a) Each Log without a CorrelatedEvent creates an activation. A
CorrelatedEvent will not apear until there are 20 Logs with the same src/dst combination.
This means that there are up to 19 pending accumulates until the first
one reaches the threshold, inserts the CorrelatedEvent and cancels the others.
b) Additional inefficiency is produced by the separate accumulation
of the markers. This set does not participate in LHS constraints, and hence could be
easily computed on the RHS.
I tried to remove accumulation of the markers.
These are the changes:
New function:
function Set getMarkers(Set matchedLogs) {
HashSet<String> markerSet = new HashSet<String>();
HashSet<String> idSet = new HashSet<String>();
for (Object matchedLogObj : matchedLogs) {
Log matchedLog = (Log) matchedLogObj;
String id = getUniqueId(matchedLog);
if (!idSet.contains(id)) {
idSet.add(id);
markerSet.add(matchedLog.fieldsMap.get("marker").toString());
if (markerSet.size() == 25) break;
}
}
return markerSet;
}
In LHS:
accumulate($accumulatedLog : Log(eval(filter($accumulatedLog)), this after[0s,60s] $log,
fieldsMap.get("src") == $log.fieldsMap.get("src"),
fieldsMap.get("dst") == $log.fieldsMap.get("dst"), $id :
getUniqueId(this));
$idSet : collectSet($id) , $matchedLogs:
collectSet($accumulatedLog);
$idSet.size > 19)
In RHS:
fieldsMap.put("cu_markers_list", getMarkers($matchedLogs));
$matchedLogs return as empty set. Why? How can I make it work?
Thanks.
-----Original Message-----
From: rules-users-bounces(a)lists.jboss.org [mailto:rules-users-bounces@lists.jboss.org] On
Behalf Of Wolfgang Laun
Sent: Sunday, October 27, 2013 7:56 PM
To: Rules Users List
Subject: Re: [rules-users] Drools memory consumption
A few remarks on the code snippets.
* function boolean filter: Is it a good idea to do a lot of operations that could be done
on 32-bit quantities (using bit operations) based on String values? Converting the log
values to int will pay back nicely.
* function String markersToString: i is never changed. StringBuffer? "\n"?
* String getUniqueId(Log log)
return log.fieldsMap.get("service") +
log.fieldsMap.get("service").toString()
would do.
* The rule.
a) Each Log without a CorrelatedEvent creates an activation. A CorrelatedEvent will not
apear until there are 20 Logs with the same src/dst combination. This means that there are
up to 19 pending accumulates until the first one reaches the threshold, inserts the
CorrelatedEvent and cancels the others.
b) Additional inefficiency is produced by the separate accumulation of the markers. This
set does not participate in LHS constraints, and hence could be easily computed on the
RHS.
This looks like a scenario where Drools is kept overly busy which, in turn, will create
some extra GC effort. Analysing the data and counting activations coming and going and
firing would shed some more light on this.
-W
On 27/10/2013, Elran Dvir <elrand(a)checkpoint.com> wrote:
Thanks for the response.
I will try to implement eval(filter(...)) in DRL.
I tried to change insert to insertLogical and the rule isn't working
correctly. The condition not CorrelatedEvent... is always true.
If I set expiration to CorrelatedEvent isn't Drools supposed to manage
CorrelatedEvent's memory consumption by itself?
Thanks.
From: rules-users-bounces(a)lists.jboss.org
[mailto:rules-users-bounces@lists.jboss.org] On Behalf Of Stephen
Masters
Sent: Thursday, October 24, 2013 5:49 PM
To: Rules Users List
Subject: Re: [rules-users] Drools memory consumption
I haven't dug through your code in serious detail, but I did notice
that you're using insert instead of insertLogical.
That could lead to huge numbers of CorrelatedEvent instances in
working memory, unless you have rules elsewhere to retract them.
>From a performance perspective, I would be very suspicious of your
eval(filter(...)) condition. It looks as though it could easily be
implemented in DRL.
And we all know that "eval" is only one vowel away from "evil". ;)
On 24 Oct 2013, at 15:03, Elran Dvir
<elrand@checkpoint.com<mailto:elrand@checkpoint.com>> wrote:
Hi All,
Can someone please take a look at my drl file and say if and how I can
improve its memory consumption?
I attach my case again.
For example, one rule is supposed to identify a port scan event.
The basic fact is connection log. For each combination of src (source
IP) and dst (destination IP) , detect a port scan event, if over 60
seconds there were at least 20 connection logs with different service and protocol.
The event will stay closed for 10 minute - no event will be sent
during this time for this combination of src and dst. The event the connection
logs'
ids (markers).
This is its drl file:
package com.checkpoint.correlation.impl.drools.package30;
import java.util.Date
import java.util.HashMap
import java.util.Set
import com.checkpoint.correlation.impl.drools.Log
import com.checkpoint.correlation.impl.drools.CorrelatedEvent
global com.checkpoint.correlation.server.EventsHandler
externalEventsHandler;
import function
com.checkpoint.correlation.impl.utils.UserDefinedFunctions.isInDayHour
Range
import function
com.checkpoint.correlation.impl.utils.UserDefinedFunctions.isInIpRange
function boolean filter(Log log) {
return (!((log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "10.80.0.0",
"10.80.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "124.0.0.0",
"124.255.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "192.168.0.0",
"192.168.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "195.158.7.0",
"195.158.7.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "11.25.0.0",
"11.25.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "128.157.0.0",
"128.157.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "213.114.0.0",
"213.114.255.255")))); }
function String markersToString(Set markersSet) {
int i = 0;
String markersString = "";
for (Object marker : markersSet) {
if (i == 25) break;
String markerStr = marker.toString();
if (i > 0) markersString += "\n";
markersString += markerStr;
}
return markersString;
}
function String calcSeverity(Log log) {
return "High";
}
function String getUniqueId(Log log) {
String uniqueId="";
uniqueId += (log.fieldsMap.get("service") != null ?
log.fieldsMap.get("service").toString() : "null");
uniqueId += (log.fieldsMap.get("proto") != null ?
log.fieldsMap.get("proto").toString() : "null");
return uniqueId;
}
declare Log
@role(event)
end
declare CorrelatedEvent
@role(event)
@expires(600s)
end
rule "Port scan from external network"
enabled true
dialect "java"
no-loop
when
$log : Log(eval(filter($log)))
not CorrelatedEvent(getId() ==
"{8AC52BA8-1EE8-4f18-9BB4-54492116501C}", groupByFieldsMap.get("src")
== $log.fieldsMap.get("src"), groupByFieldsMap.get("dst") ==
$log.fieldsMap.get("dst"))
accumulate($accumulatedLog :
Log(eval(filter($accumulatedLog)), this after[0s,60s] $log,
fieldsMap.get("src") == $log.fieldsMap.get("src"),
fieldsMap.get("dst") == $log.fieldsMap.get("dst"), $id :
getUniqueId(this));
$idSet : collectSet($id);
$idSet.size > 19)
accumulate($accumulatedLog :
Log(eval(filter($accumulatedLog)), this after[0s,60s] $log,
fieldsMap.get("src") == $log.fieldsMap.get("src"),
fieldsMap.get("dst") == $log.fieldsMap.get("dst"),
$idSet.contains(getUniqueId(this)), $marker :
fieldsMap.get("marker"));
$markerSet : collectSet($marker))
then
CorrelatedEvent $ce = new
CorrelatedEvent("{8AC52BA8-1EE8-4f18-9BB4-54492116501C}");
$ce.groupByFieldsMap.put("src",
$log.fieldsMap.get("src"));
$ce.groupByFieldsMap.put("dst",
$log.fieldsMap.get("dst"));
insert($ce);
HashMap<String,Object> fieldsMap = new
HashMap<String,Object>();
fieldsMap.put("cu_rule_id",
"{8AC52BA8-1EE8-4f18-9BB4-54492116501C}");
fieldsMap.put("event_name", "Port scan from external
network");
fieldsMap.put("cu_rule_severity", calcSeverity($log));
fieldsMap.put("cu_rule_category", "Scans");
fieldsMap.put("cu_log_count", $markerSet.size());
fieldsMap.put("time", new Date());
fieldsMap.put("cu_markers_list",
markersToString($markerSet));
fieldsMap.put("src", $log.fieldsMap.get("src"));
fieldsMap.put("src_machine_name",
$log.fieldsMap.get("src_machine_name"));
fieldsMap.put("src_user_name",
$log.fieldsMap.get("src_user_name"));
fieldsMap.put("dst", $log.fieldsMap.get("dst"));
fieldsMap.put("dst_machine_name",
$log.fieldsMap.get("dst_machine_name"));
fieldsMap.put("dst_user_name",
$log.fieldsMap.get("dst_user_name"));
fieldsMap.put("service",
$log.fieldsMap.get("service"));
fieldsMap.put("proto", $log.fieldsMap.get("proto"));
fieldsMap.put("product",
$log.fieldsMap.get("product"));
externalEventsHandler.handleEvent(fieldsMap);
end
Thanks.
From: Elran Dvir
Sent: Thursday, October 24, 2013 11:23 AM
To: Rules Users List
Subject: RE: [rules-users] Drools memory consumption
I am using 5.5.0.Final.
Thanks.
From:
rules-users-bounces@lists.jboss.org<mailto:rules-users-bounces@lists.j
boss.org>
[mailto:rules-users-bounces@lists.jboss.org<mailto:users-bounces@lists
.jboss.org>]
On Behalf Of Geoffrey De Smet
Sent: Thursday, October 24, 2013 10:46 AM
To: Rules Users List
Subject: Re: [rules-users] Drools memory consumption
Which version of drools?
Memory consumption improved a lot in 5.2 or 5.3 IIRC.
On 24-10-13 09:14, Elran Dvir wrote:
Thanks for the quick response.
I don't use entrypoints or windows in my rules.
I am using temporal constraints.
I can't retract the Log facts because I don't know whether other rules
need them.
Maybe It's because there is no expiration set for Log. But I am not
sure if I can.
(CorrelatedEvent has multiple expirations defined over the drl files)
I wrote this question two days ago and no one responded:
I have several drl files.
Each drl file contains one rule.
Each drl file is built to a knowledge package.
Each drl file contains declaration of the same fact (referencing to a
imported class defined outside of the drl). The fact's role is of
course event.
Can I define different expiration (@expires) for the fact in each drl?
Is expiration relevant only to the same drl?
Is the answer different if the fact is created in the rule's RHS
(CorrelatedEvent) or if it inserted to the session's working memory
outside of the drl (Log)?
Thanks.
From:
rules-users-bounces@lists.jboss.org<mailto:rules-users-bounces@lists.j
boss.org> [mailto:rules-users-bounces@lists.jboss.org] On Behalf Of
Matteo Cusmai
Sent: Thursday, October 24, 2013 9:32 AM
To: Rules Users List
Subject: Re: [rules-users] Drools memory consumption
Are you using any entrypoint?
Time or length based?
It could be possible you have to retract manually the events.
On 24 Oct 2013 08:23, "Elran Dvir"
<elrand@checkpoint.com<mailto:elrand@checkpoint.com>> wrote:
I am sending this message again because maybe the last wasn't sent
because of the attached snapshots.
I removed them now.
Thanks.
Hi all,
I am using Drools Fusion. I am getting OutOfMemoryError rather fast.
My JVM is running with -Xmx4g flag.
I have rules defined in another (not Drools) language.
Every rule is translated programmatically to a drl file. This is
because the user can add and remove rules (in the other language) dynamically.
The default configuration contains 125 rules.
For example, one rule is supposed to identify a port scan event.
The basic fact is connection log. For each combination of src (source
IP) and dst (destination IP) , detect a port scan event, if over 60
seconds there were at least 20 connection logs with different service and protocol.
The event will stay closed for 10 minute - no event will be sent
during this time for this combination of src and dst. The event the connection
logs'
ids (markers).
(other rules are very similar in structure, but different in logic, of
course)
This is its programmatic drl file:
package com.checkpoint.correlation.impl.drools.package30;
import java.util.Date
import java.util.HashMap
import java.util.Set
import com.checkpoint.correlation.impl.drools.Log
import com.checkpoint.correlation.impl.drools.CorrelatedEvent
global com.checkpoint.correlation.server.EventsHandler
externalEventsHandler;
import function
com.checkpoint.correlation.impl.utils.UserDefinedFunctions.isInDayHour
Range
import function
com.checkpoint.correlation.impl.utils.UserDefinedFunctions.isInIpRange
function boolean filter(Log log) {
return (!((log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "10.80.0.0",
"10.80.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "124.0.0.0",
"124.255.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "192.168.0.0",
"192.168.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "195.158.7.0",
"195.158.7.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "11.25.0.0",
"11.25.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "128.157.0.0",
"128.157.255.255")) || (log.fieldsMap.get("src")!= null &&
isInIpRange(log.fieldsMap.get("src").toString(), "213.114.0.0",
"213.114.255.255")))); }
function String markersToString(Set markersSet) {
int i = 0;
String markersString = "";
for (Object marker : markersSet) {
if (i == 25) break;
String markerStr = marker.toString();
if (i > 0) markersString += "\n";
markersString += markerStr;
}
return markersString;
}
function String calcSeverity(Log log) {
return "High";
}
function String getUniqueId(Log log) {
String uniqueId="";
uniqueId += (log.fieldsMap.get("service") != null ?
log.fieldsMap.get("service").toString() : "null");
uniqueId += (log.fieldsMap.get("proto") != null ?
log.fieldsMap.get("proto").toString() : "null");
return uniqueId;
}
&n! bsp;
declare Log
@role(event)
end
declare CorrelatedEvent
@role(event)
@expires(600s)
end
rule "Port scan from external network"
enabled true
dialect "java"
no-loop
when
$log : Log(eval(filter($log)))
not CorrelatedEvent(getId() ==
"{8AC52BA8-1EE8-4f18-9BB4-54492116501C}", groupByFieldsMap.get("src")
== $log.fieldsMap.get("src"), groupByFieldsMap.get("dst") ==
$log.fieldsMap.get("dst"))
accumulate($accumulatedLog :
Log(eval(filter($accumulatedLog)), this after[0s,60s] $log,
fieldsMap.get("src") == $log.fieldsMap.get("src"),
fieldsMap.get("dst") == $log.fieldsMap.get("dst"), $id :
getUniqueId(this));
$idSet : collectSet($id);
$idSet.size > 19)
accumulate($accumulatedLog :
Log(eval(filter($accumulatedLog)), this after[0s,60s] $log,
fieldsMap.get("src") == $log.fieldsMap.get("src"),
fieldsMap.get("dst") == $log.fieldsMap.get("dst"),
$idSet.contains(getUniqueId(this)), $marker :
fieldsMap.get("marker"));
$markerSet : collectSet($marker))
then
CorrelatedEvent $ce = new
CorrelatedEvent("{8AC52BA8-1EE8-4f18-9BB4-54492116501C}");
$ce.groupByFieldsMap.put("src",
$log.fieldsMap.get("src"));
$ce.groupByFieldsMap.put("dst",
$log.fieldsMap.get("dst"));
insert($ce);
HashMap<String,Object> fieldsMap = new
HashMap<String,Object>();
fieldsMap.put("cu_rule_id",
"{8AC52BA8-1EE8-4f18-9BB4-54492116501C}");
fieldsMap.put("event_name", "Port scan from external
network");
fieldsMap.put("cu_rule_severity", calcSeverity($log));
fieldsMap.put("cu_rule_category", "Scans");
fieldsMap.put("cu_log_count", $markerSet.size());
fieldsMap.put("time", new Date());
fieldsMap.put("cu_markers_list",
markersToString($markerSet));
fieldsMap.put("src", $log.fieldsMap.get("src"));
fieldsMap.put("src_machine_name",
$log.fieldsMap.get("src_machine_name"));
fieldsMap.put("src_user_name",
$log.fieldsMap.get("src_user_name"));
fieldsMap.put("dst", $log.fieldsMap.get("dst"));
fieldsMap.put("dst_machine_name",
$log.fieldsMap.get("dst_machine_name"));
fieldsMap.put("dst_user_name",
$log.fieldsMap.get("dst_user_name"));
fieldsMap.put("service",
$log.fieldsMap.get("service"));
fieldsMap.put("proto", $log.fieldsMap.get("proto"));
fieldsMap.put("product",
$log.fieldsMap.get("product"));
externalEventsHandler.handleEvent(fieldsMap);
end
I am sending logs in a rate of up to 200 logs/sec. After about 3
minutes, my application starts to be unresponsive.
I monitored the JVM with VisualVM. Two snapshots of VisualVM are attached.
I found out that the class consuming most memory is FromNodeLeftTuple
of drools (as can be seen in "instances.png").
1) Is my inserting rate is too high?
2) Is There a way I can make my rules more memory efficient?
Thanks.
Inserting logs:
public void insertEvents(Collection<Map<String, Object>> logs) {
for (Map<String, Object> map : logs) {
Log log = new Log();
Log.fieldsMap.putAll(map);
session.insert(log);
session.fireAllRules();
}
}
Log class:
public class Log
{
public HashMap<String, Object> fieldsMap = new HashMap<>(); }
CorrelatedEvent class:
public class CorrelatedEvent
{
public Map<String, Object> groupByFieldsMap;
private String id;
public CorrelatedEvent(String id)
{
groupByFieldsMap = new HashMap<>();
this.id<http://this.id> = id;
}
public String getId()
{
return id;
}
}
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org<mailto:rules-users@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/rules-users
Email secured by Check Point
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org<mailto:rules-users@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org<mailto:rules-users@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/rules-users
Email secured by Check Point
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org