Cacti log file reporting "Poller Output Table not Empty

[TheWitness]

One last gasp. When you download and install the new PIA. Do the following:

1) Edit <path_cacti>/poller.php
2) Goto Line 453.
3) Comment out poller_top and poller_bottom
4) Test
5) Uncomment poller_top and poller_bottom,
6) Turn 'usleep($bhah * something)' to 'sleep(intval($blah))'
7) Test again.

Tell me which one fixes the problem. Here is the reference code:

Code: Select all

			api_plugin_hook('poller_bottom');
			usleep($sleep_time * 1000000);
			api_plugin_hook('poller_top');

[philipz]

Hey Larry,

Some observations:
- I have PIA 2.9beta installed
- Commenting out poller_top and poller_bottom did not make any difference.
- I changed
usleep($sleep_time * 1000000);
into
sleep(intval($sleep_time));
but that did not make any difference either.

The only workaround I found is to use spine 0.8.7b (our previous version); at least this one processes RRD's (although sometimes with gaps). The 0.8.7g spine processes 0 RRD's.

[TheWitness]

You are running Cacti 0.8.7g right? I beginning to wonder. This is the problem when you cross-post, you confuse an already busy monitor.

TheWitness

[philipz]

You are running Cacti 0.8.7g right? I beginning to wonder. This is the problem when you cross-post, you confuse an already busy monitor.

TheWitness

I'm running 0.8.7g indeed.
One thing I found is that running 2 cacti/spine instances on the same box is not the way to go. I turned off 1 instance, and now spine 0.8.7b doesn't give me any error anymore. Changing to spine 0.8.7g immediately gives me the timeouts again and no graph updates.
This means I have a working cacti0.8.7g with spine 0.8.7b (only running for an hour now without errors anyway). Any advice against that setup?
One day I will start from scratch and see where that will lead me...

Thanks for all the help!

[ARC1450]

Is the 1 minute polling/cron an issue still and does it apply to Windows?

I've got a server that is having a "Poller Output Table not Empty" with a "There are '1' detected as overrunning a polling process, please investigate"

There seems to be no issues that I can overtly detect. The polling process hovers around 30 seconds and I don't always have both errors.

Any input?