User Interface

The user interface of Observatory is divided into two parts: input and output. For input, we read events from various sourced, configured in the Manifest. The output part is the front-end, the graph visualization.

Input: Sending events

Sending input to Observatory is easy. At this stage, you just send an event via HTTP.

Here is a sample payload:

{
   "origin": "web-server",
   "tracing": "1234abcd",
   "timestamp": "2016-11-13T22:00:00+02:00"
}

The fields are the following:

origin:The application where the event occurred tracing:The tracing token unique for this event timestamp:A ISO 8601 date when the event occurred.

This payload can be sent to Observatory with a simple POST request to the /events endpoint.

$ curl -X POST -H 'Content-type: application/json' -d@event.json http://observatory/events

HTTP/1.1 202 Accepted
Server: Observatory 0.1
Date: Sun, 13 Nov 2016 21:46:08 GMT
Content-Type: application/json; charset=utf-8

Output: Getting status

The monitor is the Observatory front-end. As we saw in the example graph, at this point, the interface is just a constantly refreshing Graphviz rendering of the graph.

Internally, Observatory is accessed with a REST API. If we configure a system called my-sample-system, we get its graph as JSON like this:

$ curl -i http://observatory/streams/my-sample-system
HTTP/1.1 200 OK
Server: Observatory 0.1
Date: Sun, 13 Nov 2016 21:46:08 GMT
Content-Type: application/json; charset=utf-8

{
  "name": "my-sample-system",
  "description": "blah blah",
  "graph": [
    {
      "from": "web-server",
      "to": "event-processor",
      "status": "OK",
      "pass": 3,
      "total": 3,
      "last": "2016-11-14T11:33:32.000Z",
      "failed": "2016-11-12T17:22:32+02:00",
      "check": {
        "within": 10,
        "unit": "seconds",
        "ok": 3,
        "warn": 0,
        "fail": 0
      }
    },
    {
      "from": "event-processor",
      "to": "journal",
      "status": "NOK",
      "pass": 1,
      "total": 4,
      "last": "2016-11-14T11:00:00+02:00"
      "failed": "2016-11-14T11:33:32+02:00",
      "check": {
        "within": 500,
        "unit": "msec",
        "ok": 3,
        "warn": 1,
        "fail": 0
      }
    },
    {
      "from": "event-processor",
      "to": "database",
      "status": "WARN",
      "pass": 4,
      "total": 5,
      "last": "2016-11-14T12:00:00+02:00"
      "failed": "2016-11-14T12:01:22+02:00",
      "check": {
        "within": 500,
        "unit": "msec",
        "ok": 3,
        "warn": 1,
        "fail": 2
      }
    }
  ]
}

This can be then rendered in the front-end, which is just a dead simple HTTP server rendering this graph to a Graphviz page, with a meta refresh directive of 1 second.