_____ __ ____
/ ___// /____ / / /___ ______
\__ \/ __/ _ \/ / / __ `/ ___/
___/ / /_/ __/ / / /_/ / /
/____/\__/\___/_/_/\__,_/_/
Simplified Container Runtime Cluster
Stellar is designed to provide simple container runtime clustering. One
or more nodes are joined together to create a cluster. The cluster
is eventually consistent making it ideal for transient workloads or edge
computing where nodes are not always guaranteed to have high bandwidth, low
latency connectivity.
Why
There are several container platforms and container orchestrators out there.
However, they are too complex for my use. I like simple infrastructure that
is easy to deploy and manage, tolerates failure cases and is easy to debug
when needed. With the increased tolerance in failure modes, this comes at
a consistency cost. It may not be for you. Use the best tool for your use
case. Enjoy :)
Features
- Container execution via containerd
- Multihost Networking via CNI
- Service Discovery via DNS
- Builtin Proxy using Blackbird (zero downtime reloads, canary deploys, health checks, automatic HTTPS)
- Masterless design
- Efficient use of system resources
- Simple daemon deployment
Building
In order to build Stellar you will need the following:
Once you have the requirements you can build.
If you change / update the protobuf definitions you will need to generate:
make generate
To build the binaries (client and server) run:
make
Docker
Alternatively you can use Docker to build:
To generate protobuf:
make docker-generate
To build binaries:
make docker-build
Running
To run Stellar, once you have a working containerd installation follow these steps:
- Install Containerd
- Build binaries or get a release
- Copy
/bin/sctl
to /usr/local/bin/
- Copy
/bin/stellar
to /usr/local/bin/
- Copy
/bin/stellar-cni-ipam
to /opt/containerd/bin/
or /opt/cni/bin
To start the initial node run stellar -D server --bind-addr <node-ip> --advertise-addr <node-ip>
.
For example:
$> stellar -D server --bind-addr 10.0.1.70 --advertise-addr 10.0.1.70
To join additional nodes simply add the --peer
flag. For example:
$> stellar -D server --bind-addr 10.0.1.71 --advertise-addr 10.0.1.71 --peer 10.0.1.70:7946
You will now have a two node cluster. To see node information, use sctl
.
$> sctl --addr 10.0.1.70:9000 cluster nodes
NAME ADDR OS UPTIME CPUS MEMORY (USED)
stellar-00 10.0.1.70:9000 Linux (4.17.0-3-amd64) 7 seconds 2 242 MB / 2.1 GB
stellar-01 10.0.1.71:9000 Linux (4.17.0-3-amd64) 6 seconds 2 246 MB / 2.1 GB
Deploying an Application
To deploy an application, create an application config. For example, create the following as example.conf
:
{
"name": "example",
"labels": [
"env=prod",
"region=us-east"
],
"services": [
{
"name": "redis",
"image": "docker.io/library/redis:alpine",
"runtime": "io.containerd.runtime.v1.linux",
"process": {
"uid": 0,
"gid": 0,
"args": ["redis-server"]
},
"labels": [
"env=prod"
],
"network": true
}
]
}
Then run the following to deploy:
$> sctl --addr 10.0.1.70:9000 apps create -f ./example.conf
You should now see the application deployed:
$> sctl --addr 10.0.1.70:9000 apps list
NAME SERVICES
example 1
$> sctl --addr 10.0.1.70:9000 apps inspect example
Name: example
Services:
- Name: example.redis
Image: docker.io/library/redis:alpine
Runtime: io.containerd.runtime.v1.linux
Snapshotter: overlayfs
Labels:
containerd.io/restart.status=running
stellar.io/application=example
stellar.io/network=true
By default all applications that have networking enabled will have a corresponding nameserver record
created. To view the records use the following:
$> sctl --addr 10.0.1.70:9000 nameserver list
NAME TYPE VALUE OPTIONS
example.redis.stellar A 172.16.0.4
example.redis.stellar TXT node=stellar-00; updated=2018-09-08T10:71:02-04:00