serviceDiscoveryAgent

This is the service discovery agent. If for some reason your application or service can't report to zookeper about itself, you can leverage this code to do so. It runs on the server/instance and monitors configured target(plugin), if target running/available this agent creates ephemeral zookeeper node in configured zookeeper. If healtcheck fails for some reason, zookeeper node gets removed. Same behaviour occurs if agent for some reason crash or lost connection with zk. It's also place information about your service into that zk node for later usage by others services.

IMPORTANT

This piece of software was written for fun and never being used in production.

How it works?

It checks the target and if getting success response, creates ephemeral zookeeper node(you have to have zookeeper server being installed) with data from zookeeper/zkData config value. If target for some reason became unavailable or agent dies - that node is getting removed. The interval between check as configured in interval config key.

Twitter-server

It's works on top of twitter server, so you can get all goodies twitter-server provides for free. Check out http://localhost:9990/admin

Pre-requirements

1. Installed sbt
2. Installed java machine

Requirements

1. Apache Zookepeer
2. Java Machine

Build jar file

To build fat jar file with all dependencies:

sbt assembly

new jar will be placed into target/scala-2.11

Running from jar

You have to have java installed on that server.

java -jar target/scala-2.11/serviceDiscoveryAgent-assembly-1.0.jar -config example.json

Configuration

By default agent looks for config at /etc/serviceDiscovery/serviceDiscovery.json, but this behaviour could be override by option -config. It works with json format and not automatically re-read by agent on the change. So you have to restart it manually every each change.

Name	Value(s)	Description
zookeeper/Base	"/services/web"	path to zookeeper base
zookeeper/zkData	"{"127.0.0.1": "80"}"	value to put into new node
target/ProcessCheck/processPidFile	"/tmp/123.pid"	path to process pid file (ProcessCheck only)
target/checkName	"ProcessCheck	HttpCheck"
target/target	localhost:80/health"	http endpoint (HttpCheck only)
target/responseCode	200	Success HTTP response code (HttpCheck only)
target/requestTimeout	5	Request timeout (HttpCheck only)
interval	10	interval between check in seconds

Macros

Is not implemented yet.

Targets

each instance can have only one target setup. So far there are only two targets available ProcessCheck and http HttpCheck.

Process

This plugin is monitoring configured process. It has extra parameter mandatory processPidFile. It must be pointed to monitored process pid file. Example:

  "processPidFile": "/tmp/123.pid",
  "checkName": "ProcessCheck",

HTTP healthcheck

This plugin is checking http page setup in checkTarget and expects to get checkResponseCode. two config params should be set checkTarget and checkResponseCode. Example:

  "checkName": "HttpCheck",
  "checkTarget": "localhost:80/health",
  "checkResponseCode": 200,

Configuration examples

Local web server

---

  "zookeeper": {
    "base": "/services/web",
    "connectString": "localhost:2181",
    "connectTimeout": 1,
    "sessionTimeout": 10,
    "zkData": "{\"%%hostname%%\": \"8080\"}"
  },
  "target": {
    "name": "HttpCheck",
    "params": {
      "target": "localhost:8080/health",
      "responseCode": 200,
      "requestTimeout": 5
    }
  },
  "interval": 10

Local process

  "zookeeper": {
    "base": "/services/web",
    "connectString": "localhost:2181",
    "connectTimeout": 1,
    "sessionTimeout": 10,
    "zkData": "{\"%%hostname%%\": \"12345\"}"
  },
  "target": {
    "name": "ProcessCheck",
    "params": {
      "processPidFile": "/tmp/123.pid"
    }
  },
  "interval": 10

TODO

• better handle zookeeper failure
• implement ip macro for Linux
• fix application metrics
• add unit testing

Development

Compile and run

cd serviceDiscoveryAgent
sbt "run -config example.json -log.level=DEBUG"