poller_output table problem?

MagnaChef · Post by **MagnaChef** » Thu Mar 16, 2006 10:49 pm

I will give this a try either tomorrow or over the weekend and let you know.

mahuani · Post by **mahuani** » Sat Mar 18, 2006 11:23 pm

Ever since I changed the memory limit in PHP, I haven't seen any problems with my graphs. Of the two servers I am running Cacti on, the one that broke had the limit set to 8M and the one that didn't was set to 128M.

MagnaChef · Post by **MagnaChef** » Mon Mar 20, 2006 10:13 am

TheWitness wrote:Magna,

If that patch I sent works for you, please post. Also, please post results of backing off of RRDtool 1.0.50. I believe that 1.0.48 was known good.

Larry

When I came in to work on Friday (a day after running my "meta-glue" script to update the RRDs) I ran into a whole slew of system problem (journal commit I/O errors, etc). After rebooting the machine, it came up with some SCSI errors, and the filesystem is now read only.

Luckily I had another server that I was able to migrate the platform over to. I swapped the file you gave me with the original, and things are now back to normal.

I have yet to try and downgrade the version of RRDTool to 1.0.48. I've been busy fighting fires and I'm looking for a minimal downtime for removing RRDTool 1.0.50 and then installing 1.0.48 (any suggestions?).

Larry, Thanks for your help.

Attached is the modified rrd.php file.

Grogman · Post by **Grogman** » Tue Apr 11, 2006 12:43 pm

Hello to everybody,

I have the same problem with poller_output being not empty. My graphs still work, but they show missing data very often.

I've increased the amount of memory of PHP to 600MB, and tried the patch, but the problem is still happening...

Thanks!

MagnaChef · Post by **MagnaChef** » Tue Apr 11, 2006 12:45 pm

How many devices are you polling?

Grogman · Post by **Grogman** » Tue Apr 11, 2006 12:49 pm

Hello,

Currently, cactid is polling 12 devices, and it takes about 40 seconds.

Grogman · Post by **Grogman** » Thu Apr 13, 2006 5:24 am

Hi again,

Now, after installing the patch, I've also upgraded the rrdtools from 1.0.49 to 1.2.12. Nice graphs, but still a lot of data in poller_output, and a lot of gaps in the graph's data

So, I've downgraded to version 1.0.48 of rrdtools, and still the same problem

Any hint? Am I missing something?

Thanks!

MagnaChef · Post by **MagnaChef** » Thu Apr 13, 2006 9:27 am

This may not affect it, but did you restart apache after making the memory change in php.ini?

Grogman · Post by **Grogman** » Thu Apr 13, 2006 9:36 am

Hi,

Yes, I've restarted the httpd service a couple of time since changing php.ini...

I've also discovered something quite strange: If I run poller.php from the console, poller_output gets empty. But when poller.php is executed by cron, entries start appearing in poller_output. First I tought that it could be a problem of permissions of cactiuser, but I've changed crontab making the root user execute poller.php instead of cactiuser, and poller_output continues growing...

Thanks for your response!

Grogman · Post by **Grogman** » Mon Apr 24, 2006 2:56 am

Hi

The problem appears to be turning worse... The gaps in the graphs are bigger and more frequent each day

Any hint?

Thank you

fmangeant · Post by **fmangeant** » Mon Apr 24, 2006 4:24 am

Hi

I had the same problem last week on a production server, running RedHat AS 4 Update 2, RRDtool 1.2.12 and Cacti 0.8.6h + the 4 patches.

The poller_output table contained 15 millions lines, and it took more than 4 minutes to truncate it

Grogman · Post by **Grogman** » Wed Apr 26, 2006 2:55 am

Was your problem fixed when you truncated poller_output, or is the table repopulating again?

I've truncated poller_output a lot of times, but in the next polling cycle, the table gets populated again...

fmangeant · Post by **fmangeant** » Wed Apr 26, 2006 3:12 am

Truncating the poller_output table fixed the problem, it goes back to 0 after each polling cycle.

Grogman · Post by **Grogman** » Wed Apr 26, 2006 3:25 am

You are lucky then

I'm still wondering how to fix the matter...

Grogman · Post by **Grogman** » Tue May 02, 2006 10:42 am

Hi,

I think I'm isolating the origin of the problem. The holes in the graphs only appear when I'm monitoring two specific devices of my network. When I disable these two devices, the holes almost disappear. Now I'm working towards discovering what have both switches in common, and why Cacti doesn't like them.

It looks like the only thing these two devices have in common, is that they're a bit slow when replying to snmp requests. However, the snmp requests don't time out, the devices reply, and still, the global polling time remains in about 35 seconds. Of course, another thing I can't understand is the matter regarding that the poller output gets empty if I run the poller from de CLI, even using Cactiuser...

Any hint?

Thanks!

Cacti

poller_output table problem?

Same Problem

Who is online