To be successful with modelling a device you need a few things before you start.
| Table of Contents |
|---|
Once you have this in hand checkout the more detailed device modelling documentation here: Developing Device Models for NMIS
Device Modelling Preparation
Device Access
...
For devices with proprietary MIB's or Enterprise MIBS, you should obtain them from the vendor, Google is very helpful and add them to your MIB's in ~/mibs, before doing the SNMP walk.
Commands to run:
.
For a large collection of MIB files which have been error corrected, you can find many MIBS on Keith Sinclair's GitHub MIB Repo
Organise your MIBS.
If you do not have any vendor MIBS, download the MIBS from the GitHub Repo
| Code Block |
|---|
cd ~
wget https://github.com/kcsinclair/mibs/archive/master.zip
unzip master.zip
mv mibs-master mibs |
If you have MIBS from your vendor
| Code Block |
|---|
mkdir ~/mibs
cp <vendor mib file(s)> ~/mibs
cp /usr/local/nmis8/mibs/traps/* ~/mibs
snmpwalk - |
Test your SNMP WALK and MIBS
| Code Block |
|---|
snmpwalk -m ALL -M ~/mibs -v 2c -c GOODCOMMUNITY <HOSTNAME or IP ADDRESS> system |
Does SNMP Bulk Walk work?
| Code Block |
|---|
snmpbulkwalk -m ALL -M ~/mibs -v 2c -c GOODCOMMUNITY <HOSTNAME or IP ADDRESS> system |
...
Run a full walk and redirect to a file, note that you can use snmpbulkwalk on many devices when using SNMPv2C, this greatly improves the speed of collection on very large devices.
| Code Block |
|---|
snmpwalk -m ALL -M ~/mibs -v 2c -c GOODCOMMUNITY <HOSTNAME or IP ADDRESS> .1 > ~/VENDOR-Product.mib or snmpbulkwalk -m ALL -M ~/mibs -v 2c -c GOODCOMMUNITY <HOSTNAME or IP ADDRESS> .1 > ~/VENDOR-Product.mib |
Now you have an SNMP Dump.
SNMP WALK with Numerical OIDS
If you are seeking support for device modelling, getting the SNMP walk results as numerical OID's only is very useful, as this can be used in an SNMP simulator. There is a translate tool available to map the OIDs we well, https://github.com/kcsinclair/mibs/blob/master/translate_snmpwalk.pl
| Code Block |
|---|
snmpwalk -v2c -c COMMUNITY String -ObentU IPADDRESS_OR_NODENAME 1.3.6 > myagent.snmpwalk |
Goal for Modelling
What is the goal for the modelling? Just standard type support, or more advanced collection, do you want to collect some performance data about how a protocol is operating, or verify the number of sessions a firewall is running.
...
This page has great information about Modelling MIBS that use Indexes using the systemHealth section.
Developing Device Models for NMIS
Checkout the handy training we developed for NMIS users, this is available here:
Checklist for Model Development
The training covers this in details, but here is a quick summary of what needs to be updated and changed when adding new device support with modelling
- Update nmis8/models/Model.nmis
- New model file created and updated
- update the nodeVendor, nodeModel, nodeType fields (among others)
- Any new resolvable OID's added to nmis_oids.nmis, e.g. the sysObjectID names.
- Update Common-heading.nmis
- Update Common-database.nmis
- Create any needed graphs
- Update Common-threshold.nmis
- Update Common-stats.nmis
Creating Thresholds with NMIS Modelling
Main collection is /systemHealth/rrd/hrProcessorLoad
Common-database entry called hrProcessorLoad
Common-heading entry for the /systemHealth/rrd/hrProcessorLoad/graphtype => hrprocload
To create a threshold, statistical data is needed, this is extracted using Common-stats and this is based on the rrd name, so for this hrProcessorLoad collection, the Common-stats entry would be hrProcessorLoad
To compare the data to the threshold, there needs to be a Common-threshold section, this is named after the /systemHealth/rrd/hrProcessorLoad/threshold => arCpuLoad
So that the threshold system can match data from the stats results, we use the item to match the data, so NMIS is going to look for the item in the resulting stats data, in this case the item name is hrProcessorLoad, which matches the print statement: 'PRINT:hrProcessorLoad:AVERAGE:hrProcessorLoad=%1.0f' in the Common-stats section.
...
.
...