Skip to content

Latest commit

 

History

History
258 lines (198 loc) · 9.19 KB

deployment-guide.md

File metadata and controls

258 lines (198 loc) · 9.19 KB

Riff Platform AWS deployment as Docker swarm

Deploying

We include two scripts to make deployment to a public Docker swarm "easy": cloudformation.py and swarm.py. (if you have problems getting those scripts to run see the setup python doc.) You will also need to have docker and the AWS CLI installed.

The basic usage for deployment involves:

  1. Provision and start a cloudformation stack, creating a docker swarm on AWS
  2. SSH tunnel to a manager in the cloudformation docker swarm you just made
  3. Initialize the swarm by making a registry and starting everything in swarm mode
  4. Build and push docker images to the remote registry in your swarm
  5. Deploy the docker stack to the docker swarm running in the cloudformation stack. (Note that there are 2 different types of stacks referenced in this step!)

I go through the steps in detail below:

1. Provision and start a cloudformation stack, creating a docker swarm on AWS

By default the cloudformation script will use the cloudformation template provided by docker, see docker for AWS.

To provision a cloudformation stack using cloudformation.py, first make sure you have your AWS credentials configured correctly by running aws configure, and make sure you know the name of the key you'll use to SSH to the created instances.

To create a swarm with 4 worker nodes that are t2.micro instances, 3 (the default) managers that are t2.small instances, name it riffdeploy, and use the ssh key pair named "dan", run:

cloudformation create_docker_stack -s 4 -i t2.micro -I t2.small riffdeploy dan

You can monitor the creation status of your swarm using:

❯ cloudformation stack_status riffdeploy
CREATE_IN_PROGRESS

once the status is CREATE_COMPLETE, you can move onto the next step.

2. SSH tunnel to a manager in the cloudformation docker swarm you just made

To ssh tunnel into your swarm and use the remote manager as your DOCKER_HOST, you can run:

cloudformation.py tunnel -i ~/attic/aws/dan.pem riffdeploy

You should see output like:

❯ cloudformation.py tunnel -i ~/attic/aws/dan.pem riffdeploy
CREATE_COMPLETE
Instance IDs:
        0) i-02f4bd917994cf277
        1) i-0e13c196d1d0b8c59
        2) i-0f5153edf558f3af6
Enter the instance index:

The swarm you created has some number of manager nodes. To run properly, you need to tunnel into just one. It usually doesn't matter which one you choose.

❯ cloudformation.py tunnel -i ~/attic/aws/dan.pem riffdeploy
CREATE_COMPLETE
Instance IDs:
        0) i-02f4bd917994cf277
        1) i-0e13c196d1d0b8c59
        2) i-0f5153edf558f3af6
Enter the instance index: 0
Manager IP: 54.91.27.98
No existing docker tunnel ssh processes found. Continuing to tunnel...
Continue tunneling? (y/n): y

You should see some output like this:

>> tunneling using:  ssh -i /home/dcalacci/attic/aws/dan.pem -f -NL localhost:2374:/var/run/docker.sock docker@54.91.27.98
The authenticity of host '54.91.27.98 (54.91.27.98)' can't be established.
ECDSA key fingerprint is SHA256:GxwBwJ358jbH8eLUpn8mpkFkYsEDK5mSsHG+GNqnbiU.
Are you sure you want to continue connecting (yes/no)? yes


Tunnel up.
Matching tunnel process IDs:
26414
To kill this tunnel, use the command kill_tunnel

 To use this tunnel with your local docker config, run:

         export DOCKER_HOST=localhost:2374

 To reset to your local environment, run:

         export DOCKER_HOST=

To complete tunneling, do what the output says, and run:

export DOCKER_HOST=localhost:2374

3. Initialize the swarm by making a registry and starting everything in swarm mode

To initialize the swarm, you can run:

❯ ./swarm.py init

❯ ./swarm.py init
Initializing docker swarm...
> docker swarm init
Starting registry at localhost:5000 named registry
> docker service create --name registry --publish ppublished=5000,target=5000 registry:2
Error response from daemon: This node is already part of a swarm. Use "docker swarm leave" to leave this swarm and join another one.
Registry 'registry' up, docker in swarm mode
Ready to deploy to local swarm.

4. Build and push docker images to the remote registry in your swarm

To build and push a configuration from a docker stack config named production.yml, run:

❯ ./swarm.py push -f production.yml

You should see output like this, while it builds your image. It's being built on the remote manager node:

❯ ./swarm.py push -f production.yml

Building from production.yml
Continue? (y/n): y
WARNING: Some services (mongo-server, riff-rtc, riff-server, signalmaster, visualizer, web-server) use the 'deploy' key, which will be ignored. Compose does not support deploy configuration - use `docker stack deploy` to deploy to a swarm.
visualizer uses an image, skipping
mongo-server uses an image, skipping
Building riff-server
Step 1/50 : FROM node:8
8: Pulling from library/node
3d77ce4481b1: Pull complete
7d2f32934963: Pull complete
0c5cf711b890: Pull complete
9593dc852d6b: Pull complete
4b16c2786be5: Pull complete
5fcdaabfa451: Pull complete
5c8b2b2e4dd1: Pull complete

Eventually, it should push, and you should see a message like this:

.
.
.
Pushing signalmaster (127.0.0.1:5000/signalmaster:latest)...
The push refers to repository [127.0.0.1:5000/signalmaster]
9cfe4beca73b: Pushed
1bbd865744b9: Pushed
e6b14dd6688f: Pushed
9ef882f5c2ea: Pushed
9dfa40a0da3b: Pushed
latest: digest: sha256:4dcf926009f0750c730f2ae9ef5d135eb66d9abe6c7a6bb5d73ddd373d00f869 size: 1366

To deploy the images you just pushed, run `swarm.py deploy`

5. Deploy and bring your swarm up!

To finally deploy, use swarm.py deploy. You have to give your stack a name. To deploy production.yml and name it riffdeploy, run:

❯ ./swarm.py deploy -f production.yml -n riffdeploy

❯ ./swarm.py deploy -f production.yml -n riffdeploy
Have you already pushed this project to your repo? (y/n): y
Ignoring unsupported options: build

Creating network riffdeploy_default
Creating service riffdeploy_mongo-server
Creating service riffdeploy_signalmaster
Creating service riffdeploy_visualizer
Creating service riffdeploy_web-server
Creating service riffdeploy_riff-server
Creating service riffdeploy_riff-rtc

If you now run docker service ls, you should see the status of your services:

❯ docker service ls
ID                  NAME                      MODE                REPLICAS            IMAGE                                PORTS
4l7hp7iqdfmt        riffdeploy_signalmaster   replicated          1/1                 127.0.0.1:5000/signalmaster:latest   *:0->8888/tcp
hpnpv8trec3t        riffdeploy_web-server     replicated          1/1                 127.0.0.1:5000/web-server:latest     *:80->80/tcp
q4w9u2di1fmv        riffdeploy_mongo-server   replicated          1/1                 mongo:latest
q9eou1krhfa9        riffdeploy_visualizer     replicated          1/1                 dockersamples/visualizer:stable      *:8080->8080/tcp
t0gb5ujssskj        registry                  replicated          1/1                 registry:2                           *:5000->5000/tcp
w92gmp9ne454        riffdeploy_riff-rtc       replicated          0/1                 127.0.0.1:5000/riff-rtc:latest       *:0->3001/tcp
xw1xvel2ii6r        riffdeploy_riff-server    replicated          0/1                 127.0.0.1:5000/riff-server:latest    *:0->3000/tcp

6. Get the public DNS, and test!

You can get the public DNS of your swarm using cloudformation.py get_stack_dns. To get the DNS of the stack we just deployed, named riffdeploy, run:

❯ cloudformation get_stack_dns riffdeploy
riffdeplo-External-QH833A1J8HE8-1547422939.us-east-1.elb.amazonaws.com

This URL is the url you can use to access your newly deployed app!

Bugs

init failed but continued.

(venv) mjl@lm18x ~/Projects/riff/riff-docker $ ./swarm.py init
Initializing docker swarm...
> docker swarm init
Starting registry at localhost:5000 named registry
> docker service create --name registry --publish published=5000,target=5000 registry:2
Error response from daemon: could not choose an IP address to advertise since this system has multiple addresses on different interfaces (10.0.2.15 on enp0s3 and 192.168.56.101 on enp0s8) - specify one with --advertise-addr

Failed to create registry. You may already have another one running:                                                    
Error response from daemon: This node is not a swarm manager. Use "docker swarm init" or "docker swarm join" to connect this node to swarm and try again.
Do you  want to use an existing registry? (y/n): n
Enter registry name: registry
Registry 'registry' up, docker in swarm mode
Ready to deploy to local swarm.