SOLVED: SNMP - Get Mounted Partitions "Used" Issue

[tymbow]

I've suddenly got this bloody annoying issue on one of my Windows hosts.

Caveat: I've just upgraded to 0.8.7c but I have seen this briefly in earlier versions but it "went away".

The "Used" value for the "C:" drive on one Windows 2003 server keeps graphing as 0 but it was working fine. I ran the poller in Debug mode and 0 is the value that Spine (or "cmd.php" - I've tried both) both gets from SNMP and writes to the RRD.

To make it annoying an SNMPGET/SNMPWALK reveals the counter is NOT 0 - it returns the correct value.

I tried reboots, restarting SNMP, re-installing SNMP, re-cache and changing SNMP timeout. All have the same result - it works again for 1 or 2 poller cycles then goes back to reporting 0.

Any thoughts?

EDIT: This is getting weirder. Now it seems to be alternating between "Used" being 0 for a while with "Total" working OK then "Total" is zero for a while and "Used" starts working again. As before an SNMPGET/SNMPWALK shows counters are not 0 (I've setup a little script that runs before/after the poller runs that saves the values to a file so I can check).

[BSOD2600]

I tried reboots, restarting SNMP, re-installing SNMP, re-cache and changing SNMP timeout. All have the same result - it works again for 1 or 2 poller cycles then goes back to reporting 0.

Any thoughts?

Sadly, no. You've tried all of the common solutions/fixes.

What method is your 'SNMP - Get Mounted Partitions' using? Go to Data Queries, find SNMP - Get Mounted Partitions, and look at the XML Path. If it's still using script_queries, then you might give the script server method a shot instead. http://docs.cacti.net/?q=node/516

[tymbow]

It's always been the script server method.

[tymbow]

I've solved it.

It's being caused by the dreaded Windows profile unload problem (Windows Event ID 1517). The account that is running the poller is not unloading its profile properly anymore for some reason. When it does unload properly the SNMP counters do get read by Cacti properly.

[BSOD2600]

hmm, weird. snmp is a service which shouldn't have any effect by a user or its profile not unloading.

Glad it's working.

[lottabull]

I've solved it.

It's being caused by the dreaded Windows profile unload problem (Windows Event ID 1517). The account that is running the poller is not unloading its profile properly anymore for some reason. When it does unload properly the SNMP counters do get read by Cacti properly.

Thanks for posting this. I had the same exact problem with returned disk space total and used values.
Your reply got me looking at the server's event log and I found a bunch of entries from userenv.

Tracking those down let me discover the sysvol\domain\Policies folder was gone. It must be a common enough problem that W2K3 has a supplied program to recreate it.

After recreating the Policies folder the SNMP responses are working again.

[ataro]

How do I change the polling interval/SNMP timeout value?

I tried the below URL, works fine for similar issue.

http://docs.cacti.net/usertemplate:data ... disk_usage