вторник, 18 сентября 2012 г.

Temperature monitoring with OpenNMS and brocade FC switch

We are expanding! And we have to monitor temperature level in our new data center. So, we have:
  1. existing OpenNMS monitoring system,
  2. Brocade 300 SAN FC switch (or something very similar).
Our task is the same: to monitor temperature and send SMS when it's too high. Later part was already discussed.
The above mentioned FC switch has three temperature sensors, which data is available by SNMP (oids .1.3.6.1.4.1.1588.2.1.1.1.1.22.1.4.1 - .1.3.6.1.4.1.1588.2.1.1.1.1.22.1.3). The only problem is that OpenNMS doesn't know about them. Let's teach it!
At first, we have to modify datacollection-config.xml and add to "groups" section:
<group  name="brocade-temperature" ifType="ignore">
         <mibObj oid=".1.3.6.1.4.1.1588.2.1.1.1.1.22.1.4" instance="1" alias="temperature1" type="integer" />
         <mibObj oid=".1.3.6.1.4.1.1588.2.1.1.1.1.22.1.4" instance="2" alias="temperature2" type="integer" />
         <mibObj oid=".1.3.6.1.4.1.1588.2.1.1.1.1.22.1.4" instance="3" alias="temperature3" type="integer" />
      </group>
I know, "temperature[1-3]" names are ugly, but I didn't care at first and now when the whole thing is working I don't want to change anything :)
And modify "Brocade FC Switches" systemDef in systems section:
     <systemDef name="Brocade FC Switches">
        <sysoidMask>.1.3.6.1.4.1.1588.</sysoidMask>
        <collect>
          <includeGroup>brocade-temperature</includeGroup>
          <includeGroup>brocade-switch-fctable</includeGroup>
        </collect>
     </systemDef>
Now we are ready and can define thresholds. To be sure that requested parameters are gathered you can look at rrdRepository. Path to the repository is specfied in rrdRepository property of datacollection-config resource (look at top of datacollection-config.xml). If you have temperature[1-3].jrb files in ${rrdRepository}/${NodeID} and these files contains some data, everything is OK. NodeId is displayed in URL, when you look at node (/opennms/node.jsp?node=353) or you can query "node" table in OpenNMS database:
SELECT nodeid from node where nodelabel='your label';

To be sure that temperature[1-3].jrb contains correct data, you can dump it (your version of jrobin-*.jar file may be different):
# cd ${rrdRepository}/${NodeID}
# echo "dump temperature1.jrb" |  java -jar ${OPENNMS_HOME}/opennms/lib/jrobin-1.5.12.jar
However, OpenNMS still doesn't know how to display this data. Let's help it. To define new graph we will add the following parts to snmp-graph.properties. In the begining of file we have to add reports names to reports definition:
reports=mib2.HCbits, mib2.bits, mib2.percentdiscards, mib2.percenterrors, \
...
brocade.switch.temperature1, brocade.switch.temperature2, brocade.switch.temperature3, \
...
and define reports later (here only one report is shown, other two are essentially the same, just change [Tt]emperature1 to [Tt]emperature[23]):

report.brocade.switch.temperature1.name=Brocade switch temperature1
report.brocade.switch.temperature1.columns=temperature1
report.brocade.switch.temperature1.type=nodeSnmp
report.brocade.switch.temperature1.command=--title="Brocade switch temperature1" \
 --vertical-label="degrees celsius" \
 DEF:temperature1={rrd1}:temperature1:AVERAGE \
 AREA:temperature1#0000ff:"Temperature1" \
 GPRINT:temperature1:AVERAGE:" Avg \\: %8.2lf %s" \
 GPRINT:temperature1:MIN:"Min \\: %8.2lf %s" \
 GPRINT:temperature1:MAX:"Max \\: %8.2lf %s\\n" 
.. 
Now after "service opennms restart" we'll get pretty graphs if we build graph based on "Node-level Performance Data" resources for brocade node.

P.S. And lastly I just must give you a link to a good document concerning data collection configuration in OpenNMS and another one concerning SNMP configuration.