TECHZEN Zenoss User Community ARCHIVE  

Ignoring data from a failed sampling

Subject: Ignoring data from a failed sampling
Author: [Not Specified]
Posted: 2014-02-06 21:56

Hi Everyone,

I am using the winexe command along with wmic to scan Windows Event Logs for certain messages. Basically, my command looks like something like this, where it counts the number of times a certain message appears:

winexe -U '${here/zWinUser}%${here/zWinPassword}' //${here/getDeviceName} "cmd.exe /C wmic ntevent where (message like '%widgets%' and LogFile = 'Application' and SourceName='Application Error') list brief" | echo "OK|Errors=$$(wc -l)"

I want to be alerted only when a new instance of the message I am looking for is recorded in the Event Log. I am using the ValueChangeThreshold to do this.

My problems is that occasionally the Event Log scan fails and I get a count of "None" for the number of relevant messages. At some point, Zenoss reconnects and records the correct number during the sampling. This triggers the alert and creates a false positive.

What's the best way to handle this Any advice would be appreciated. Thank you!



Subject: I suspect that the WMI
Author: [Not Specified]
Posted: 2014-02-10 11:30

I suspect that the WMI service on the windows box is the culrpit. In many of my environments, I restart that service once a day to help fight it, but even that isn't perfect.

Hydruid



Subject: Thank you. I would most
Author: [Not Specified]
Posted: 2014-02-10 11:58

Thank you. I would most certainly agree that there is either a comm issue or wmi is acting up. What if I don't care about missing the sampling once in a while. Is there a way to ignore it in Zenoss. My main concern is not producing false positives.



Subject: You could use an escalation
Author: [Not Specified]
Posted: 2014-02-11 08:28

You could use an escalation transform of some sorts.....in other words, after it returned a different value 5 times (5 polling cycles) then it would alert you. That would help fight false positives.

Take a look at these:
http://wiki.zenoss.org/Transforms_-_Escalate_by_Count
http://wiki.zenoss.org/Transforms_-_Escalate_by_Count_in_a_Time_Window

Hydruid



Subject: You could use an escalation
Author: [Not Specified]
Posted: 2014-02-11 08:45

You could use an escalation transform of some sorts.....in other words, after it returned a different value 5 times (5 polling cycles) then it would alert you. That would help fight false positives.

Take a look at these:
http://wiki.zenoss.org/Transforms_-_Escalate_by_Count
http://wiki.zenoss.org/Transforms_-_Escalate_by_Count_in_a_Time_Window

Hydruid



< Previous
survey of zenoss for useful OSS monitoring tool
  Next
Some Information Windows event log still be collected by Zenoss
>