docker 26 内核要求

在Docker1.12版本中，一个大的功能点是swarm集群（基于swarmkit项目），通过docker命令可以直接实现docker-engine相互发现，并组建成为一个容器集群。有关集群的docker命令如下：

• （1）docker swarm：集群管理，子命令有init, join, leave, update
• （2）docker service：服务创建，子命令有create, inspect, update, remove, tasks
• （3）docker node：节点管理，子命令有accept, promote, demote, inspect, update, tasks, ls, rm
• （4）docker stack/deploy：试验特性，用于多应用部署 下面在一个三节点环境中，针对各个命令进行试用，详细展开说明

1、swarm

swarm命令用于集群管理，子命令如下：

[root@host1 ~]# docker swarm --help
 
Usage:  dockerswarmCOMMAND
 
ManageDockerSwarm
 
Options:
      --help  Printusage
 
Commands:
  init        Initialize a Swarm.
  join        Join a Swarmas a nodeand/or manager.
  update      updatetheSwarm.
  leave      Leave a Swarm.
  inspect    InspecttheSwarm
 
Run 'docker swarm COMMAND --help' for moreinformationon a command.

可以看到，提供了init、join、update、leave、inspect命令

（1）swarm init命令用于初始化一个集群

[root@host1 ~]# docker swarm init --help
 
Usage:  dockerswarminit [OPTIONS]
 
Initialize a Swarm.
 
Options:
      --auto-acceptvalue  Autoacceptancepolicy (worker, manager, or none)
      --force-new-cluster  Forcecreate a new clusterfromcurrentstate.
      --help                Printusage
      --listen-addrvalue  Listenaddress (default 0.0.0.0:2377)
      --secretstring      Setsecretvalueneededto acceptnodesintocluster

在第一个节点上运行init命令初始化一个集群，初始化完成后会监听2377端口，接受其他节点的加入请求

[root@host1 ~]# docker swarm init
Swarminitialized: currentnode (6mjmescd8473lh6jpvmx2khkq) is now a manager.

（2）swarm join命令用于加入一个现有集群

[root@host1 ~]# docker swarm join --help
 
Usage:  dockerswarmjoin [OPTIONS] HOST:PORT
 
Join a Swarmas a nodeand/or manager.
 
Options:
      --ca-hashstring      HashoftheRootCertificateAuthoritycertificateusedfor trustedjoin
      --help                Printusage
      --listen-addrvalue  Listenaddress (default 0.0.0.0:2377)
      --manager            Try joiningas a manager.
      --secretstring      Secretfor nodeacceptance

在第二个、第三个节点上通过join命令加入集群

[root@host2 ~]# docker swarm join 186.100.40.90:2377
This nodejoined a Swarmas a worker.
[root@host3 ~]# docker swarm join 186.100.40.90:2377
This nodejoined a Swarmas a worker.

在第一个节点上，通过docker node ls查看节点信息

[root@host1 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
3xto6bh4htx7thj7b222datfd    host2  Accepted    Ready  Active                        
6mjmescd8473lh6jpvmx2khkq *  host1  Accepted    Ready  Active        Reachable      Yes
bwkk0n044fuod36dvghi9jbkv    host3  Accepted    Ready  Active

可以看到当前集群中有3个节点，Manager节点为host1，由于当前只有一个manager，所以这一节点也为是leader（swarmkit采用raft协议构建集群）

（3）swarm leave命令由于离开集群

（4）swarm inspect命令用于查看集群详细信息

[root@host1 ~]# docker swarm inspect
[
    {
        "ID": "90wkw2fmy0x5fjlh0mtuhtkrd",
        "Version": {
            "Index": 11
        },
        "CreatedAt": "2016-06-19T12:47:40.260925387Z",
        "UpdatedAt": "2016-06-19T12:47:40.665401298Z",
        "Spec": {
            "Name": "default",
            "AcceptancePolicy": {
                "Policies": [
                    {
                        "Role": "worker",
                        "Autoaccept": true
                    },
                    {
                        "Role": "manager",
                        "Autoaccept": false
                    }
                ]
            },
            "Orchestration": {
                "TaskHistoryRetentionLimit": 10
            },
            "Raft": {
                "SnapshotInterval": 10000,
                "LogEntriesForSlowFollowers": 500,
                "HeartbeatTick": 1,
                "ElectionTick": 3
            },
            "Dispatcher": {
                "HeartbeatPeriod": 5000000000
            },
            "CAConfig": {
                "NodeCertExpiry": 7776000000000000
            }
        }
    }
]

2、service

通过service命令可以管理一个服务，子命令如下：

[root@host1 ~]# docker service --help
 
Usage:  dockerserviceCOMMAND
 
ManageDockerservices
 
Options:
      --help  Printusage
 
Commands:
  create      Create a new service
  inspect    Inspect a service
  tasks      Listthetasksof a service
  ls          Listservices
  rm          Remove a service
  scale      Scaleoneor multipleservices
  update      Update a service
 
Run 'docker service COMMAND --help' for moreinformationon a command.

（1）service create命令用于创建一个服务[root@host1 ~]# docker service create –help

Usage:  dockerservicecreate [OPTIONS] IMAGE [COMMAND] [ARG...]
 
Create a new service
 
Options:
      --constraintvalue            Placementconstraints (default [])
      --endpoint-modestring        Endpointmode(Validvalues: VIP, DNSRR)
  -e, --envvalue                    Setenvironmentvariables (default [])
      --help                        Printusage
  -l, --labelvalue                  Servicelabels (default [])
      --limit-cpuvalue              LimitCPUs
      --limit-memoryvalue          LimitMemory
      --modestring                  Servicemode (replicatedor global) (default "replicated")
  -m, --mountvalue                  Attach a mountto theservice
      --namestring                  Servicename
      --networkvalue                Networkattachments (default [])
  -p, --publishvalue                Publish a portas a nodeport (default [])
      --replicasvalue              Numberoftasks (default none)
      --reserve-cpuvalue            ReserveCPUs
      --reserve-memoryvalue        ReserveMemory
      --restart-conditionstring    Restartwhenconditionis met (none, on_failure, or any)
      --restart-delayvalue          Delaybetweenrestartattempts (default none)
      --restart-max-attemptsvalue  Maximumnumberofrestartsbeforegivingup (default none)
      --restart-windowvalue        Windowusedto evalulatetherestartpolicy (default none)
      --stop-grace-periodvalue      Timeto waitbeforeforcekilling a container (default none)
      --update-delayduration        Delaybetweenupdates
      --update-parallelismuint      Maximumnumberoftasksupdatedsimultaneously (default 1)
  -u, --userstring                  Usernameor UID
  -w, --workdirstring              Workingdirectoryinsidethecontainer

在参数中有两个比较重要的参数： * replicas用于描述服务对应的实例数 * mode用来描述服务的类型，replicated多实例类型，global全局服务类型（在每一个节点上都会创建）

在第一个节点上，通过create命令创建服务：

[root@host1 ~]# docker service create --name test_service --replicas 2 busybox ping 186.100.40.1
et1i181ni4dj2ocm1l4frbqzm

（2）service ls命令用户查看service列表

[root@host1 ~]# docker service ls
ID            NAME          SCALE  IMAGE    COMMAND
et1i181ni4dj  test_service  2      busybox  ping 186.100.40.1

（3）service tasks可以查看service对应的任务

[root@host1 ~]# docker service tasks test_service
ID                        NAME            SERVICE      IMAGE    LASTSTATE              DESIREDSTATE  NODE
1yewid0xbviyggyvjv10adjrs  test_service.1  test_service  busybox  RunningAbout a minute  Running        host3
9l4z9aqlg9iwums14yxu0lym7  test_service.2  test_service  busybox  RunningAbout a minute  Running        host1

可以看到，刚才创建的service对应的2个实例分别在host3和host1上

（4）service update可以对服务的参数进行更新

对刚才创建的service实力数目进行更新

[root@host1 ~]# docker service update --replicas 3 test_service
test_service
[root@host1 ~]# docker service tasks test_service
ID                        NAME            SERVICE      IMAGE    LASTSTATE          DESIREDSTATE  NODE
1yewid0xbviyggyvjv10adjrs  test_service.1  test_service  busybox  Running 5 minutes  Running        host3
9l4z9aqlg9iwums14yxu0lym7  test_service.2  test_service  busybox  Running 5 minutes  Running        host1
emnuxg8sb37hbxbpdt63mpy1a  test_service.3  test_service  busybox  Running 18 seconds  Running        host2

（5）service inspect命令用户查看service详细信息

[root@host1 ~]# docker service inspect test_service
[
    {
        "ID": "et1i181ni4dj2ocm1l4frbqzm",
        "Version": {
            "Index": 33
        },
        "CreatedAt": "2016-06-19T13:07:29.549242777Z",
        "UpdatedAt": "2016-06-19T13:12:23.210552545Z",
        "Spec": {
            "Name": "test_service",
            "TaskTemplate": {
                "ContainerSpec": {
                    "Image": "busybox",
                    "Args": [
                        "ping",
                        "186.100.40.1"
                    ]
                },
                "Resources": {
                    "Limits": {},
                    "Reservations": {}
                },
                "RestartPolicy": {
                    "Condition": "any",
                    "MaxAttempts": 0
                },
                "Placement": {}
            },
            "Mode": {
                "Replicated": {
                    "Replicas": 3
                }
            },
            "UpdateConfig": {
                "Parallelism": 1
            },
            "EndpointSpec": {
                "Mode": "vip"
            }
        },
        "Endpoint": {
            "Spec": {}
        }
    }
]

3、node

node命令用户节点管理，比如将一个节点变为manager节点、同意节点加入请求等。

[root@host1 ~]# docker node --help
 
Usage:  dockernodeCOMMAND
 
ManageDockerSwarmnodes
 
Options:
      --help  Printusage
 
Commands:
  accept      Accept a nodein theswarm
  demote      Demote a nodefrommanagerin theswarm
  inspect    Inspect a nodein theswarm
  ls          Listnodesin theswarm
  promote    Promote a nodeto a managerin theswarm
  rm          Remove a nodefromtheswarm
  tasks      Listtasksrunningon a node
  update      Update a node
 
Run 'docker node COMMAND --help' for moreinformationon a command.

（1）node accept

accpet命令用户同意节点加入请求，在初始化swarm集群时候通过–auto-accept可以指定节点默认加入的较色，如果为none的话，需要通过accpet命令同意加入节点，成为worker节点 重新初始化三节点，集群初始化时，采用–auto-accept=none，最终可以看到新加入的两个节点MEMBERSHIP为Pending

[root@host1 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
15wyq8vc1vxe7r4n3yyl0p93j *  host1  Accepted    Ready    Active        Reachable      Yes
2fh91rbwtj2xosqixzlc7s3mg          Pending    Unknown  Active                        
5okhpif49b4d6byohpdr4vdmw          Pending    Unknown  Active

通过accept命令同意节点加入：

[root@host1 ~]# docker node accept 2fh91rbwtj2xosqixzlc7s3mg
2fh91rbwtj2xosqixzlc7s3mg
[root@host1 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
15wyq8vc1vxe7r4n3yyl0p93j *  host1  Accepted    Ready    Active        Reachable      Yes
2fh91rbwtj2xosqixzlc7s3mg    host3  Accepted    Ready    Active                        
5okhpif49b4d6byohpdr4vdmw          Pending    Unknown  Active  
[root@host1 ~]# docker node accept 5okhpif49b4d6byohpdr4vdmw
5okhpif49b4d6byohpdr4vdmw
5okhpif49b4d6byohpdr4vdmw attemptingto accept a nodein theswarm.
[root@host1 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
15wyq8vc1vxe7r4n3yyl0p93j *  host1  Accepted    Ready  Active        Reachable      Yes
2fh91rbwtj2xosqixzlc7s3mg    host3  Accepted    Ready  Active                        
5okhpif49b4d6byohpdr4vdmw    host2  Accepted    Ready  Active

（2）node promote命令用于将worker节点提升为manager节点，用户manager的HA

[root@host1 ~]# docker node promote 2fh91rbwtj2xosqixzlc7s3mg
2fh91rbwtj2xosqixzlc7s3mg
2fh91rbwtj2xosqixzlc7s3mg attemptingto promote a nodeto a managerin theswarm.
[root@host1 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
15wyq8vc1vxe7r4n3yyl0p93j *  host1  Accepted    Ready  Active        Reachable      Yes
2fh91rbwtj2xosqixzlc7s3mg    host3  Accepted    Ready  Active        Reachable      
5okhpif49b4d6byohpdr4vdmw    host2  Accepted    Ready  Active  
[root@host1 ~]# docker node promote 5okhpif49b4d6byohpdr4vdmw
5okhpif49b4d6byohpdr4vdmw
5okhpif49b4d6byohpdr4vdmw attemptingto promote a nodeto a managerin theswarm.
[root@host1 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
15wyq8vc1vxe7r4n3yyl0p93j *  host1  Accepted    Ready  Active        Reachable      Yes
2fh91rbwtj2xosqixzlc7s3mg    host3  Accepted    Ready  Active        Reachable      
5okhpif49b4d6byohpdr4vdmw    host2  Accepted    Ready  Active        Reachable

（3）node demote命令用户将manger节点变为worker节点

[root@host1 ~]# docker node demote 15wyq8vc1vxe7r4n3yyl0p93j
15wyq8vc1vxe7r4n3yyl0p93j
15wyq8vc1vxe7r4n3yyl0p93j attemptingto demote a managerin theswarm.
[root@host1 ~]# docker node ls
Errorresponsefromdaemon: rpcerror: code = 4 desc = contextdeadlineexceeded

可以看到变为worker节点以后，就无法再此节点上对集群进行管理

[root@host2 ~]# docker node ls
ID                          NAME  MEMBERSHIP  STATUS  AVAILABILITY  MANAGERSTATUS  LEADER
15wyq8vc1vxe7r4n3yyl0p93j    host1  Accepted    Ready  Active                        
2fh91rbwtj2xosqixzlc7s3mg    host3  Accepted    Ready  Active        Reachable      Yes
5okhpif49b4d6byohpdr4vdmw *  host2  Accepted    Ready  Active        Reachable

可以看到此时，host2已经变为了leader

（4）node inspect查看节点详细信息

（5）node ls查看节点列表

（6）node rm删除一个节点

（7）node tasks可以查看节点上运行的任务task

（8）node update可以更新一个节点

[root@host2 ~]# docker node update --help
 
Usage:  dockernodeupdate [OPTIONS] NODE
 
Update a node
 
Options:
      --availabilitystring  Availabilityofthenode (active/pause/drain)
      --help                  Printusage
      --membershipstring    Membershipofthenode (accepted/rejected)
      --rolestring          Roleofthenode (worker/manager)

4、stack/deploy

这两个命令都是experiment阶段，用来部署一个STACK，一个STACK可以描述多个service。 在创建Stack前，需要通过compose.yml文件描述一个组应用，然后再用compose bundle产生一个dsb文件，用户stack的创建。

[root@host2 ~]# docker stack --help
 
Usage:  dockerstackCOMMAND
 
ManageDockerstacks
 
Options:
      --help  Printusage
 
Commands:
  config      Printthestackconfiguration
  deploy      Createand update a stack
  rm          Removethestack
  tasks      Listthetasksin thestack
 
Run 'docker stack COMMAND --help' for moreinformationon a command.

5、总结

Docker1.12通过swarm实现了容器集群，在集群之上可以service命令创建一个服务，实现服务的多实例。在服务之上，通过stack可以描述一组服务，最终实现一个大系统的部署。可见，Docker在公司在下一盘很大的棋，从镜像生态到容器引擎，从容器引擎到容器集群，Anything is possible in Docker。