02-18-2010 01:22 PM
I loaded MIB's for the device but I do not see any components being discoverd when I run a configuration poll.
Do I need to recycle the NNMi processes?
I tried nnmloadmib -sync and that doesn't seem to help
02-22-2010 04:18 AM
02-23-2010 02:32 AM
In the NNM Configuration workspace select Monitoring Configuration and examine the Device settings.
Note, if it is greyed out CH polling only becomes available when you have installed a Performance SPI license. An instant on license is automatically installed if the SPI is on the NNM server, if it is remote you need to go to your NNM server and run nnmenableperfspi.ovpl
02-23-2010 02:41 AM
the Component Health is an out of the box feature which does not need to load mibs. But it uses the vendor specific mibs from Cisco and Nortel and the Host Resource MIB. Please have a look in the online help (search for: "Node Capabilities Provided by NNMi").
The Custom Poller is used to monitor (other) SNMP values. To use it you have to load the mib(s) and you must set up a Custom Poller collection (there is a step-by-step white paper on SSO manuals). With the iSPI Performance for Metrics (8.13) you are also able to store the collected values and to generate reports about that data ( whithout the iSPI no data collection and no reporting - only thresholding).
02-23-2010 07:20 AM
How do I turn off a capability? For example, we have Juniper EX network switches which run a form of Linux. The Host Resources MIB capability has been applied to these devices and causing us to receive false memory alarms. It is alarming on the procfs file system being full but in reality, this means nothing to the overall component health of the device. I've unloaded the HOST-RESOURCES-MIB, performed a resync, and cleared all previous alarms; however, during the next poll they all came back again. Any advice is greatly appreciated. Thanks.
02-23-2010 07:23 AM
Then click on "Monitoring"
Create a new node setting specific for your node group.
Uncheck the "Enable component health"
This should exclude all devices in that group from component health.
02-23-2010 07:29 AM
Somehow NNM is aware that certain devices like Cisco should automatically have their CPU/Power Supply monitored from discovery.
I wanted to understand how this is automatically done for certain devices and append to it new device types.
02-23-2010 07:33 AM
02-23-2010 09:13 AM
You could try reducing (or increasing depending on the sense of the metric) the thresholds for particular devices (or groups) within NNM. That would avoid thresholds being breached.
02-23-2010 09:56 AM
Yes, I am referring to threshold breaches that trigger the "MemoryOutOfRangeOrMalfunctioning" incident. The issue is that these are being triggered based on the procfs file system which is a pseudo-file system on most Unix-variant OSs that is allocated very little memory and storage space but utilizes it fully. For example, on these switches, procfs is always allocated 4k and when the switch is up and working it utilizes the full 4k (100%). It's similar to monitoring a CD drive when a CD is loaded - it will always show 100% full but it has no bearing on performance; therefore it is pointless to monitor.
It is definitely coming from the HOST-RESOURCES capability and I'd like to figure out how I can turn this capability off. Unloading the MIB has not worked. Any ideas?
02-23-2010 11:02 AM
Can you not set thresholds for these devices for fault too?
I shall ask someone more expert in NNM.
08-03-2010 03:38 AM
I have an issue related to component health. There is a Juniper (firewall) device being monitored. For this device, component health polling is not happening. In the configuration poll, it says "Get Memory (CPU & Buffer) retrieved no data because this device does not support a way of collecting data". So in the reporting menu, "Component Health" tab shows no data. But, Interface health data is getting polled.
My question is, do I need to load MIB for component health for Juniper? If yes, which MIB i should load? If no, Why it is giving the error?
08-03-2010 03:44 AM
I think the required MIBs are automatically loaded by NNM.
If the device itself does not support the required MIB (or sometimes OIDs within that MIB) then there is simply no data to retrieve, nothing you do in NNM will fix that.
I think I have some information on the devices and MIBs supported, I'll try and dig it out and post it.
05-11-2012 01:35 AM
Prompted by an off-board message I did ask an NNM expert and this is what they said:
My recommendation would be to un-manage those memory sensors by selecting the node sensor in NNMi and the either setting the management mode to "Not Managed" or run the management mode action. The other possibility is to setup node groups for the offending nodes and then to disable node component monitoring but this would disable monitoring of all node components for those nodes which the customer probably does not want to do.