ConductR Sandbox

Build Status

This plugin has been replaced by sbt-conductr. Please use this instead.

Introduction

sbt-conductr-sandbox aims to support the running of a Docker-based ConductR cluster in the context of a build. The cluster can then be used to assist you in order to verify that endpoint declarations and other aspects of a bundle configuration are correct. The general idea is that this plugin will also support you when building your project on CI so that you may automatically verify that it runs on ConductR.

Note that because the sandbox uses Docker, you cannot run bundles that also use Docker - Docker cannot run within Docker! If you have a Docker bundle then we recommend that you setup a VM for development and follow the regular installation guide of our docs; perhaps just for a single node to make your setup a bit easier.

Prerequisites

Docker is required so that you can run the ConductR cluster as if it were running on a number of machines in your network. You won't need to understand much about Docker for ConductR other than installing it as described in its "Get Started" section. If you are on Windows or Mac then you will become familiar with docker-machine which is a utility that controls a virtual machine for the purposes of running Docker.

The conductr-cli is used to communicate with the ConductR cluster.

Usage

  1. Add sbt-conductr-sandbox to the project/plugins.sbt:

    addSbtPlugin("com.typesafe.conductr" % "sbt-conductr-sandbox" % "1.4.3")
    
  2. Verify that sbt-bundle keys have been set. For further information refer to the sbt-bundle documentation:

    import ByteConversions._
    BundleKeys.nrOfCpus := 1.0
    BundleKeys.memory := 64.MiB
    BundleKeys.diskSpace := 10.MB
    
  3. Specify the ConductR Developer Sandbox version in the build.sbt. This version is available gratis during development with registration at lightbend.com. Please visit the ConductR Developer page for usage instructions.

    SandboxKeys.imageVersion in Global := "YOUR_CONDUCTR_SANDBOX_VERSION"
    
  4. Reload the sbt session:

    reload
    

The plugin is then enabled automatically for your entire project. Additionally it includes sbt-conductr and sbt-bundle so these plugins doesn’t need to be added separately. The conduct commands such as conduct info will automatically communicate with the Docker cluster managed by the sandbox. There is no need to set ConductR’s ip address with controlServer manually.

To start the sandbox use the run command:

sandbox run
[info] Running ConductR...
[info] Running container cond-0 exposing 192.168.59.103:9000...

Checkout the commands section to see all available sandbox commands.

Debugging application in ConductR sandbox

It is possible to debug your application inside of the ConductR sandbox:

  1. Start ConductR sandbox in debug mode:

    sandbox debug
  2. This exposes a debug port to Docker and the ConductR container. By default the debug port is set to 5005 which is the default port for remote debugging in IDEs such as IntelliJ. If you are happy with the default setting skip this step. Otherwise specify another port in the build.sbt:

    SandboxKeys.debugPort := 5432
  3. Load and run your bundle to the sandbox:

    conduct load <HIT THE TAB KEY AND THEN RETURN>
    conduct run my-app

You should be now able to remotely debug the bundle inside the ConductRs sandbox with your favorite IDE.

Features

In order to use the following features you should ensure that the machine that runs Docker has enough memory, typically at least 2GB. VM configurations such as those provided via docker-machine and Oracle's VirtualBox can be configured like so:

docker-machine stop default
VBoxManage modifyvm default --memory 2048
docker-machine start default

Features are specified with a --with-features visualization logging option convention. The following features are available:

Name ConductR ports Docker ports Description
logging 9200, 5601 9200, 5601 Consolidates the logging output of ConductR itself and the bundles that it executes. To view the consolidated log messsages run `conduct logs {bundle-name} or access the Kibana UI at http://{docker-host-ip}:5601.
monitoring N/A N/A Enables the Takipi based monitoring via Typesafe monitoring.
visualization 9999 9909 Provides a web interface to visualize the ConductR cluster together with deployed and running bundles. After enabling the feature, access it at http://{docker-host-ip}:9909.

Docker Container Naming

Each node of the Docker cluster managed by the sandbox is given a name of the form cond-n where n is the node number starting at zero. Thus cond-0 is the first node (and the only node given default settings).

Port Mapping Convention

The following ports are exposed to the ConductR and Docker containers:

  • BundleKeys.endpoints of sbt-bundle
  • SandboxKeys.ports
  • SandboxKeys.debugPort if sandbox is started in debug mode

Example

Your application defines these settings in the build.sbt:

lazy val root = (project in file(".")).enablePlugins(JavaAppPackaging)

BundleKeys.endpoints := Map("sample-app" -> Endpoint("http", services = Set(uri("http://:9000"))))
SandboxKeys.ports in Global := Set(1111)
SandboxKeys.debugPort := 5095

We specify that the web application should serve traffic on port 9000. Additionally we expose port 1111. The debug port gets only mapped if we start the ConductR sandbox cluster in debug mode with sandbox debug.

Now we create a ConductR cluster with 3 nodes:

sandbox run --nr-of-containers 3

These settings result in the following port mapping:

Docker container ConductR port Docker internal port Docker public port
cond-0 9000 9000 9000
cond-1 9000 9000 9010
cond-2 9000 9000 9020
cond-0 1111 1111 1111
cond-1 1111 1111 1121
cond-2 1111 1111 1131
cond-0 5095 5095 5095
cond-1 5095 5095 5005
cond-2 5095 5095 5015

Each specified port is mapped to a unique public Docker port in order to allow multiple nodes within the sandbox cluster. By convention, the first node is using the ConductR port as the public Docker port, e.g. 1111 becomes 1111. The following nodes using a different public Docker port by increasing the second last digit of the ConductR port by the node number, e.g. the second node is mapping the port 1111 to 1121, the third is mapping the port 1111 to 1131. The web application becomes available on the Docker host IP address. The sandbox cluster is configured with a proxy and will automatically route requests to the correct instances that you have running in the cluster. Therefore any one of the addresses with the 9000, 9010 or 9020 ports will reach your application.

As a convenience, sandbox run and sandbox debug is reporting each of the above mappings along with the IP addresses:

[info] Starting ConductR nodes..
[info] Starting container cond-0 exposing 192.168.99.100:9000, 192.168.99.100:1111, 192.168.99.100:5095...
[info] f7f030f8b97cc375642f5a0321654594547b1e7d3a327c6f120715ba3199dda0
[info] Starting container cond-1 exposing 192.168.99.100:9010, 192.168.99.100:1121, 192.168.99.100:5005...
[info] 7f2caf1053810465589ac9710e0676a39d6828687c06020fa2d4a80ba815d80a
[info] Starting container cond-2 exposing 192.168.99.100:9020, 192.168.99.100:1131, 192.168.99.100:5015...
[info] 5644465b394617fb8d18bdb64ce488257776eea4437d7d1d205b481b775ab42a

Roles

ConductR uses roles to load and scale bundles to the respective nodes. For more information about it please refer to Managing services documentation.

To keep things simple sbt-conductr-sandbox ignores roles by default. Therefore, your bundles can be loaded and scaled to all ConductR nodes. To have more fine-grained control of the bundles it is also possible to provide a custom role configuration by specifying the SandboxKeys.conductrRoles setting:

SandboxKeys.conductrRoles in Global := Seq(Set("muchMem", "myDB"))

This is only adding the roles muchMem and myDB to all ConductR nodes. The bundle roles are ignored in case SandboxKeys.conductrRoles is specified. Assigning different roles to nodes is also possible by specifying a set for each of the nodes:

SandboxKeys.conductrRoles in Global := Seq(Set("muchMem"), Set("myDB"))

In the above example the bundles with the role muchMem would only be loaded and scaled to the first node and bundles with the role myDB only to the second node.

If the --nr-of-containers startup option is larger than the size of the conductrRoles sequence then the specified roles are subsequently applied to the remaining nodes.

SandboxKeys.conductrRoles in Global := Seq(Set("muchMem", "frontend"), Set("myDB"))
sandbox run --nr-of-containers 4

This would load and scale all bundles with the roles muchMem and frontend to the first and third node and bundles with the role myDB to the second and fourth node.

Functions

There is a ConductRSandbox.runConductRs function for starting the sandbox environment programmatically. This is useful for setting up test environments. Conversely there is a ConductRSandbox.stopConductRs function call can be made to close down the sandbox environment programmatically. The following listing shows how these functions can be used within an sbt integration test:

testOptions in IntegrationTest ++= Seq(
  Tests.Setup { () =>
    ConductRSandbox.stopConductRs(streams.value.log)
    ConductRSandbox.runConductRs(
      1,
      Seq.empty,
      Set.empty,
      Map.empty,
      (conductrImage in Global).value,
      (conductrImageVersion in Global).value,
      Set.empty,
      streams.value.log,
      "info")
  },

  Tests.Cleanup (() => ConductRSandbox.stopConductRs(streams.value.log))
)

Settings

The following settings are provided under the SandboxKeys object:

Name Scope Description
envs Global A Map[String, String] of environment variables to be set for each ConductR container.
image Global The Docker image to use. By default typesafe-docker-registry-for-subscribers-only.bintray.io/conductr/conductr-dev is used i.e. the single node version of ConductR.
imageVersion Global The version of the Docker image to use. Must be set. Please visit the ConductR Developer page on Typesafe.com for the current version and additional information.
ports Global A Seq[Int] of ports to be made public by each of the ConductR containers. This will be complemented to the endpoints setting's service ports declared for sbt-bundle.
debugPort Project Debug port to be made public to the ConductR containers if the sandbox gets started in debug mode. The debug ports of each sbt project setting will be used. If sbt-bundle is enabled the JVM argument -jvm-debug $debugPort is additionally added to the startCommand of sbt-bundle. Default is 5005.
logLevel Global The log level of ConductR which can be one of "debug", "warning" or "info". By default this is set to "info". You can observe ConductR's logging via the docker logs command. For example docker logs -f cond-0 will follow the logs of the first ConductR container.
conductrRoles Global Sets additional roles to the ConductR nodes. Defaults to Seq.empty. If this settings is not set the bundle roles specified with BundleKeys.roles are used. To provide a custom role configuration specify a set of roles for each node. Example: Seq(Set("megaIOPS"), Set("muchMem", "myDB")) would add the role megaIOPS to first ConductR node. muchMem and myDB would be added to the second node. If nrOfContainers is larger than the size of the conductrRoles sequence then the specified roles are subsequently applied to the remaining nodes.

Commands

The following commands are provided:

Name Description
sandbox run Starts the ConductR sandbox cluster without remote debugging facilities.
sandbox debug Starts the ConductR sandbox cluster and exposes the debugPort to docker and ConductR to enable remote debugging.
sandbox stop Stops the ConductR sandbox cluster.

© Typesafe Inc., 2015