[solved] Sporadic "Get Mounted Partitions" data in

[jberanek]

Hello,

I've had a Cacti installation for many years, but have recently seen a strange occurrence on a few graphs. Some introductory info:

Cacti 0.9.7d
CentOS 5.5
PHP 5.1.6
NET-SNMP 5.3.2.2

I have some Windows hosts configured, included in this configuration is the "SNMP - Get Mounted Partitions" data query. This generally works fine, and you can see all the systems drives, and create graphs for them.

However, on some of these Windows systems, the "used space" value for (seemingly) just the C: drive shows incorrect values. The systems in question use as an example 8GB on the C: drive, but frequently show either 0GB used, or a few hundred MB.

I've dug into how the script (ss_host_disk.php) works a bit, and found it uses the SNMP OID tree .1.3.6.1.2.1.25.2.3.1 (HOST-RESOURCES-MIB). When the data is showing the incorrect value in Cacti, I can run an snmpwalk and see the correct data there.

So, seemingly a bug in the script server script? How can I go about debugging the problem? I tried to follow debugging guidance I found at http://docs.cacti.net/manual:087:3a_adv ... ipt_server but I couldn't get it to work. Here's an attempted run (<cacti_dir> is obviously expanded in the real run):

$ php <cacti_dir>/script_server.php
PHP Script Server has Started - Parent is cmd
<cacti_dir>/scripts/ss_host_disk.php ss_host_disk host.example.com 76 1:161:10:5:50 query used

and only get back:

PHP Script Server has Started - Parent is cmd
U

Can anyone help me debug this? I've attached an example graph showing the oddness.

Cheers,

John.

[BSOD2600]

1) increase the snmp timeout for that device... like 5000ms. This is a known an inherit problem with the windows snmp agent regarding disks.

2) some people have had better success converting the template over to pure OIDs -- there is a howto in the doc site.

[jberanek]

1) increase the snmp timeout for that device... like 5000ms. This is a known an inherit problem with the windows snmp agent regarding disks.

This does indeed seem to fix it, many thanks for the fix!

2) some people have had better success converting the template over to pure OIDs -- there is a howto in the doc site.

Hmm, I might have a look for that.