[NMS-7839] Counter variables reported as strings (like Net-SNMP extent) are not stored properly when using RRDtool - The OpenNMS Issue Tracker

Details

Type: Bug
Status: Resolved (View Workflow)
Priority: Blocker
Resolution: Fixed
Affects Version/s: 14.0.3, Meridian-2015.1.0, 15.0.2, 16.0.2
Fix Version/s: 16.0.3, Meridian-2015.1.1, 17.0.0
Component/s: Data Output - RRD
Security Level: Default (Default Security Scheme)
Labels:
- support

Description

Here is how to reproduce the problem.

On my CentOS 6.7 VM running Meridian 2015.1.0, I configured the local snmpd.conf with this:

/etc/snmp/snmpd.conf

extend sample /etc/snmp/counter.pl

The script that simulates the counter is very simple:

/etc/snmp/counter.pl

#!/usr/bin/perl

use strict;

my $data = 0;
my $source_file = "/tmp/.counter.data";

if (-e $source_file) {
   open READ, $source_file;
   $data = <READ>;
   close READ;
}

$data += int(rand(100));
print $data, "\n";

open WRITE, ">$source_file" or die "Can't write data on $source_file\n";
print WRITE $data;
close WRITE;

If you execute the script it will always return an increasing number:

[agalue@centos6srv ~]$  /etc/snmp/counter.pl 
403
[agalue@centos6srv ~]$  /etc/snmp/counter.pl 
480
[agalue@centos6srv ~]$  /etc/snmp/counter.pl 
562

In OpenNMS, I have a file called /opt/opennms/etc/datacollection/sample.xml with the following content:

/opt/opennms/etc/datacollection/sample.xml

<datacollection-group name="Sample">
 <group name="sample" ifType="ignore">
   <mibObj oid=".1.3.6.1.4.1.8072.1.3.2.4.1.2.6.115.97.109.112.108.101" instance="1" alias="sample" type="counter"/>
 </group>
 <systemDef name="Net-SNMP Counter Sample">
   <sysoidMask>.1.3.6.1.4.1.8072.3.2.</sysoidMask>
   <collect>
     <includeGroup>sample</includeGroup>
   </collect>
 </systemDef>
</datacollection-group>

Of course, I've added a reference to it on datacollection-config.xml:

[agalue@centos6srv ~]$ grep Sample /opt/opennms/etc/datacollection-config.xml 
     <include-collection dataCollectionGroup="Sample"/>

Now, let's try the OID using snmp-request (I use snmp-request, a clone of Net-SNMP's tools based on SNMP4J, because the snmpwalk command is not as good as SNMP4J in terms of the SNMP Protocol implementation):

[agalue@centos6srv ~]$ /opt/opennms/bin/snmp-request -v 2c -c public -Ow localhost .1.3.6.1.4.1.8072.1.3.2 | grep 6.115.97.109.112.108.101
SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder".
SLF4J: Defaulting to no-operation (NOP) logger implementation
SLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details.
1.3.6.1.4.1.8072.1.3.2.2.1.2.6.115.97.109.112.108.101 = /etc/snmp/counter.pl
1.3.6.1.4.1.8072.1.3.2.2.1.3.6.115.97.109.112.108.101 = 
1.3.6.1.4.1.8072.1.3.2.2.1.4.6.115.97.109.112.108.101 = 
1.3.6.1.4.1.8072.1.3.2.2.1.5.6.115.97.109.112.108.101 = 5
1.3.6.1.4.1.8072.1.3.2.2.1.6.6.115.97.109.112.108.101 = 1
1.3.6.1.4.1.8072.1.3.2.2.1.7.6.115.97.109.112.108.101 = 1
1.3.6.1.4.1.8072.1.3.2.2.1.20.6.115.97.109.112.108.101 = 4
1.3.6.1.4.1.8072.1.3.2.2.1.21.6.115.97.109.112.108.101 = 1
1.3.6.1.4.1.8072.1.3.2.3.1.1.6.115.97.109.112.108.101 = 1102
1.3.6.1.4.1.8072.1.3.2.3.1.2.6.115.97.109.112.108.101 = 1102
1.3.6.1.4.1.8072.1.3.2.3.1.3.6.115.97.109.112.108.101 = 1
1.3.6.1.4.1.8072.1.3.2.3.1.4.6.115.97.109.112.108.101 = 0
1.3.6.1.4.1.8072.1.3.2.4.1.2.6.115.97.109.112.108.101.1 = 1102

The last entry is what we have on the datacollection-group.

Now, let's see how Collectd is handling that data:

[agalue@centos6srv logs]$ grep "Visiting attribute.*sample" collectd.log 
2015-08-13 11:14:51,886 DEBUG [Collectd-Thread-2-of-50] o.o.n.c.s.AbstractCollectionAttribute: Visiting attribute node[1].sample [.1.3.6.1.4.1.8072.1.3.2.4.1.2.6.115.97.109.112.108.101] = 1559
2015-08-13 11:15:23,005 DEBUG [Collectd-Thread-3-of-50] o.o.n.c.s.AbstractCollectionAttribute: Visiting attribute node[1].sample [.1.3.6.1.4.1.8072.1.3.2.4.1.2.6.115.97.109.112.108.101] = 1625
2015-08-13 11:15:54,132 DEBUG [Collectd-Thread-4-of-50] o.o.n.c.s.AbstractCollectionAttribute: Visiting attribute node[1].sample [.1.3.6.1.4.1.8072.1.3.2.4.1.2.6.115.97.109.112.108.101] = 1654
2015-08-13 11:16:24,396 DEBUG [Collectd-Thread-5-of-50] o.o.n.c.s.AbstractCollectionAttribute: Visiting attribute node[1].sample [.1.3.6.1.4.1.8072.1.3.2.4.1.2.6.115.97.109.112.108.101] = 1705

[agalue@centos6srv logs]$ grep "updating RRD.*sample" collectd.log 
2015-08-13 11:14:51,932 INFO  [Collectd-Thread-2-of-50] o.o.n.r.RrdUtils: updateRRD: updating RRD file /opt/opennms/share/rrd/snmp/1/sample.rrd with values '1439478892:1559.0'
2015-08-13 11:15:23,006 INFO  [Collectd-Thread-3-of-50] o.o.n.r.RrdUtils: updateRRD: updating RRD file /opt/opennms/share/rrd/snmp/1/sample.rrd with values '1439478923:1625.0'
2015-08-13 11:15:54,132 INFO  [Collectd-Thread-4-of-50] o.o.n.r.RrdUtils: updateRRD: updating RRD file /opt/opennms/share/rrd/snmp/1/sample.rrd with values '1439478954:1654.0'
2015-08-13 11:16:24,396 INFO  [Collectd-Thread-5-of-50] o.o.n.r.RrdUtils: updateRRD: updating RRD file /opt/opennms/share/rrd/snmp/1/sample.rrd with values '1439478984:1705.0'

As you can see, the values are stored like float numbers (i.e. numbers with decimals), which is a non valid operation for counters on RRDtool.

Of course, the values are not being stored on the RRDtool file:

[agalue@centos6srv logs]$ /opt/opennms/bin/rrdtool dump /opt/opennms/share/rrd/snmp/1/sample.rrd | grep "2015-08-13 11:1[456]"
			<!-- 2015-08-13 11:14:00 EDT / 1439478840 --> <row><v>NaN</v></row>
			<!-- 2015-08-13 11:14:30 EDT / 1439478870 --> <row><v>NaN</v></row>
			<!-- 2015-08-13 11:15:00 EDT / 1439478900 --> <row><v>NaN</v></row>
			<!-- 2015-08-13 11:15:30 EDT / 1439478930 --> <row><v>NaN</v></row>
			<!-- 2015-08-13 11:16:00 EDT / 1439478960 --> <row><v>NaN</v></row>
			<!-- 2015-08-13 11:16:30 EDT / 1439478990 --> <row><v>NaN</v></row>

Seeing the source code, I found the following:

org.opennms.netmgt.collectd.SnmpAttribute

    @Override
    public String getNumericValue() {
        if (getValue() == null) {
            LOG.debug("No data collected for attribute {}. Skipping", this);
            return null;
        } else if (getValue().isNumeric()) {
            return Long.toString(getValue().toLong());
        } else {
            // Check to see if this is a 63-bit counter packed into an octetstring
            Long value = SnmpUtils.getProtoCounter63Value(getValue());
            if (value != null) {
                return value.toString();
            }

            try {
                return Double.valueOf(getValue().toString()).toString();
            } catch(NumberFormatException e) {
                LOG.trace("Unable to process data received for attribute {} maybe this is not a number? See bug 1473 for more information. Skipping.", this);
                if (getValue().getType() == SnmpValue.SNMP_OCTET_STRING) {
                    try {
                        return Long.valueOf(getValue().toHexString(), 16).toString();
                    } catch(NumberFormatException ex) {
                        LOG.trace("Unable to process data received for attribute {} maybe this is not a number? See bug 1473 for more information. Skipping.", this);
                    }
                }
            }
            return null;
        }
    }

The key element here is how isNumeric is implemented:

org.opennms.netmgt.snmp.snmp4j.Snmp4JValue

    public boolean isNumeric() {
        switch (m_value.getSyntax()) {
        case SMIConstants.SYNTAX_INTEGER:
        case SMIConstants.SYNTAX_COUNTER32:
        case SMIConstants.SYNTAX_COUNTER64:
        case SMIConstants.SYNTAX_TIMETICKS:
        case SMIConstants.SYNTAX_UNSIGNED_INTEGER32:
            return true;
        default:
            return false;
        }
    }

Because the extent feature of Net-SNMP returns the value as string, isNumeric will return false, so it will converted into a double, which is not correct for these case.

Attachments

Activity

People

Assignee:

Alejandro Galue

Reporter:

Alejandro Galue

Votes:

0 Vote for this issue

Watchers:

1 Start watching this issue

Dates

Created:

13/Aug/15 11:42 AM

Updated:

13/Aug/15 12:09 PM

Resolved:

13/Aug/15 12:09 PM