Gaps in Cacti graphs

BobDistad · Post by **BobDistad** » Mon Nov 23, 2009 5:02 pm

I am running cacti 087e on a Windows 2003 server. We have been noticing gaps of data in some of out router network graphs. There are errors in the cacti log that say: POLLER: Poller[0] WARNING: Scheduled Task is out of sync with the Poller Interval! The Poller Interval is '300' seconds, with a maximum of a '300' second Scheduled Task, but 600 seconds have passed since the last poll! Not sure if they are related or not.

Post by **gandalf** » Tue Nov 24, 2009 4:53 pm

This may be related.
But it is not considered a good style to post graphs as part of some strange "image container format" like *.doc; the forums allow for attaching images straight away.
You may refer to the 2nd link of my sig for a start on debugging
R.

BobDistad · Post by **BobDistad** » Tue Nov 24, 2009 4:57 pm

"You may refer to the 2nd link of my sig for a start on debugging "

I don' understand what you mean by that.?.

BobDistad · Post by **BobDistad** » Tue Nov 24, 2009 5:02 pm

Never mind. I found the link.

BobDistad · Post by **BobDistad** » Tue Dec 01, 2009 3:28 pm

I turned on debugging and captured some of the messages from the last gap. The graphs that are gapping are
aubmqr01
dtnmqr01
fbomqr02
fbomqr01
mlamqr01
owtmqr02
owtmqr03
owtmqr01
plymqr01
plymqr02
stnmqr01

The gaps are at 13:30 and 13:35. Any suggestions? I do see that "WARNING: Scheduled Task is out of sync with the Poller Interval! " message at 13:30.

Post by **gandalf** » Tue Dec 01, 2009 4:14 pm

Are ALL graphs gapping at that time? This will lead to a problem with the poller.
Or are only some gapping? This will lead to a problem with those targets or data sources
R.

BobDistad · Post by **BobDistad** » Tue Dec 01, 2009 4:45 pm

The gaps are occuring on all graphs. So I would think it is the poller. The devices are all active during this time.

BobDistad · Post by **BobDistad** » Thu Dec 03, 2009 12:55 pm

More information:
I am also noticing multiple php-win.exe processes in Task Manager - between 50-80 each, and there is a cmd.exe process for each processes. Around 2:00 yesterday afternoon, I killed all the php-win.exe processes on the server. After that, the number of processes began increasing by 10 approximately every 2 hours. There were over 100 when I cleaned up all the processes a few minutes ago. These processes are submitted every 5 minutes, so at this rate, it would imply that some of them are finishing and some are hanging. The gaps seem to become more frequent as more proccesses showed up in Task Manager.

Post by **gandalf** » Thu Dec 03, 2009 2:51 pm

Moving to windows due to your last post.
R.

BSOD2600 · Post by **BSOD2600** » Thu Dec 03, 2009 8:38 pm

Try using Sysinternals Process Monitor/explorer to see what all those cmd/php processes are running. Typically its some sort of script which isn't behaving/timing out. Otherwise, can either use "User Profile Hive Cleanup" or a taskkill scheduled task to mitigate the issue.

BobDistad · Post by **BobDistad** » Mon Dec 07, 2009 12:33 pm

I have set up a job that kills the outstanding tasks every 15 minutes. That has resolved the graphs stopping, but I am still getting the gapsm and the "POLLER: Poller[0] WARNING: Scheduled Task is out of sync with the Poller Interval!" messages several times an hour. What do these messages mean?

nuinco · Post by **nuinco** » Tue Dec 08, 2009 2:27 pm

I had the same issue when using Spine as the poller type.

Switched back to cmd.php and it is working fine.

I have about 400 graphs and it still completes polling in less than 30 seconds. So performance is not an issue yet.

BobDistad · Post by **BobDistad** » Mon Dec 14, 2009 3:15 pm

Thanks for the update, but we are already using cmd.php as the poller type.

Gaps in Cacti graphs

Gaps in Cacti graphs

Who is online