OpenStack Heat AutoScaling详解及实例代码

openstack heat autoscaling

一、背景

openstack的heat是在h版之后加入的组件，旨在创建一套业务流程，更轻松的管理一个集群。集群内的虚拟机可以作为一个整体，统一的为客户提供服务。heat中把功能定义成资源，在heat中会用到nova，neutron，ceilometer等组件，这些都可以看成是资源，通过模板文件来描述，模板文件可以是yaml格式，也可以是json格式，一般是yaml格式。
autoscaling的概念最早出现在aws，autoscaling是一项web服务，目的是根据用户定义的策略，时间表的运行状态检查启动或终止虚拟机，达到自动伸缩。

openstack里的auto scale是由heat和ceilometer模块一起配合完成的。ceilometer负责收集处理性能数据，一旦达到heat模版里定义的阀值，就发告警信息给heat-engine，由heat-engine调动heat模版里定义的其它的openstack资源实现auto scale。

二、heat autoscaling resources

实现autoscaling功能涉及到的资源如下：

1.aws::autoscaling::autoscalinggroup

伸缩组是具有相同应用场景的实例的集合，定义了组内实例数的最大值和最小值，冷却时间等等。
注：冷却时间是指一个伸缩活动后的一段锁定时间，在这个时间内不能进行其他的伸缩活动。

语法如下：

{
 "type" : "aws::autoscaling::autoscalinggroup",
 "properties" : {
  "availabilityzones" : [ string, ... ],
  "cooldown" : string,
  "desiredcapacity" : string,
  "healthcheckgraceperiod" : integer,
  "healthchecktype" : string,
  "instanceid" : string,
  "launchconfigurationname" : string,
  "loadbalancernames" : [ string, ... ],
  "maxsize" : string,
  "metricscollection" : [ metricscollection, ... ]
  "minsize" : string,
  "notificationconfigurations" : [ notificationconfigurations, ... ],
  "placementgroup" : string,
  "tags" : [ auto scaling tag, ..., ],
  "targetgrouparns" : [ string, ... ],
  "terminationpolicies" : [ string, ..., ],
  "vpczoneidentifier" : [ string, ... ]
 }
}

2.aws::autoscaling::launchconfiguration

伸缩配置定义了用于弹性伸缩的实例的配置。由autoscalinggroup用于配置组内的实例。

语法如下：

{
  "type" : "aws::autoscaling::launchconfiguration",
  "properties" : {
   "associatepublicipaddress" : boolean,
   "blockdevicemappings" : [ blockdevicemapping, ... ],
   "classiclinkvpcid" : string,
   "classiclinkvpcsecuritygroups" : [ string, ... ],
   "ebsoptimized" : boolean,
   "iaminstanceprofile" : string,
   "imageid" : string,
   "instanceid" : string,
   "instancemonitoring" : boolean,
   "instancetype" : string,
   "kernelid" : string,
   "keyname" : string,
   "placementtenancy" : string,
   "ramdiskid" : string,
   "securitygroups" : [ securitygroup, ... ],
   "spotprice" : string,
   "userdata" : string
  }
}

3.aws::autoscaling::scalingpolicy

为auto scale group添加伸缩的策略，定义了具体的扩展或者收缩的操作，以及伸缩的数量。

语法如下：

{
  "type" : "aws::autoscaling::scalingpolicy",
  "properties" : {
   "adjustmenttype" : string,
   "autoscalinggroupname" : string,
   "cooldown" : string,
   "estimatedinstancewarmup" : integer,
   "metricaggregationtype" : string,
   "minadjustmentmagnitude" : integer,
   "policytype" : string,
   "scalingadjustment" : integer,
   "stepadjustments" : [ stepadjustments, ... ]
  }
}   

此外，heat中autoscaling还需配合os::ceilometer::alarm使用，由alarm监控实例的运行状况，一旦超过阈值，则会产生告警。

三、 heat autoscaling template

下面是一个简单的例子:

heat_template_version: 2013-05-23
description: heat template for autoscaling
parameters:#定义一些变量
 flavor:
  type: string
  default: m1.small
 image:
  type: string
  default: 1a2b3c4f-1a2b-3c4f-5d6e-4130ff5203de
 availability_zone:
  type: string
  default: nova
 alarm_scaleout_threshold:#阈值
  type: number
  default: 80
 alarm_scalein_threshold:#阈值
  type: number
  default: 20
resources:
 neutron_network:
  type: os::neutron::net
  properties:
   name: {get_param: "os::stack_name"}

 neutron_subnet:
  type: os::neutron::subnet
  properties:
   name: {get_param: "os::stack_name"}
   network_id: { get_resource: neutron_network }
   cidr: '192.168.111.0/24'
   gateway_ip: '192.168.111.1'
   allocation_pools:
    - start: '192.168.111.2'
     end: '192.168.111.254'
 neutron_router:
  type: os::neutron::router
  properties:
   name: {get_param: "os::stack_name"}
 add_router_interface:
  type: os::neutron::routerinterface
  properties:
   router_id: { get_resource: neutron_router }
   subnet_id: { get_resource: neutron_subnet }
 nova_server_security_group:
  type: os::neutron::securitygroup
  properties:
   description: 'security group for vm'
   name: {get_param: "os::stack_name"}
   rules: [
    {direction: 'ingress',
     remote_ip_prefix: '0.0.0.0/0',
     port_range_min: 0,
     port_range_max: 30000,
     ethertype: ipv4,
     protocol: 'tcp'},
    {direction: 'egress',
     remote_ip_prefix: '0.0.0.0/0',
     port_range_min: 0,
     port_range_max: 65535,
     ethertype: 'ipv4',
     protocol: 'tcp'},
    {direction: 'egress',
     remote_ip_prefix: '0.0.0.0/0',
     port_range_min: 0,
     port_range_max: 65535,
     ethertype: 'ipv4',
     protocol: 'udp'},
    {direction: 'ingress',
     remote_ip_prefix: '0.0.0.0/0',
     port_range_min: null,
     port_range_max: null,
     ethertype: 'ipv4',
     protocol: 'icmp'},
    {direction: egress,
     remote_ip_prefix: '0.0.0.0/0',
     port_range_min: null,
     port_range_max: null,
     ethertype: 'ipv4',
     protocol: 'icmp'}
   ]
 launch_config:#scale group中的实例的配置
  type: aws::autoscaling::launchconfiguration
  properties:
   imageid: { get_param: image }#实例使用的image
   instancetype: { get_param: flavor }#实例使用的flavor
   securitygroups: [ get_resource: nova_server_security_group ]
   userdata: |#实例启动时运行的脚本
     #!/bin/bash
     passwd root << eod
     123456
     123456
     eod
 server_group:#伸缩组
  type: aws::autoscaling::autoscalinggroup
  properties:
   availabilityzones: []
   cooldown: '60'#冷却时间
   launchconfigurationname: { get_resource: launch_config }#组中实例的配置
   minsize: '1'#最小实例数
   maxsize: '4'#最大实例数
   vpczoneidentifier: [ get_resource: neutron_subnet ]
 scaleout_policy:#向上扩展的策略
  type: aws::autoscaling::scalingpolicy
  properties:
   adjustmenttype: changeincapacity
#heat 支持三种调整方式：change_in_capacity (new = current + adjustment), #exact_capacity (new = adjustment), percent_change_in_capacity (在current 的基#础上上按照 adjustment 的 百分比调整)
   autoscalinggroupname: { get_resource: server_group }
   scalingadjustment: '1'#每次的调整量，即增加一个实例
 scalein_policy:#向下收缩的策略
  type: aws::autoscaling::scalingpolicy
  properties:
   adjustmenttype: changeincapacity
   autoscalinggroupname: { get_resource: server_group }
   scalingadjustment: '-1'#每次的调整量，即减少一个实例
 neutron_port:
  type: os::neutron::port
  properties:
   network_id: { get_resource: neutron_network }
   fixed_ips:
    - subnet_id: { get_resource: neutron_subnet }
   security_groups: [ { get_resource: nova_server_security_group } ]
 alarm_scaleout: #定义一个 ceilometer alarm
  type: os::ceilometer::alarm
  properties:
   description: scale-up if the average cpu > 80% for 10 minute
   meter_name: cpu_util #监控虚拟机的 cpu_util
   statistic: avg #statistic 的计算方法为 avg 即平均值法
   period: 600 #统计周期
   evaluation_periods: 1 #连续几个周期才算有效
   repeat_actions: true
   threshold: { get_param: alarm_scaleout_threshold }# cpu_util 的阈值
   alarm_actions: #该告警在alarm 状态时的 action。
    - {get_attr: [scaleout_policy, alarmurl]}
   matching_metadata: {'metadata.user_metadata.groupname': {get_resource: 'server_group'}}
   comparison_operator: gt #检测值和阈值的比较方式为 gt 即大于
 alarm_scalein:
  type: os::ceilometer::alarm
  properties:
   description: scale-down if the average cpu < 20% for 10 minutes
   meter_name: cpu_util
   statistic: avg
   period: 600
   evaluation_periods: 1
   repeat_actions: true
   threshold: { get_param: alarm_scalein_threshold }
   alarm_actions:
    - {get_attr: [scalein_policy, alarmurl]}
   matching_metadata: {'metadata.user_metadata.groupname': {get_resource: 'server_group'}}
   comparison_operator: lt#检测值和阈值的比较方式为 lt 即小于
outputs:
 scale_in_url:
  value: { get_attr: [ scalein_policy, alarmurl ] }
 scale_out_url:
  value: { get_attr: [ scaleout_policy, alarmurl ] }

这个stack的功能是监控实例的cpu使用率,当cpu使用率大于80%时,将会启动一个新的实例,当cpu使用率小于20%,将会减少一个实例。

感谢阅读，希望能帮助到大家，谢谢大家对本站的支持！