worker.controller_worker [-] Task 'octavia.controller.worker.tasks.amphora_driver_tasks.ListenersUpdate' (0f588287-a383-4c70-9847-20187dd19f9f) transitioned into state 'SUCCESS' from state 'RUNNING' with result 'None' _task_receiver /usr/lib/python2.7/site-packages/taskflow/listeners/logging.py:178

10.2

octavia agent分析

在9443端口创建监听端口给worker和health-manager 访问

2016-10-17 12:10:41.344 1043 INFO werkzeug [-]  * Running on https://0.0.0.0:9443/ (Press CTRL+C to quit)

octavia agent的似乎有bug，不显示debug信息。

11 高可用测试

将/etc/octavia/octavia.conf配置文件中的loadbalancer\_topology = SINGLE 改成 ACTIVE\_STANDBY 可以启用高可用模式，目前不持双ACTIVE

生成loadbalancer之后，可以看到生成两个虚拟机

\[root@controller octavia\]\# neutron lbaas-loadbalancer-create --name test-lb1238 lbtest2

看到3个ip:vip是192.168.235.14,两台机器出口

ip是192.168.235.10和192.168.235.19

登陆虚拟机验证一下，注意登陆的ip是管理网络的ip：

[root@controller ~]# ssh 192.168.0.8 "ps -ef |grep keepalived; cat  /var/lib/octavia/vrrp/octavia-keepalived.conf"
    root      1868     1  0 04:40 ?        00:00:00 /usr/sbin/keepalived -D -d -f /var/lib/octavia/vrrp/octavia-keepalived.conf
    root      1869  1868  0 04:40 ?        00:00:00 /usr/sbin/keepalived -D -d -f /var/lib/octavia/vrrp/octavia-keepalived.conf
    root      1870  1868  0 04:40 ?        00:00:00 /usr/sbin/keepalived -D -d -f /var/lib/octavia/vrrp/octavia-keepalived.conf
    root      5448  5377  0 05:00 ?        00:00:00 bash -c ps -ef |grep keepalived; cat  /var/lib/octavia/vrrp/octavia-keepalived.conf
    root      5450  5448  0 05:00 ?        00:00:00 grep keepalived
    vrrp_script check_script {
      script /var/lib/octavia/vrrp/check_script.sh
      interval 5
      fall 2
      rise 2
    }
    vrrp_instance 4e43f3c7c0f644c78dabe2fc8ed16e0f {
     state MASTER
     interface eth1
     virtual_router_id 1
     priority 100
     nopreempt
     garp_master_refresh 5
     garp_master_refresh_repeat 2
     advert_int 1
     authentication {
      auth_type PASS
      auth_pass ee46125
     }
     unicast_src_ip 192.168.235.10
     unicast_peer {
           192.168.235.19
     }
     virtual_ipaddress {
      192.168.235.14
     }
     track_script {
        check_script
     }
    }
    [root@controller ~]# ssh 192.168.0.8 "ps -ef |grep haproxy; cat  /var/lib/octavia/836053f0-ea72-46ae-9fae-8b80153ef593/haproxy.cfg"
    nobody    2195     1  0 04:43 ?        00:00:00 /usr/sbin/haproxy -f /var/lib/octavia/836053f0-ea72-46ae-9fae-8b80153ef593/haproxy.cfg -L jrwLnRhlvXcPd21JhvXEMStRHh0 -p /var/lib/octavia/836053f0-ea72-46ae-9fae-8b80153ef593/836053f0-ea72-46ae-9fae-8b80153ef593.pid -sf 2154
    root      6745  6676  0 05:06 ?        00:00:00 bash -c ps -ef |grep haproxy; cat  /var/lib/octavia/836053f0-ea72-46ae-9fae-8b80153ef593/haproxy.cfg
    root      6747  6745  0 05:06 ?        00:00:00 grep haproxy
    # Configuration for test-lb1238
    global
        daemon
        user nobody
        group nogroup
        log /dev/log local0
        log /dev/log local1 notice
        stats socket /var/lib/octavia/836053f0-ea72-46ae-9fae-8b80153ef593.sock mode 0666 level user
    
    defaults
        log global
        retries 3
        option redispatch
        timeout connect 5000
        timeout client 50000
        timeout server 50000
    
    peers 836053f0ea7246ae9fae8b80153ef593_peers
        peer 3OduZJiPzm475Q7IgyshE5oq1Jk 192.168.235.19:1025
        peer jrwLnRhlvXcPd21JhvXEMStRHh0 192.168.235.10:1025
    
    
    frontend 836053f0-ea72-46ae-9fae-8b80153ef593
        option tcplog
        bind 192.168.235.14:22
        mode tcp
        default_backend 457d4de5-3213-4969-8f20-1f2d3505ff1e
    
    backend 457d4de5-3213-4969-8f20-1f2d3505ff1e
        mode tcp
        balance leastconn
        timeout check 5
        server fa28676f-a762-4a8e-91ab-7a83f071b62b 192.168.235.20:22 weight 1 check inter 5s fall 3 rise 3
        server 1ded44da-cba5-434c-8578-95153656c392 192.168.235.24:22 weight 1 check inter 5s fall 3 rise 3

另一台结果结果类似。

结论：octavia的高可用是通过haproxy加keepalived来完成的。

四、其他

1、在services\_lbaas.conf下有个选项

\[octavia\]

request\_poll\_timeout = 200

此选项的定义了，创建loadbalancer之后，当超过这个时间以后，如果octavia还没有的状态没有变成active，neutron就会把这个loadbalancer设置为error，默认值是100，在我的环境下高可用模式会来不及。日志如下：

2016-10-19 09:38:26.392 6256 DEBUG neutron_lbaas.drivers.octavia.driver [req-bee3619a-f9d4-4463-adcd-3cb99826b600 - - - - -] Octavia reports load balancer 2676dac6-c41d-4501-9c41-781a176c6baf has provisioning status of PENDING_CREATE thread_op /usr/lib/python2.7/site-packages/neutron_lbaas/drivers/octavia/driver.py:75
    2016-10-19 09:38:29.393 6256 DEBUG neutron_lbaas.drivers.octavia.driver [req-bee3619a-f9d4-4463-adcd-3cb99826b600 - - - - -] Timeout has expired for load balancer 2676dac6-c41d-4501-9c41-781a176c6baf to complete an operation.  The last reported status was PENDING_CREATE thread_op /usr/lib/python2.7/site-packages/neutron_lbaas/drivers/octavia/driver.py:94

2、源代码小修改例子：

当neutron的loadbalancer状态发生变成active或者error时候时候推送到报警系统

修改/usr/lib/python2.7/site-packages/neutron\_lbaas/drivers/octavia/driver.py

if prov_status == 'ACTIVE' or prov_status == 'DELETED':
                kwargs = {'delete': delete}
                if manager.driver.allocates_vip and lb_create:
                    kwargs['lb_create'] = lb_create
                    # TODO(blogan): drop fk constraint on vip_port_id to ports
                    # table because the port can't be removed unless the load
                    # balancer has been deleted.  Until then we won't populate the
                    # vip_port_id field.
                    # entity.vip_port_id = octavia_lb.get('vip').get('port_id')
                    entity.vip_address = octavia_lb.get('vip').get('ip_address')
                manager.successful_completion(context, entity, **kwargs)
                if prov_status == 'ACTIVE':
                  urllib2.urlopen('http://********')
                  LOG.debug("report  status to******* {0}{1}".format(entity.root_loadbalancer.id, prov_status))
                return
            elif prov_status == 'ERROR':
                manager.failed_completion(context, entity)
                urllib2.urlopen('http://*******')
                LOG.debug("report status to ******* {0}{1}".format(entity.root_loadbalancer.id, prov_status))
                return

3、octavia的数据库和neutron不是同一张表，但是里面有很多数据要求要保持一致，一定要保持两者相关数据的同步，不同步的话会带来很多问题，亲身经历。

[http_lingxiankong.github.io_blog_2016_03_30_octavia_utm_source_tuicool_utm_medium_referral]: http://lingxiankong.github.io/blog/2016/03/30/octavia/?utm_source=tuicool&utm_medium=referral