I have updated this repository to be a demonstration of a docker swarm that uses a few tools together to deliver a simple load balanced solution. It uses the following technologies:
- Docker
- Docker Machine
- Docker Swarm
- Consul
- Registrator
- Tyk
To give you the most details I can about this system I will go over how to stand it up, how all the pieces interact and how the final product will perform.
This repository is created to give you a system that you can use to test a docker cluster (setup using dockers supported swarm). To be able to use this repository you will need to have a virtualmachine setup that docker-machine can use to create its instances, this can be any of the below options:
VirtualBox: virtualbox
VMWare Fusion: vmwarefusion
XHyve: xhyve
There are 2 main operations that exist in the base script.
./setup-cluster start
This operation will create a docker machine that has a consul service used to track the swarm nodes and other nodes added to the
swarm. This consul service can be opened by running open http://$(docker-machine ip consul):8500/. It will then create a docker
swarm that consists of 3 nodes: swarmmaster, swarms11, swarms2. This swarm has the swarmmaster node as the swarm master
(yeah I know, figured I would call it out anyways). This swarm is created using an overlay network which means all boxes should
have an address 10.0.0.x that they can use to communicate with each other. After creating the swarm it will then install consul
onto every node to actas the service discovery for each service (and dns on the boxes). Then it installs registrator onto each
of the nodes to have them automatically register the started services with consul. Finally it will stand up the services necessary
for tyk to run which include: redis, mongo, tyk_gateway and tyk_dashboard.
Once you are finished with a cluster you can free up the machines and destroy everything by running the below command:
./setup-cluster stop
The default driver that will be used is virtualbox, however if you would like to use vmware or xhyve just use the below examples.
bash
DOCKER_MACHINE_DRIVER=vmwarefusion ./setup-cluster start
DOCKER_MACHINE_DRIVER=xhyve ./setup-cluster start
fish
env DOCKER_MACHINE_DRIVER=vmwarefusion ./setup-cluster start
env DOCKER_MACHINE_DRIVER=xhyve ./setup-cluster start
So now you have the cluster running, lets go through a few commands to see how utilize this swarm cluster.
Once you have a cluster up and running you can start to deploy services to this cluster. This can be accomplished by running the following command.
bash
eval $(docker-machine env --swarm swarmmaster)
fish
eval (docker-machine env --swarm swarmmaster)
This will setup a connection so that all docker command will be run against the cluster and not against a single machine. After this you just
need to run your docker run operation to get the new service up an running. For our example we have a flask app that has been included
in the repository so just un the following commands.
cd webapp/
docker-compose up -d
docker-compose scale flaskapp=3
We have created our simple webservice and scaled it to 3 instances. To see these all running, you just need to run the below command.
docker ps
Since this new service will automatically be registered to consul by registrator so we should be able to see this in the consul
ui which can be opened by running the following command:
open http://$(docker-machine ip swarmmaster):8500/ui
We can also see the results of this service using the consul http api by running the curl command listed below.
curl http://$(docker-machine ip swarmmaster):8500/v1/catalog/service/webapp_flaskapp
We need this endpoint to return the list of services that we started (since we spun up 3 there should be 3 that are returned).
Now we have our new webapp registered as a service in consul, we can use tyk to be our api frontend and have a custom endpoint forward to our webapp.
The first thing you will need to do is to login to the tyk dashboard portal. The portal address and the username, password should have been
at the end of the output from ./setup-cluster start. Right now the username defaults to [email protected] and the password defaults to changeme3.
Once you have the portal open you will want to click on the Apis link that is under the System Management section.
Next click on the Add new API button that can be found on the upper right of the page.
You should now be in the API Designer page. At this point you will want to fill out the form providing an API Name and a corresponding Listen Path.
(Note: the listen path should be automatically matched to the API Name). You will also want to fill in the Target URL. If you are planning to use
the service discovery feature (through consul) then you will want the Target URL to be /.
If you are going to use service discovery then click on the checkbox next to Enable service discovery in the section Service Discovery. To make
this work with consul you will need to fill in the Query Endpoint with a url similar to the one below.
http://x.x.x.x:8500/v1/catalog/service/webapp_flask
Note: If you need to get the ip address you can get it by running docker-machine ip swarmmaster
Click the checkbox next to the entry Does this endpoint return a list?. In the Data path field you want to put ServiceAddress. Click on the
checkbox next to Is port information separate from the hostname?, and then in the Port data path: field you want to put ServicePort.
Finally in the Target Details section you will want to change the Authentication Mode to be Open (keyless).
Click on the Save button which can be found at the upper right of the page and that will have created a new api.
To use this api you will want to go through the tyk_gateway. The gateway ip should have been printed out with the tyk_dashboard login details.
To test out our new service open curl and run the following command.
curl http://x.x.x.x:8080/<API Name>/
There is a possibility that if you have used vagrant on you machine before then docker-machine will have problems with virtualbox, as vagrant seems to disable the DHCP server option of the host-only network. More info can be found here
There are a number of great references that exist online that I used when going through this process including those listed below.