docker swarm auto scale

1. 主旨介绍

使用mac和fusion vm两台主机搭建一个小的docker集群, 采用docker swarm方式搭建. Swarm基本概念:

• manager:用来管理swarm服务
• node: docker enginer节点, 通过manager可以在node上操作, manager本身也是node
• worker: 非管理的node, 一般是实际部署和运行container的docker节点

目的是是探索使用swarm完成动态伸缩的可行性.

2. 操作步骤

• swarm init(vm中操作,因为docker for mac有网络问题 manager)

  docker swarm init --advertise-addr 192.168.132.148
  Swarm initialized: current node (b1uem2hg4y2ojnauvre3x2rdw) is now a manager.
  
  To add a worker to this swarm, run the following command:
  
      docker swarm join --token SWMTKN-1-1snrbxsenp4v1npcusox9l3n9jzsa4n8jbkvlhjc19e3z89c8a-7vfrbr1o1e1sr2g2r6f97xhne 192.168.132.148:2377
  
  To add a manager to this swarm, run 'docker swarm join-token manager' and follow the instructions.
  

  初始化swarm之后可以使用docker node ls查看信息:

  docker node ls
  ID                            HOSTNAME                STATUS              AVAILABILITY        MANAGER STATUS      ENGINE VERSION
  b1uem2hg4y2ojnauvre3x2rdw *   linuxkit-025000000001   Ready               Active              Leader              18.06.1-ce            18.06.1-ce
  

• join node(worker)

  docker swarm join --token SWMTKN-1-3h1rzm2cdu6ms935f6kysy4rhzs54mqofnhr1thlwu0fiewqfq-9q6fwc7eqk70zwb1kwi2vdklj 192.168.132.148:2377
  This node joined a swarm as a worker.
  

• deploy service 部署一个alpine容器, 运行ping命令视为内部服务

  docker service create --replicas 1 --name helloword alpine ping docker.com
  vyw5nezmiv37j3fk5kgu4231x
  overall progress: 1 out of 1 tasks 
  1/1: running   [==================================================>]
  

  查看结果如下, 可以发现一个alpine被部署到worker节点(vm是manager机器名是unbuntu所以下面的linuxkit-xxxx是worker, 即mac server)

  docker service ps helloword
  ID                  NAME                IMAGE               NODE                    DESIRED STATE       CURRENT STATE           ERROR               PORTS
  ht9npghzzux9        helloword.1         alpine:latest       linuxkit-025000000001   Running             Running 4 minutes ago 
  

• scale service 横向扩展服务为2个实例

  docker service scale helloword=2
  helloword scaled to 2
  overall progress: 1 out of 2 tasks 
  1/2: running   [==================================================>] 
  overall progress: 2 out of 2 tasks 
  1/2: running   [==================================================>] 
  2/2: running   [==================================================>] 
  verify: Service converged 
  

  查看扩展的结果如下, 可以发现两个实例被部署到worker和manager, 因为worker不够用, 所以使用了manager, 也可以设置manager只是管理, 不运行task

  docker service ps helloword
  ID                  NAME                IMAGE               NODE                    DESIRED STATE       CURRENT STATE            ERROR               PORTS
  ht9npghzzux9        helloword.1         alpine:latest       linuxkit-025000000001   Running             Running 5 minutes ago                        
  v0i6ghtx8eed        helloword.2         alpine:latest       ubuntu                  Running             Running 26 seconds ago
  

3. 总结

使用swarm集群可以优雅的进行docker部署和扩展, 结合docker-compose操作会更加方便, 如果增加了产线的流量监控+服务发现, 就可以实现产线环境的动态伸缩部署, 很棒!!

4. 问题和解答

• unavailable desc = all SubConns are in TransitentFailure

  这个问题以为manager是mac, docker for mac 因为网络的问题不能支持work去链接到mac docker swarm manager

• Error response from daemon: manager stopped: failed to listen on remote API address: listen tcp 10.0.1.34:2377: bind: cannot assign requested address

  这是因为指定了 --listen-addr和--advertise-addr ip一样但是端口默认也是一直的, 去除listen-addr或者指定不同的端口即可

• 结束swarm

  在manager服务器运行: docker swarm leave --force, 这样manager就变成inactive了, 可以重新操作执行init或者其他

5. 参考资料

• Getting started with swarm mode
• Add nodes to the swarm
• Deploy a service to the swarm
• Scale the service in the swarm

来源：oschina

链接：https://my.oschina.net/u/2623130/blog/2872189