Introduction

This tutorial will provide an overview of working with docker-machine, a nifty tool for managing and deploying docker hosts.

Dependencies

Eris uses a handful of docker-machine features, both as an integral part of the tool and for continous integration/testing.

Many features of the eris platform are available without docker-machine (even on linux which does not require it as part of the docker installation process). However, your blockchain superpowers are greatly limited without it. See linux install guidelines here. Windows and OSX users are highly recommended to install the Docker Toolbox as a means of installing eris; and as part of that docker-machine will be installed.

This tutorial requires eris be installed (see links above) for all Sections except for the first.

Overview

The first section of this tutorial will highlight basic docker-machine commands and functioning. If it feels oddly familiar, that’s because I’ve modeled after this intro but with a marmot twist.

Section two will elaborate on eris <=> docker-machine orchestration.

Section three focuses on cloud deployments.

N.B.

The first section is helpful in navigating the nuances of installing the eris tool on Windows/OSX operating systems but also introduces important docker-machine concepts relevant across os’.

Future releases of eris will further simplify the process (and abstract away the use of docker-machine directly). On this last point, eris is primarily a convenience wrapper around various docker commands leveraging the go-dockerclient library. If you are new to docker, hang on!

End N.B.

Protip The marmots use docker-machine a lot. As such we normally alias dm=docker-machine to our ~/.bashrc, ~/.zshrc, etc.

Setup & Basic Commands

This section is tailored for OSX/Windows. If you are on Linux you can use if you like, but mostly it should be thought of as informational.

After installing the toolbox, clicking the Docker Quickstart Terminal will open a new terminal window.

You’ll likely see:

(a) a success message, or
(b) something like:

Host does not exist: "default"
docker is configured to use the default machine with IP
For help getting started, check out the docs at https://docs.docker.com
Host does not exist: "default"

or;

(c) that certificates require regeneration.

These error messages will depend on where you are in the toolbox/docker/docker-machine/eris experimentation process.

If (c), follow the instructions. If you don’t have a default machine, go ahead and create it:

docker-machine create default --driver virtualbox

Your output should be similar to the one a few lines below. Since eris <=> docker-machine can be testy on osx/windows, create a machine of the same name to please the marmots:

docker-machine create eris --driver virtualbox

You’ll see something like:

Running pre-create checks...
Creating machine...
(eris) Creating VirtualBox VM...
(eris) Creating SSH key...
(eris) Starting VM...
Waiting for machine to be running, this may take a few minutes...
Machine is running, waiting for SSH to be available...
Detecting operating system of created instance...
Detecting the provisioner...
Provisioning with boot2docker...
Copying certs to the local machine directory...
Copying certs to the remote machine...
Setting Docker configuration on the remote daemon...
Checking connection to Docker...
Docker is up and running!
To see how to connect Docker to this machine, run: docker-machine env eris

That last line is critical to understanding what docker-machine does and how eris leverages it under the hood. It will be a running theme throughout this tutorial.

What does it do, you ask? We’ll get there. First, let’s take a look at our machines:

docker-machine ls

There are two things to note here: the - under ACTIVE indicates that neither machine is active (“in scope”). Specifically, docker has not been told to use either machines’ environment variables. If you try, for example, docker run hello-world, you’ll probably get an error like: Cannot connect to the Docker daemon. Is the docker daemon running on this host?. Yet there are two machine running! (Note: depending on how you started the toolbox, default may have a *, in which case the docker run command would have worked. More on this further below.

NAME      ACTIVE   DRIVER       STATE     URL                         SWARM   ERRORS
default   -        virtualbox   Running   tcp://192.168.99.100:2376
eris      -        virtualbox   Running   tcp://192.168.99.101:2376

Well, the machines are running, but the docker daemon doesn’t know about them. Let’s fix that:

docker-machine env eris

The output you see below is a list of environment variables that need to be set if you’d like the docker daemon to talk to the machine named eris. In fact, all this command did was stdout to the terminal.

export DOCKER_TLS_VERIFY="1"
export DOCKER_HOST="tcp://192.168.99.101:2376"
export DOCKER_CERT_PATH="/Users/zicter/.docker/machine/machines/eris"
export DOCKER_MACHINE_NAME="eris"
# Run this command to configure your shell:
# eval "$(docker-machine env eris)"

To put the eris machine “in scope”, run:

eval "$(docker-machine env eris)"

This command evaluates and sets the environment variables output from the env command. There ought not be any output to your screen. Note: this command has shell/operating system nuances. See here for solutions.

Now, re-run docker-machine ls and “eris” will be in scope:

NAME      ACTIVE   DRIVER       STATE     URL                         SWARM   ERRORS
default   -        virtualbox   Running   tcp://192.168.99.100:2376
eris      *        virtualbox   Running   tcp://192.168.99.101:2376

To confirm: docker-machine active, which should output “eris”.

If you’re an OSX/Windows user wanting to quickly get started on the eris platform, here is where you would:

  • install eris via go (go get github.com/eris-ltd/eris-cli/cmd/eris) or via the latest binary
  • run eris init
  • note the ip of your machine with docker-machine ip eris, which should match the ip seen in docker-machine ls.

This ip replaces the use of localhost (re: linux) in some of our tutorials and maps to 0.0.0.0 of a container running with exposed ports. Similar logic applies for cloud deployments with docker-machine, discussed further below.

There are several more docker-machine commands that we will save you from since they aren’t immediately relevant to our purposes though they’re definitely worth checking out.

N.B.

There is no reason why you need to have your eris box be a “local” machine using virtualbox. While we are opinionated (more so on OSX and Windows) about the machineName, if you work from a small-ish laptop then there are benefits to having your eris machine be a remote machine (on any of a range of cloud hosting providers). More on how to do this below.

End N.B.

Machine Thinking

Above, we created two machines.

These should be thought of as two individual hosts, in addition to the ‘master’ host. Indeed, the containers within them are accessed via different ip addresses.

While a machine is active (in scope - remember eval/env), any docker command executed on the host will apply to that machine and that machine only. This means that if you docker run a handful of containers while one machine is in scope then run the eval command for another machine, docker ps -a won’t show the containers you had just started because that machine is no longer in scope. Which is exactly what we want and why docker-machine is so awesome.

It gets even better; all machines created (and subsequently managed) from a host have access to the hosts filesystem (but not docker images, unfortunately). This logic applies equally to all eris commands, which have a pre-run check to connect to docker via the machine “in scope”.

N.B. for linux

eris (on linux) (by default) does not require an active machine (but does require a running docker daemon, i.e., docker installed) whereas OSX/Windows run by default with a machine named eris in scope, as per the previous section. On linux we routinely run remote machines and connect into them from linux.

Bringing a machine into scope on linux will override the default way eris connects to a docker daemon just as it overrides connecting into the eris machine on OSX/Windows.

End N.B.

Because eris respects the machine that is in scope (save for a few bugs on !linux), it’s possible to do something like:

docker-machine create marmot1 --driver virtualbox
docker-machine create marmot2 --driver virtualbox
docker-machine create marmot3 --driver virtualbox
eval "$(docker-machine env marmot1)"
eris init
eris services start ipfs
eval "$(docker-machine env marmot2)"
eris init
eris services start ipfs
eval "$(docker-machine env marmot3)"
eris init
eris services start ipfs

resulting in three docker machines each running an ipfs node. Note: executing that series of commands can take awhile.

If your service has any dependencies, they’ll also be running on each machine. Any files that need to be copied in at runtime can be sourced from the host. So that’s pretty cool.

But what if you have ten machines? A hundred? This would mean running the eval ... command each and every time you wanted to put a machine in scope prior to executing any eris command. Thankfully, there’s a better way. To list the running ipfs node on each machine from above, rather than using the eval/env combo:

eris services ls --running --machine marmot1
eris services ls --running --machine marmot2
eris services ls --running --machine marmot3

where the --machine flag tells eris to execute the listing function on that machine, so long as it exists and is running.

This flag is global (applies to all eris commands) and overides any machine/environment variables in scope.

This means that when using eris, you never have to check which one of your docker-machines is active and put it into scope before executing a command (that will simply provide eris with its default that can always be overridden with the cli flag).

We think this is pretty functionality for wrangling the blockchain dragons. Up next, cloud deployments with docker-machine.

Robots In The Sky

First, Create the Machines

Note: this section is a modified (& more generalized) version of the chain deploying tutorial.

Above, we deployed machines locally (using the virtualbox driver). With docker-machine, this is all done from the host. Instead of having to log into Digital Ocean/AWS, provision an instance, ssh into it, harden the server, install docker, install eris, copy in some files, and start a service/chain; it’s simply a matter of:

docker-machine create mach1 --driver digitalocean --digitalocean-access-token $DO_TOKEN --digitalocean-region sgp1
docker-machine create mach2 --driver azure --azure-subscription-id $AZURE_SUBSCRIPTION_ID --azure-location "West US" --azure-subscription-cert $AZURE_SUBSCRIPTION_CERT
docker-machine create mach3 --driver amazonec2 --amazonec2-access-key $AWS_A_KEY --amazonec2-secret-key $AWS_S_KEY --amazonec2-vpc-id $AWS_VPC

whereby specifying a region/location is usually optional (one will be chosen for you).

For more info on (the many) supported docker-machine drivers, see here. Vendor specific tutorials coming soon!

For maximum resilence, create machines in plenty of regions and on different providers.

N.B.

If you’re company has a subscription with a cloud provider, do not fret, docker-machine (probably) has you covered.

End N.B.

Second, Get eris Sorted

Once you’ve got those machines created:

eris init --yes --machine mach1
eris init --yes --machine mach2
eris init --yes --machine mach3

which will initialize eris on each machine (--yes overides the command-line prompts) by pulling eris’ default docker images and definition files into each machine.

Note that we didn’t need to install eris (or go, or docker) on any of those machines. Because eris is an opiniated orchestration tool that wraps docker, it goes where docker goes.

And docker-machine provisions docker hosts for you.

We’ve got three remote hosts, provisioned with eris images and ready for action. What a wonderful world we’re in.

Third, Get Going!

Now eris can “plug in” to each one of these machines with only a flag. Some things you could do with, say, 30 machines each named btcd/core/class0-9, respectively:

eris services start btcd --machine btcd0
eris services start btcd --machine btcd1
...
eris services start bitcoincore --machine core0
eris services start bitcoincore --machine core1
...
eris services start bitcoinclassic --machine class0
eris services start bitcoinclassic --machine class1
...

and you’d have ten of each full nodes to run simulations against or something else.

Perhaps performance of different tor relays is more interesting. In this case, to overide the default relay to run (exit node) you’ll have to make a small modification to the tor service definition file:

eris services edit tor

will open the file in your default editor if your shell has an $EDITOR set; the last line here can be edited next:

# replace 'exit' with 'bridge' or 'middle'
# depending on the type of relay you'd like to run
command = "-f /etc/tor/torrc.exit"

but first:

eris services start tor --machine exit0
eris services start tor --machine exit1
...

then edit the file, replacing ‘exit’ with ‘bridge’:

eris services start tor --machine bridge0
eris services start tor --machine bridge1
...

and so on, until you have 10 of each an exit, bridge, and middle relay.

As an aside, it would be trivial to make a new service definition file for each type of relay named tor_exit.toml, tor_bridge.toml, tor_middle.toml that would be marshalled by eris services start tor_exit/tor_bridge/tor_middle, respectively.

This avoids having to edit the service definition file every time a different relay is to be run. At any rate, this should give you a good idea of how to leverage docker-machine on the eris platform.

Coming Soon

But wait! What about the hassle of having to eris services start NAME --machine number99 a service many times over, not to mention that each machine had to have been created manually from the command line? Not to fret, the command eris remotes is in our pipeline and will likely take the form of a configuration file to specify N machines in X region from Y cloud provider, starting services Z+n; be it a chain, ipfs, bitcoin, or any combination of any processes that can be dockerized. Stay tuned!

Edit this page