github.com/jmitchell/nomad@v0.1.3-0.20151007230021-7ab84c2862d8/website/source/docs/drivers/docker.html.md (about)

     1  ---
     2  layout: "docs"
     3  page_title: "Drivers: Docker"
     4  sidebar_current: "docs-drivers-docker"
     5  description: |-
     6    The Docker task driver is used to run Docker based tasks.
     7  ---
     8  
     9  # Docker Driver
    10  
    11  Name: `docker`
    12  
    13  The `docker` driver provides a first-class Docker workflow on Nomad. The Docker
    14  driver handles downloading containers, mapping ports, and starting, watching,
    15  and cleaning up after containers.
    16  
    17  ## Task Configuration
    18  
    19  The `docker` driver supports the following configuration in the job specification:
    20  
    21  * `image` - (Required) The Docker image to run. The image may include a tag or
    22    custom URL. By default it will be fetched from Docker Hub.
    23  
    24  * `command` - (Optional) The command to run when starting the container.
    25  
    26  * `network_mode` - (Optional) The network mode to be used for the container.
    27     Valid options are `default`, `bridge`, `host` or `none`. If nothing is
    28     specified, the container will start in `bridge` mode. The `container`
    29     network mode is not supported right now.
    30  
    31  ### Port Mapping
    32  
    33  Nomad uses port binding to expose services running in containers using the port
    34  space on the host's interface. For example, Nomad host running on `1.2.3.4` may
    35  allocate port `22333` to a task, so you would access that service via
    36  `1.2.3.4:22333`.
    37  
    38  Nomad provides automatic and manual mapping schemes for Docker. You can use
    39  either or both schemes for a task. Nomad binds both tcp and udp protocols to
    40  ports used for Docker containers. This is not configurable.
    41  
    42  Note: You are not required to map any ports, for example if your task is running
    43  a crawler or aggregator and does not provide a network service. Tasks without a
    44  port mapping will still be able to make outbound network connections.
    45  
    46  #### Automatic Port Mapping
    47  
    48  Typically when you create a Docker container you configure the service to start
    49  listening on a port (or ports) when you start the container. For example, redis
    50  starts listening on `6379` when you `Docker run redis`. Nomad supports this by
    51  mapping the random port to the port inside the container.
    52  
    53  You need to tell Nomad which ports your container is using so Nomad can map
    54  allocated ports for you. You do so by specifying a **numeric port value** for
    55  the `dynamic_ports` option in your job specification.
    56  
    57  ```
    58  dynamic_ports = ["6379"]
    59  # or
    60  dynamic_ports = [6379]
    61  ```
    62  
    63  This instructs Nomad to create a port mapping from the random port on the host
    64  to the port inside the container. So in our example above, when you contact the
    65  host on `1.2.3.4:22333` you will actually hit the service running inside the
    66  container on port `6379`. You can see which port was actually bound by reading the
    67  `NOMAD_PORT_6379` [environment variable](/docs/jobspec/environment.html).
    68  
    69  In most cases, the automatic port mapping will be the easiest to use, but you
    70  can also use manual port mapping (described below).
    71  
    72  #### Manual Port Mapping
    73  
    74  The `dynamic_ports` option takes any alphanumeric string as a label, so you could
    75  also specify a label for the port like `http` or `admin` to designate how the
    76  port will be used.
    77  
    78  In this case, Nomad doesn't know which container port to map to, so it maps 1:1
    79  with the host port. For example, `1.2.3.4:22333` will map to `22333` inside the
    80  container.
    81  
    82  ```
    83  dynamic_ports = ["http"]
    84  ```
    85  
    86  Your process will need to read the `NOMAD_PORT_HTTP` environment variable to
    87  determine which port to bind to.
    88  
    89  ## Client Requirements
    90  
    91  Nomad requires Docker to be installed and running on the host alongside the Nomad
    92  agent. Nomad was developed against Docker `1.8.2`.
    93  
    94  By default Nomad communicates with the Docker daemon using the daemon's
    95  unix socket. Nomad will need to be able to read/write to this socket. If you do
    96  not run Nomad as root, make sure you add the Nomad user to the Docker group so
    97  Nomad can communicate with the Docker daemon.
    98  
    99  For example, on ubuntu you can use the `usermod` command to add the `vagrant` user to the
   100  `docker` group so you can run Nomad without root:
   101  
   102      sudo usermod -G docker -a vagrant
   103  
   104  For the best performance and security features you should use recent versions of
   105  the Linux Kernel and Docker daemon.
   106  
   107  ## Client Configuration
   108  
   109  The `docker` driver has the following configuration options:
   110  
   111  * `docker.endpoint` - Defaults to `unix:///var/run/docker.sock`. You will need
   112    to customize this if you use a non-standard socket (http or another location).
   113  
   114  ## Client Attributes
   115  
   116  The `docker` driver will set the following client attributes:
   117  
   118  * `driver.Docker` - This will be set to "1", indicating the
   119    driver is available.
   120  
   121  ## Resource Isolation
   122  
   123  ### CPU
   124  
   125  Nomad limits containers' CPU based on CPU shares. CPU shares allow containers to
   126  burst past their CPU limits. CPU limits will only be imposed when there is
   127  contention for resources. When the host is under load your process may be
   128  throttled to stabilize QOS depending how how many shares it has. You can see how
   129  many CPU shares are available to your process by reading `NOMAD_CPU_LIMIT`. 1000
   130  shares are approximately equal to 1Ghz.
   131  
   132  Please keep the implications of CPU shares in mind when you load test workloads
   133  on Nomad.
   134  
   135  ### Memory
   136  
   137  Nomad limits containers' memory usage based on total virtual memory. This means
   138  that containers scheduled by Nomad cannot use swap. This is to ensure that a
   139  swappy process does not degrade performance for other workloads on the same host.
   140  
   141  Since memory is not an elastic resource, you will need to make sure your
   142  container does not exceed the amount of memory allocated to it, or it will be
   143  terminated or crash when it tries to malloc. A process can inspect its memory
   144  limit by reading `NOMAD_MEMORY_LIMIT`, but will need to track its own memory
   145  usage. Memory limit is expressed in megabytes so 1024 = 1Gb.
   146  
   147  ### IO
   148  
   149  Nomad's Docker integration does not currently provide QOS around network or
   150  filesystem IO. These will be added in a later release.
   151  
   152  ### Security
   153  
   154  Docker provides resource isolation by way of
   155  [cgroups and namespaces](https://docs.docker.com/introduction/understanding-docker/#the-underlying-technology).
   156  Containers essentially have a virtual file system all to themselves. If you need
   157  a higher degree of isolation between processes for security or other reasons, it
   158  is recommended to use full virtualization like [QEMU](/docs/drivers/qemu.html).