Skip to:

  1. Skip to Jira Navigation
  2. Skip to Side Navigation
  3. Skip to Main Content

The timer function from the Drools correlation engine seems to be broken.

Description

I run a few rule sets in drools on OpenNMS 15.0.2 with the rule attribute timer, since the upgrade to OpenNMS 16.0.4 they doesn't work correctly. It seems that these rules runs , when an event for the rule set appears but not when the timer should trigger it.
I have write a simple rule set for tests and it confirmed the behavior.

Rule set for test:

package org.opennms.netmgt.correlation.drools;

import java.util.Date;
import java.text.SimpleDateFormat;
import java.util.Timer;
import org.opennms.netmgt.xml.event.Event;
import org.opennms.netmgt.xml.event.Parm;
import org.opennms.netmgt.model.events.EventBuilder;
import org.opennms.netmgt.events.api.EventConstants;

global org.opennms.netmgt.correlation.drools.DroolsCorrelationEngine engine;

rule "catch rule"
when
$e : Event( uei == "uei.example.org/clock/clocktest" )
then
SimpleDateFormat sdf = new SimpleDateFormat("dd.MM.yyyy-hh:mm:ss");
System.out.println("DEBUG: event was catched " + sdf.format(new Date()));
retract($e);
end

rule "timetest Cron" timer (cron: * 0/1 * * * ?) no-loop
when
then
SimpleDateFormat sdf = new SimpleDateFormat("dd.MM.yyyy-hh:mm:ss");
System.out.println("DEBUG: timetest cron :" + sdf.format(new Date()));
end

rule "timetest int" timer ( int:0 1m) no-loop
when
then
SimpleDateFormat sdf = new SimpleDateFormat("dd.MM.yyyy-hh:mm:ss");
System.out.println("DEBUG: timetest int :" + sdf.format(new Date()));
end

And the drools-engine.xml - file

<?xml version="1.0" encoding="UTF-8"?>
<engine-configuration
xmlns="http://xmlns.opennms.org/xsd/drools-engine"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://xmlns.opennms.org/xsd/drools-engine http://xmlns.opennms.org/xsd/drools-engine.xsd ">
<rule-set name="testclock">
<rule-file>testclockRules.drl</rule-file>
<event>uei.example.org/clock/clocktest</event>
</rule-set>
</engine-configuration>

Environment

On an dedicated dell server with two Intel(R) Xeon(R) CPU E5-2630 v3 , 16 GB RAM. I tested it as well in a VirtualBox-VM with 1 CPU and 4GB RAM. Both installation was base on ubuntu 14.0.4 LTS.

Acceptance / Success Criteria

None

Attachments

6
  • 22 Dec 2015, 09:47 AM
  • 22 Dec 2015, 09:47 AM
  • 22 Dec 2015, 09:47 AM
  • 08 Dec 2015, 08:59 AM
  • 08 Dec 2015, 08:59 AM
  • 08 Dec 2015, 08:59 AM

Confluence content

mentioned on

Lucidchart Diagrams

Activity

Show:

Andreas Fuchs June 9, 2016 at 3:25 AM

That was my first try to solve this problem.
I made an intensive debugging and rewrite the rules a few times,
but I didn't get the timer function run, which was the reason for me to lock in the source code.
Currently, the issue has a low priority for me, because I will use a python script to forward events, with a defined delay, to our SMS gateway ( the original use of the first rule set).
I didn't close this bug, because it still exists in 16.0.4 and 17.0.0.
I didn't check if it still exists in the newer Versions of OpenNMS.

Seth Leger June 8, 2016 at 2:46 PM

Is this an issue that can be solved by updating your rules scripts?

Andreas Fuchs December 23, 2015 at 7:28 AM

In an on-road test I figured out, that something changed by use of the rule-attribute no-loop.
In the actual version, this attribute seams to stop a timer triggered rule that catch a unmodified fact-object in the workspace.
By deleting it, the rule seams to work normal (expect you use the a cron timer, than good luck).

Andreas Fuchs December 22, 2015 at 9:47 AM

I have exterminate the source of OpenNMS and found the file
opennms/opennms-correlation/drools-correlation-engine/src/main/java/org/opennms/netmgt/correlation/drools/DroolsCorrelationEngine.java
where no session option is set. So I made a patch (see attached DroolsCorrelationEngin.java ) and test a little bit around with the OpenNMS version 17.0.1 and 16.0.4. You can find the output of my debug messages in output.log.

The int timer seems to work fine now, but the behaviour of the cron timer is very strange ( you can see it in the log files ). Partly the rules run like expected ( test 03 and 04 ),
partly the rule match every second or have unexplainable delays.

It is possible to use this as a workaround if you substitute the cron timer with the int timer, but I didn't test it for a long time, so no guarantee.
If someone can give me a tip to handle the cron problem I would be happy.

Andreas Fuchs December 11, 2015 at 11:30 AM

I have made some tests with Drools 6.0.1 Final (stand-alone) with Java 1.7.0_91 and 1.8.0_60 to check the timer function.
I used stateful sessions like opennms and recognized that the behavior is similar to the issue. In the Drools documentation is an part that seems to explain it ( https://docs.jboss.org/drools/release/6.0.1.Final/drools-docs/html/DroolsReleaseNotesChapter.html#d0e592 ).
In my test I set the option TimedRuleExectionOption.YES to the session configuration and the timer do there work correctly.
Because I cant find the part of the OpenNMS sourcecode that where the drools correlator is configured, I can't check if this option is set or must be set. Perhaps a completely different way is used.
Can anyone help me with this?

Details

Assignee

Reporter

Labels

Doc Backlog Status

Doc Backlog Grooming Date

Components

Affects versions

Priority

PagerDuty

Created October 23, 2015 at 11:22 AM
Updated September 21, 2021 at 6:24 PM

Flag notifications