Yetis WIP

Kubernetes for linux processes running on a single machine.

Use Case

Your VPS doesn't support Docker containers, but you would still like some k8s features.

Features of Yetis

Kubernetes-like declarative configuration.
Self-healing. Automatically restarts failed processes. Kills and recreates unresponsive processes.
Log management. It saves the standard output into iterative log files.
Zero downtime deployment, achieved with RollingUpdate strategy.

Installing

sudo wget -O /usr/local/bin/yetis https://github.com/glossd/yetis/raw/refs/heads/master/build/yetis && chmod +x /usr/local/bin/yetis

Commands

Start Yetis server

sudo yetis start

*Must be root to configure proxy
Yetis will start in the background. You can pass yetis configuration with -f flag.

Available commands

Deploy your process:

yetis apply -f config.yaml

Configuration

List the managed processes

yetis list will show the list of the processes.
Add flag -w to watch the updates

Full list of commands

Server Commands:
	start [-d]              start Yetis server
	shutdown                terminate Yetis server
	info                    print server status
Resources Commands:
	apply -f FILENAME       apply a configuration from yaml file
	list [-w]               print a list the managed deployment
	logs [-f] NAME          print the logs of the deployment with NAME
	describe NAME           print a detailed description of the selected deployment
	delete NAME             delete the deployment, terminating its process
	restart NAME            restart the deployment according to its strategy type 
	help                    print the list of the commands

Configuration examples

A simple process to watch over and restart if port becomes unavailable

spec:
  name: frontend # must be unique
  cmd: npm start
  workdir: /home/user/myfront
  livenessProbe:
    tcpSocket:
      port: 3000

Zero-downtime is achieved by specifying a proxy.port which creates an iptables port forwarding rule.
In the example below, all the traffic from port 2345 will be forwarded to the frontend process, which will be running on a dynamic port YETIS_PORT

spec:
  name: frontend
  cmd: npm start
  workdir: /home/user/myfront
  strategy:
    type: RollingUpdate # For zero downtime
  env:
    - name: APP_PORT # Copies the value from YETIS_PORT to APP_PORT
      value: $YETIS_PORT
  proxy:
    port: 2345 # Forwards traffic from 2345 to the port in YETIS_PORT env var.

Full configuration

kind: Deployment # Defaults to Deployment
spec:
  name: hello-world # Must be unique
  preCmd: javac HelloWorld.java # Command to execute before starting the process.  
  cmd: java HelloWorld
  workdir: /home/user/myproject # Directory where command is executed. Defaults to the path in 'apply -f'. 
  logdir: /home/user/myproject/logs # Directory where the logs are stored. Defaults to the path in 'apply -f'.
  strategy:
    type: Recreate # Recreate or RollingUpdate. Defaults to Recreate.
  livenessProbe: # Checks if the command is alive and if not then restarts it
    tcpSocket:
      port: 8080 # Should be specified if proxy is not configured. Defaults to $YETIS_PORT 
    initialDelaySeconds: 5 # Defaults to 10
    periodSeconds: 5 # Defaults to 10
    failureThreshold: 3 # Defaults to 3
    successThreshold: 1 # Defaults to 1
  env: # YETIS_PORT env var is passed by default. You should use alongside proxy config. 
    - name: SOME_SECRET
      value: "pancakes are cakes made in a pan"
    - name: SOME_PASSWORD
      value: mellon
    - name: MY_PORT
      value: $YETIS_PORT # pass the value of the environment variable to another one.
  proxy:
    port: 8080 # Tells linux to forward from the specified port to $YETIS_PORT, allowing zero downtime restarts.

Liveness Probe

Checks if the process is alive and ready. Yetis relies on this configuration to restart the process. Plus if proxy.port is configured, then forward traffic to the deployment. For now, the probe only supports tcpSocket. Liveness also acts as Readiness and StartUp probes.

Deployment Strategies

Zero downtime is achieved with RollingUpdate strategy and restart command. Your deployment must start on $YETIS_PORT and have a proxy.port configured. restart command will spawn a new deployment, check if it's healthy, then direct traffic to the new instance, and only then will terminate the old instance. The new deployment will have the name with an index i.e. frontend-1, frontend-2 and so on. If deployment has Recreate strategy, Yetis will wait for the termination of the old instance before starting a new one with the same name. It's the same as in Kubernetes

Yetis Server Configuration

Provide configuration when starting Yetis: yetis start -f /path/to/config.yml

Alerting

You can configure server to send you alerts of your deployment's failure or recovery from failure. For now, yetis only supports mail alerting through SMTP.

Full Yetis Configuration

logdir: /tmp # yetis.log will be stored in there. Defaults to /tmp
alerting: # Alerts when a managed process fails or recovers.
  mail: # add SMPT creds of your smpt server for alerting
    host: smtp.host.com
    port: 587
    from: [email protected]
    to:
      - [email protected]
    username: authUser
    password: authPass

# Packages

# README