The reason I ask the above question is due to the following:

"net-snmp does not deal with terabyte file systems well if they are NFS mounted - they will have small hrStorageAllocationUnits, and hrStorageSize and hrStorageUsed will wrap, so you may see that the Used space is greater than the total Size.

There is currently no fix for this in the Net-snmp snmp agent, so the workaround is to monitor the NFS server, not the system that mounts them."

Ryan, I have never heard that about snmp v2 and big filesystems, have you seen this yourself? I cant wait to try it tomorrow at work.

Some version of net-snmp roll and show negative and some don't roll and say something stupid Surprisingly the showing a negative can be addresses programmatically but the stupid response cant.

I do know I have read the most recent version of net-snmp ( claims ) to have fixed the issue of monitoring filesystems greater than 2TB, I say claimed because I have not seen it myself. I so far have had to address it via other ways.

I encountered the same problem. SNMP just doesn't report them correctly. You need to use zenoss.cmd.df as your filesystem modeler plugin. It's been some time since I set this up but from memory -

- install zenplugin.py in /opt/zenoss/libexec on the target host

- I created a new device class for these devices, edited the modeler plugins to remove the snmp.filesystemmap and added zenoss.cmd.df

- edited Configuration Properties for the class and set zCommandUsername and zCommandPassword for a login account on that box; set zCommandPath to be '/opt/zenoss/libexec'

- added and modeled the device.

I _think_ that was it. Googling for zenoss.cmd.df might help more if you get stuck.

Then I would be most interested in knowing the proper way to monitor large file systems too. The standard SNMP based collector does not work out of the box in 3.2 (patched).

Marc

Do you have the following ZenPacks installed?

docs/DOC-3435

docs/DOC-3493

From your description it sounds like you don't have the proper FileSystem template in the class where the device is.  There's a specific template for collecting the data via SSH (which should be included with the above ZenPacks).

Devices where the filesystem is being monitored via SSH need to be in a class with that template (as opposed to the standard SNMP one).

This works for me.  You need to have both the command-based modeler plugin and the command-based Filesystem template active.  After that, it takes a couple of poll intervals before you see the used / free stats filled in but you should certainly get there.

Try running zenmodeler (as the zenoss user) just for a relevant device for the df plugin.  For a device called fred, that would be:

zenmodeler run -v 10 -d fred --collect zenoss.cmd.df

You should see exactly what output is reported that way.

Cheers,

Jane