Spike detection and JVM issues

lrsmith · Joined: 29 Jun 2006 Posts: 16

I've run into an problem with using Cacti to monitor a JVM that has SNMP enabled. I created a device that connects to the SNMP port of the JVM instance and query fine. However only the JVM MIB is available on that port so Cacti is not able to query the .1.3.6.1.2.1.1.3.0 MIB to see how long the host has been up.

As a work around I go into the MySQL database and modify the poller_reindex table to monitor the .1.3.6.1.4.1.42.2.145.3.163.1.1.4.11.0 MIB which is the jvmRTUptimeMs counter.

This works fine as long as the device configuration is not changed, because it reverts back to the 1st mib and I have to go into the MySQL table and change it again. I'm not sure what else would cause this yet.

It would be nice if we could specify the MIB, on a per device basis, that should be checked for uptime and spike detection.

Other than that Cacti has been working well and its a great tool/project.

Thanks,
Len

lrsmith · Joined: 29 Jun 2006 Posts: 16

I have been tracing this through the code and in poller.php the problem seems to come down to the switch statement that starts at line 150. The reindex method is backwards_uptime, which is hard-coded to reset the OID to .1.3.6.1.2.1.1.3.0 rather than use the existing OID.

It would be nice to have that switch to use a variable that can be redefined from the default, on a per device basis.

TheWitness · Posted: Mon Sep 25, 2006 7:40 pm Post subject:

lrsmith,

Please goto bugs.cacti.net and log your feature request. I believe that this is a "worthy" feature.

TheWitness

lrsmith · Joined: 29 Jun 2006 Posts: 16

http://bugs.cacti.net/view.php?id=829

As a work around I've set the reindex method to none and that seems to work.

TheWitness · Posted: Sun Dec 31, 2006 2:03 pm Post subject:

What happens with this host when you perform an snmpgetnext as follows:

snmpgetnext -c <your_community> -v <your_version> <your_host> .1

TheWitness

TheWitness · Posted: Sat Sep 15, 2007 11:58 pm Post subject:

In Cacti 0.8.7, you will be able to add the following tag to your XML file "<oid_uptime>" so an OID that supports your uptime counter therefore, resolving this issue.

TheWitness

gandalf · Posted: Sun Sep 16, 2007 2:19 pm Post subject:

Please Larry,
let me know more about this. I was not aware of this feature. How will I document it???
Reinhard

TheWitness · Posted: Sun Sep 16, 2007 2:23 pm Post subject:

Reinhard,

I just sent you a note relative to this. Simply add the following to your XML file:

<oid_uptime>.a.b.c.d.e.f</oid_uptime>

Then reindex your query.

Larry