Configuration

To enable this functionality, you must edit the opEvents "Event Actions" JSON file. We recommend using the web GUI to do this as there is a very handy "Validate" button that can be used to ensure your configuration changes are indeed valid JSON and won't break opEvents. If you must, you can also edit the JSON file directly at /usr/local/omk/conf/EventActions.json file. Beware that making changes that result in an invalid JSON file will result in your actions not functioning as intended.

Validating a JSON file on the command line can be done using the command:

python -mjson.tool /usr/local/omk/conf/EventActions.json

opCommon.json

Certain node properties must be defined in opCommon.json in order for opEvents to make use of them. In earlier versions for example, group and location we were at the root level. These are now under configuration so should be defined as below.

      "opevents_event_copy_node_properties" : [
         "configuration.group",
         "configuration.location"
      ],

Not like below.

      "opevents_event_copy_node_properties" : [
         "group",
         "location"
      ],

This is mostly of relevance to customers who have upgraded from opEvents 2.x.

Policy

There are a few sections in EventActions.json, but the one we're concerned with first is the policy section. Entries in a given section are numbered for determining the order in which they are processed. Usually we make forwarding the last entry (the entry with the highest number).

Check All

First we want to check all events and filter them in sub-sections. This just makes things easier for a human to read and understand. So our IF section at the first level is a simple node.any and event.any. For any node and any event. The THEN section is again numbered for order, that will then be processed.

Tag any up events

Next we need to decide which events we would like to send to the second instance. We're going to send (for the purposes of this example) all events with a node that has a serviceStatus of 'Production'.

These events will go into a second list, inside the top level.

Also, we should send all 'Node Up' or 'Closed' events to the second instance, regardless of the node so it cancels out any escalation. It does not matter that no 'down' event might not have been sent previously. So our JSON will now look like the below.

And one final item - the first thing we'll do is add two tags - escalateToCentral and sendToCentral. Adding them regardless of the event will allow the tests later on to check the values without throwing warnings in the logs because those tags don't exist.

Below we have added the tags and checked for any Node Up or Closed events.

Tag any Production node events

Now you may have noticed we have two items to action - escalateToMaster and sendToMaster. escalateToMaster sends an escalation where-as sendToMaster doesn't do anything other than simply send the event. No escalation required.

You may also have noticed the 'up|closed' entry. These entries take perl regular expressions, so in this case if up or closed is present in the event/state, it will match and trigger the THEN.

Next we'll add an event that we want to escalate. Let's check if the serviceStatus is production and if so, tag it for escalation.

That's all we'll bother with to keep things simple.

Check tags

Next we need to check the tags and if required, escalate or send it to the second instance.

Escalate

As you can see, we reference two different routines in the THEN sections for #30 and #40.

For the THEN action escalate.central in #30, we are simply checking if the event has a priority high enough to send and our hours of operation correspond. If it does, send it. This is defined in the escalate section of JSON.  The 360 refers to the escalation to take place after 360 seconds has passed.  Implementing like this will prevent flaps and noisy events from going to the central server.

Script

Any 'script.X' THEN items are defined in the script section of EventActions.json - hey, big surprise, I know

For #40, we don't bother testing for a priority, we just send them. This is because events such as 'up' or 'closed' have low priority - but obviously we want to send them to cancel and 'down' type events.

So our script blocks looks as below.

Secure Wrapper

If you wish to not include your username and password in the Event Actions file, you can use a wrapper script.

Create a script e.g. /usr/local/omk/bin/create_remote_event.sh and make the ownership and permissions for root only.

sudo touch /usr/local/omk/bin/create_remote_event.sh
sudo chown root:root /usr/local/omk/bin/create_remote_event.sh
sudo chmod 700 /usr/local/omk/bin/create_remote_event.sh

The contents of this script would be:

#/usr/bin/env bash

# username and password below, pass through all other arguments.
/usr/local/omk/bin/create_remote_event.exe -u "your_opevents_user" -p "your_opevents_password" $@

Event Actions Script would be updated as such

"script": {
    "sendToCentral": {
        "arguments": [
            "-s https://your_central_opevents_server/omk",
            "authority=macro.authority",
            "location=https://your_local_opevents_server/omk/opEvents/events/event._id/event_context",
            "host=node.host",
            "event=NETWORK-event.event",
            "details=event.details",
            "time=event.time",
            "date=event.date",
            "element=event.element",
            "interface_description=event.interface_description",
            "type=event.type",
            "priority=event.priority",
            "level=event.level",
            "state=event.state",
            "stateful=event.stateful"
        ],
        "exec": "/usr/local/omk/bin/create_remote_event.sh",
        "output": "save",
        "stderr": "save",
        "exitcode": "save",
    }
}

Finishing Up

And that's it!

You have selected events to be forwarded and tagged them. Tested those tags for actions and depending on the tag and priority, forwarded it to your central instance.

So for future versions, you might define an event on a node with serviceStatus of "Testing" and choose only to forward it during office hours (for example). Give it a different tag, check that tag and use escalate with an additional item (say "testing_office_hours_central"). The ways to configure opEvents really are limited only by your imagination.

Hopefully this article has given you some ideas and pointed you in the right direction.