Retrieving several fields from Canopy

[newkirk]

I'm responsible for setting up monitoring and management for a small but rapidly growing base of Motorola Canopy wireless units. There's certain critical data that we need to monitor, either in immediacy or observing long-term trends, or in a few cases both. I MUST be able to set up where someone would be able to enter the customer name, the IP address, and select "Canopy Modem" host template, then select all preset templated graphs "on", and be done.

All the info we need to monitor is available via SNMP, although broken down among 2-3 separate OID branches. I've spent about 50 hours the past 4 days trying to get it working, through 4 different versions of Cacti (a mandrake rpm feed 8.5, 8.6b, and two CVS's, latest Nov25 00:31), SNMP queries, SNMP indexed, Script indexed, etc. (OK, to be fair I've worked on other things as well, like monitoring the nodes in a mail cluster and finding where to customize colors to match our other webadmin interfaces ;^)

With each cacti version I've deleted the mysql databases and started from scratch, fixing bugs ("ERROR 1136 at line 1471: Column count doesn't match value count at row 1" and others related to table expansions, in latest CVS "cacti.sql" for example) where found.

Now I'm about to be locked up and sedated, and I'm going to be working on this straight through Thanksgiving until I get it - my ass (and sanity) are on the line to have it working before the weekend.

To create this all as a Host Template it seems I then need Graph Templates and Data Queries - is this correct? If so, then it appears that Data Queries only offers Indexed versions, while what I need to work with offers no meaningful index. (What do I index when I'm reading 7 single values for a single device??)


How, either SNMP or script, do I successfully get the '*'ed data below into a single graph in Cacti, with 'rssi' (and possibly 'jitter') graphed, and other values as captions???
snmpwalk -m all -v 2c -c public 172.24.1.1 .1.3.6.1.4.1.161.19.3.2.2
sessionStatus.0 REGISTERED
*rssi.0 1516
*jitter.0 3
*airDelay.0 35
*radioSlicing.0 568
*radioTxGain.0 15
calibrationStatus.0 Power Level, RSSI and Slicing Value are calibrated
*radioDbm.0 -68 dBM
*registeredToAp.0 0a-00-3e-90-02-04
dhcpCip.0 0.0.0.0
dhcpSip.0 0.0.0.0
dhcpClientLease.0 0
dhcpCSMask.0 0.0.0.0
dhcpDfltRterIP.0 0.0.0.0
dhcpcdns1.0 0.0.0.0
dhcpcdns2.0 0.0.0.0
dhcpcdns3.0 0.0.0.0
dhcpDomName.0

How do I handle this?

I've written a few different perl scripts that extract this data via SNMP and massage it a bit (like supplying IP instead of MAC for 'registeredToAp' - BUT how do I GPRINT a string instead of a number?? I took care of printing INTs, but not IPs...) but I don't understand exactly what they are supposed to return to Cacti when requested for an index, or how to set this all up with templates WITHOUT 'indexed' queries - I presume that once I've resolved that, I'll be able to ensure that my query/get support works... :^(

I understand the XML syntax for the script or SNMP queries, with the exception of indexing and what (if anything) of my data should be 'input'.

I've been successful if I hand-build the graph, and define single OID datasources 'on-the-fly', but not as a template.

Anybody who can offer any advice or answers, please help.

j


Joel Newkirk

[rony]

I must reply to you using CVS.

CVS is currently in an Alpha state and changing daily. Bugs are everywhere, we are working on the next version.

Please use 0.8.6b and all the patches.

You can download 0.8.6b from http://www.cacti.net/downloads/cacti-0.8.6b.tar.gz and the patches are located in http://www.cacti.net/downloads/patches/0.8.6b/.


Concerning your other questions,

I assume you are monitoring the customer units, or long hauls. Because I can see the indexes changing if the units are rebooted and a customer unit reconnectes to a the AP cluster, shoot it may even move to a different AP. But from your example, I would point you at the graph templates: "Host MIB - Logged in Users" and "Host MIB - Processes". These are good examples of using a single OID and graphing it. Then you can place these graph templates in the host template for the canopy units.

You can place text on the graph by using a graph item of COMMENT, but to automatically do this, that I don't know.

This is open source software, modify it as you need..

Good luck.

[newkirk]

I must reply to you using CVS.

CVS is currently in an Alpha state and changing daily. Bugs are everywhere, we are working on the next version.

Please use 0.8.6b and all the patches.

Thanks - after examining the two CVS versions I retrieved, I reverted back to (now patched ) 0.8.6b version. (BTW, I liked what I saw in CVS)
..

I assume you are monitoring the customer units, or long hauls. Because I can see the indexes changing if the units are rebooted and a customer unit reconnectes to a the AP cluster, shoot it may even move to a different AP. But from your example, I would point you at the graph templates: "Host MIB - Logged in Users" and "Host MIB - Processes". These are good examples of using a single OID and graphing it. Then you can place these graph templates in the host template for the canopy unit

Indeed, one of the things we are trying to track is the tendency of some customer's units to change APs. However, the graph I'm trying to achieve would NOT be a single OID, but rather something like the attached pic - RSSI level graphed, other data presented below in captions.

The problems I'm running into with the graph:

Is there any way to retrieve a string value via SNMP and present it in the graph??

How can I strip the trailing "dBM" off the SNMP value "-47 dBM" and store the "-47" (or just '47' if necessary) in a RRD?

j

[TheWitness]

In today's environment, you would write a script to get the OID value and then trim it to the appropriate level and return to Cacti. In the future, we might add some feature that strips the text inside of Cacti's poller process. Right now, non-numeric data doesn't pass our pre-checks.

Something to think about though.

TheWitness

[Curtis Doty]

I manage similar metrics on different wireless hardware. The script solution is best because it allows you to calculate more valuable stuff such as S/N rather than just RSSI. Also the net-snmp -OU option will strip off the dBm or watever units are in your MIB.

[Shaman]

Do you have any scripts, etc. that you would be willing to share with the rest of us Cacti users? I'm just now setting up some monitoring for our Motorola units.