Home

Awesome

atlantis-aquarium

Atlantis in a Vagrant VM! Excellent for testing.

Requirements

You can install aquarium in the following approaches:

#Vagrant We have a vagrant box available. To run it,

What is inside the vagrant box

Once the atlantis-aquarium vagrant box is up and running, it runs a zookeeper, a docker registry, an atlantis-builder, an atlantis-manager, two routers (one internal and one external) and two atlantis-supervisors. Each component lives within a docker container.

##FAQ

launch-aquarium.sh gets killed while downloading vagrant box, and re-running the script gives an error

=> default: Box download is resuming from prior download progress
An error occurred while downloading the remote file. The error
message, if any, is reproduced below. Please fix this error and try
again.

HTTP server doesn't seem to support byte ranges. Cannot resume.

The Vagrant box host we are using (hashicorp) does not support range request, so the previous partial download cannot be resumed. The solution is to remove the partial downloaded vagrant box image and try again.

 rm ~/.vagrant.d/tmp/*

How do I ssh into a container in aquarium VM?

Use bin/atlantis-aquarium ssh command; refer to [interacting with the components](##Interacting with the components)

What happens if my VM get reloaded?

After a VM restart, you need to run launch-aquarium.sh again in order to bring up the containers. Please remember apps hosted in atlantis supervisor will not be recovered; you have to deploy them again.

#Control

Control is primarily through the controller, bin/atlantis-aquarium (which should probably become a gem or other easily-installable package at some point).

$bin/atlantis-aquarium

Usage:
  atlantis-aquarium build   [<component>...] [options] [-C | --compile] [-I | --image]
  atlantis-aquarium start   [<component>...] [options]
  atlantis-aquarium restart  <component>...  [options]
  atlantis-aquarium stop     <component>...  [options]
  atlantis-aquarium ssh     [<component>]    [options]
  atlantis-aquarium atlantis [--] [<argument>...]
  atlantis-aquarium provision
  atlantis-aquarium register-components
  atlantis-aquarium base-cluster
  atlantis-aquarium build-layers  [--base] [--builder]
  atlantis-aquarium zookeeper

Options:
  -C, --compile  Only compile; don't build image or deploy
  -I, --image    Only compile and build Docker image; don't deploy
  -i, --instance Which instance of the component to act on [default: all]
  -h, --help     Show usage

Component names

atlantis-aquarium supports the following component names:

Interacting with the components

Once the components are built, you can start/stop/restart or obtain an ssh shell into the container:

$bin/atlantis-aquarium start builder

$bin/atlantis-aquarium restart manager

$bin/atlantis-aquarium ssh manager

Access zookeeper running in aquarium

atlantis-aquarium provides a short command to access zookeeper

$bin/atlantis-aquarium zookeeper

Convenience wrapper for atlantis-manager CLI

atlantis subcommand will pass remaining arguments to the atlantis-manager cli run within the VM; this is a convenience, e.g.

atlantis-aquarium>bin/atlantis-aquarium atlantis list-supervisors

{"build"=>false, "<component>"=>[], "--compile"=>0, "--no-cache"=>false, "--image"=>0, "--instance"=>"all", "--help"=>false, "-c"=>false, "start"=>false, "restart"=>false, "stop"=>false, "ssh"=>false, "command"=>0, "atlantis"=>true, "<arguments>"=>[], "provision"=>false, "register-components"=>false, "base-cluster"=>false, "build-layers"=>false, "--base"=>false, "--builder"=>false}
Restarting in VM...
{"build"=>false, "<component>"=>[], "--compile"=>0, "--no-cache"=>false, "--image"=>0, "--instance"=>"all", "--help"=>false, "-c"=>false, "start"=>false, "restart"=>false, "stop"=>false, "ssh"=>false, "command"=>0, "atlantis"=>true, "<arguments>"=>[], "provision"=>false, "register-components"=>false, "base-cluster"=>false, "build-layers"=>false, "--base"=>false, "--builder"=>false}
cd .; atlantis -R dev list-supervisors
LDAP Username:
2015/07/25 20:52:44 List Supervisors...
2015/07/25 20:52:44 -> status: OK
2015/07/25 20:52:44 -> supervisors:
2015/07/25 20:52:44 ->   172.17.0.8
2015/07/25 20:52:44 ->   172.17.0.9
Connection to 127.0.0.1 closed.

Manually

The following steps are needed in order to bootstrap atlantis aquarium. The steps are recorded in makeitso.sh script.

clone repos

First, create $HOME/repos folder and clone atlantis-aquarium repo under this folder. Then find $HOME/repos/atlantis-aquarium/bin/gather-files script and run it; the script will clone several atlantis coomponents repos that aquarium needs into the $HOME/repos folder;

$cd $HOME/repos/atlantis-aquarium
$./bin/clone-repos

....

$ls $HOME/repos
atlantis-aquarium atlantis-manager  atlantis-supervisor hello-atlantis
atlantis-builder  atlantis-router   go-docker-registry

provision

The provision subcommand installs go, sets up docker, etc in the vagrant VM. This step should be done with a fresh VM.

$bin/atlantis-aquarium provision

.............

$vagrant status
Current machine states:

aquarium                  running (virtualbox)

The VM is running. To stop this VM, you can run `vagrant halt` to
shut it down forcefully, or you can run `vagrant suspend` to simply
suspend the virtual machine. In either case, to restart it again,
simply run `vagrant up`.

build and start atlantis components

Aquarium requires the following components. All but base-aquarium-image are services that run in docker containers within the VM we just created; base-aquarium-image is a special target to build the base image that other components build off of.

The build subcommand compiles the component, builds the container, and (re)start the given component. build take following options

 -C, compiles only; If an instance is given, only that instance will be compiled
 -I, compiles and builds container but doesn't restart.  If an instance is given, only that instance will be build.

to build and (re)start all the components;

$bin/atlantis-aquarium build

or, to build and (re)start a single component

$bin/atlantis-aquarium build <component-name>

Known issue with build

While running build subcommand, sometimes it fails with error message like this:

INFO[0024] Error getting container 2328b2d6de9727c41ad1dcf1f977057a1cedb15e65c466a01482c09576236618 from driver devicemapper: open /dev/mapper/docker-8:1-403558-2328b2d6de9727c41ad1dcf1f977057a1cedb15e65c466a01482c09576236618: no such file or directory

This is caused by well known docker issue (https://github.com/docker/docker/issues/4036) where docker has a race condition in the storage backend. Re-running the build will pass normally.

build-layers

The build-layers subcommand builds layers required for deploying. This is only needed for the simple builder, as used by aquarium. It support following options:

$bin/atlantis-aquarium build-layers [--base] [--builder]

register-components

The register-components subcommand registers the supervisors, routers, etc. with the manager. Should be done once after all components are started, or after zookeeper is restarted.

$bin/atlantis-aquarium register-components

Note: You may be prompted to input LDAP username and/or password, just type enter to continue

hello-world app

The base-cluster subcommand sets up a sample hello-go app and deploys it to a supervisor. It is useful as a test to ensure everything is working in aquarium. It should be done after all other steps have been taken.

$bin/atlantis-aquarium base-culster

If everything works out nicely, you just launched your first atlantis app in aquarium. Now check it out:

$bin/atlantis-aquarium ssh supervisor

... ...

root@c5fb42cc6eb4:~# docker ps
CONTAINER ID        IMAGE                                                                                COMMAND                CREATED             STATUS              PORTS                                                                                                                                                                                  NAMES
d62dc9dc237e        registry.aquarium/apps/hello-go1.2-6292411629e73234fd0ecd0ded43b1fd259dfa44:latest   "runsvdir /etc/servi   6 minutes ago       Up 6 minutes        0.0.0.0:61000->61000/tcp, 0.0.0.0:61100->61100/tcp, 0.0.0.0:61200->61200/tcp, 0.0.0.0:61300->61300/tcp, 0.0.0.0:61400->61400/tcp, 0.0.0.0:61500->61500/tcp, 0.0.0.0:61600->61600/tcp   hello-go1.2-629241-test-oAZOK9


root@c5fb42cc6eb4:~# curl http://localhost:61000
<!DOCTYPE html><html><head><title>hello-go</title></head><body><pre>
Hello from Go /
</pre></body></html>

Clean up

nuke.sh, as the name suggested, tears down everything!

$./nuke.sh

Miscellaneous Errors

1. If you see an error downloading the virtual box image at
https://atlas.hashicorp.com/ghao/boxes/atlantis-aquarium/versions/0.0.1/providers/virtualbox.box its most likely
because of a URL redirect. Simply curl the url, grab the redirected URL and in the  Vagrantfile uncomment aquarium.vm.box,
aquarium.vm.box_url and comment out aquarium.vm.box. Also set aquarium.vm.box_url to the redirected url.


2. You see following error while building the app
$bin/atlantis-aquarium base-cluster
..
sh: 1: cd: can't cd to /home/vagrant/repos/hello-atlantis/
..
2015/07/31 21:54:58 Triggering Simple Build
Build Error: map[]
Connection to 127.0.0.1 closed

You are missing the hello-go repo. Run the following command:

$bin/gather-files

If the build still fails, checks repos and you will probably notice the hello-atlantis directory is not synced

$ls -al $HOME/repos
ls: cannot access hello-atlantis: No such file or directory
total 4
drwxr-xr-x 1 1001 1001  306 Jul 31 21:50 .
drwx------ 6 root root 4096 Jul 31 21:54 ..
drwxr-xr-x 1 1001 1001  612 Jul 31 19:19 atlantis-aquarium
drwxr-xr-x 1 1001 1001  544 Jul 31 21:49 atlantis-builder
drwxr-xr-x 1 1001 1001  442 Jul 31 21:48 atlantis-manager
drwxr-xr-x 1 1001 1001  442 Jul 31 21:49 atlantis-router
drwxr-xr-x 1 1001 1001  476 Jul 31 21:49 atlantis-supervisor
drwxr-xr-x 1 1001 1001  442 Jul 31 21:48 go-docker-registry
?????????? ? ?    ?       ?            ? hello-atlantis

Now reload the vagrant and launch aquarium
$vagrant reload
$./launch-aquarium.sh
$bin/atlantis-aquarium base-cluster

3. Want to cleanup state in zookeeper.
$vagrant ssh
$/usr/share/zookeeper/bin/zkCli.sh -server `cat data/zookeeper/ip`
$delete /atlantis/supervisors/dev/172.17.0.8
  1. You have moved machines and launch throws UID Error The VirtualBox VM was created with a user that doesn't match the current user running Vagrant....
$cd $HOME/repos/atlantis-aquarium/.vagrant/machines/aquarium/virtualbox
edit the file creator_uid and replace the uid there with the uid of the current user
verify by running vagrant up