zscaler aims to be an environment-agnostic, simple and flexible scaler. It plugs itself on any existing infrastructure, probe metrics and scale services according to configured rules by issuing orders to orchestration engines. Currently, the only supported target is the docker engine.

• A running docker engine
• docker-service scaler require docker 1.12 (with --replica)
• docker-compose >1.5 (with scale) in your path

Clone the repository in your $GOPATH or anywhere else

git clone [email protected]:Zenika/zscaler.git

Build it

make docker # build in docker and create docker image

Run an example with træfik on Docker

cd ./examples/docker-compose/traefik && docker-compose up

Now you can open the Træfik web UI and watch the backend scale up and down in real time!

This demo use a mock probe that report sinus-like values over 1min and causes the backend whoami to scale between 1 and 10 containers. The other probe is monitoring the CPU used across all whoami2 containers. You can stress-test it with the following benchmark:

ab -c 100 -n 10000000 -H 'Host:whoami2.docker.localhost' http://localhost/

Please refer to the dedicated wiki page for details.

Use-case configuration files can be found under the examples/ folder. Here's a sample:

orchestrator:
    engine: "docker"                   # only docker is supported ATM
    endpoint: "tcp://localhost:2376"   # optionnal, default to unix socket
    tls-cacert: "ca.pem"               # if using TLS
    tls-cert: "cert.pem"
    tls-key: "key.pem"
scalers:                               # scaler section
    whoami-compose:                    # custom id
        type: "docker-compose"         # what do we use to scale the service ?
        project: "traefik"             # parameter for docker-compose
        target: "whoami"               # parameter for docker-compose
        config: "docker-compose.yaml"  # parameter for docker-compose
    whoami2-compose:
        type: "docker-compose"
        project: "traefik"
        target: "whoami2"
        config: "docker-compose.yaml"
rules:                                 # rule section
    whoami-cpu-scale:                  # custom name of the service
        target: "whoami"               # name of service as tagged in orchestrator
        probe: "swarm.cpu_average"     # probe to use
        up: "> 0.75"                   # up rule
        down: "< 0.25"                 # down rule
        scaler: whoami-compose         # refer to any scaler id defined above
        refresh: 3s                    # scaler refresh rate
    whoami2-cpu-scale:
        target: "whoami2"
        probe: "cmd.execute"           # probe can be any binary
        cmd: "./some_script.sh"        # retrieve and write a float to stdout
        up: "> 200"
        down: "< 1.67"
        scaler: whoami2-compose
        refresh: 10s

$ zscaler [command]
Available Commands:
  dumpconfig  Dump parsed config file to stdout
  start       Start autoscaler
  version     Display version number
Flags:
  -d, --debug   Activate debug output

A REST API is available at startup, listening on :3000 (change it with -l or --api-port flag). You can find examples on the wiki page.

You'll need Go 1.5 or above. Older version may work but are still untested.

• Install Goalang and set you $GOPATH
• Run go get github.com/Zenika/zscaler and do

make all

This will download all dependencies and install the binary in $GOPATH/bin.

If you have a docker engine, you can build zScaler inside a container and run it as a docker image. To do so run:

make docker

The golang build image is tagged zscaler-build and the production image zscaler.

Some ansible scripts where crafted ahead of development to bootstrap cluster deployment when needed. You can find them under the deploy/ directory.

Project supported by Maximilien Richer, supervised by Sylvain Revereault (Zenika Rennes)