Poller running twice

[jarosoup]

After adding a script to login to a rouer I noticed the logins happening twice, at the same time. I turned on the logging level to high and notice that everything gets polled twice within 5 seconds of the first poll. I've tried changing the Poller Failure count to 1 which doesn't seem to help. The first pass of polls always returns valid values, and the second poll right on top of the first returns the same values. Latency and network performance on both ends is not a problem.

Is there anyway to stop this behavior? Is it normal?

System Specs:
Clean install of Cacti 0.8.6c (cmd.php) with the 3 official patches
Fedora Core 3, PHP 4.3.9, MySQL 3.23.58-13

Any insight would be greatly appreciated.

Edit: I have verified that there is only one entry in the crontab

[TheWitness]

Review your poller cache or your poller_items table for dups.

TheWitness

[jarosoup]

Thanks for the reply. I checked both and no duplicates. I deleted the only host I had and let cacti run for 15 minutes then recreated the same host, but all of my graph items (8 under 1 host) are showing duplicates in the log and scripts still run twice. While checking the poller cache, I noticed that clearing the poller cache and log file have no effect...Everything else seems to work fine with Cacti as far as I can tell.

[TheWitness]

Please post your log file.

[jarosoup]

Here is a chunk of my log file from this morning. I've created one host, and added to it the SNMP Interfaces included from a host template that has other templates. The only other host is the default localhost which is also running (and showing duplicates). I turned on the log to Debug mode at 9:50. Thanks for your help.

[TheWitness]

Very unusual. Do you have an issue sending my a mysqldump of the database? Send to my published e-mail address. I will review this evening.

TheWitness

[jarosoup]

While testing, I've discovered something interesting. First of all, I uninstalled the current version (rpm -e cacti..., dropped the db, removed the crontab entry) and reinstalled from scratch 0.8.6c without the patches applied. Same problem. Then did it again with 0.8.6b. Same thing. I've still go an old directory from 0.8.6b (my first cacti install - ran for about 2 months) and the log file for it...that log file doesn't show this problem.

Anyway, I've just discovered the 2 proccesses:

Code: Select all

[root@wiggum etc]# ps ax | grep cacti
19486 ?        Ss     0:00 /bin/sh -c php /var/www/html/cacti/poller.php > /dev/null 2>&1
19487 ?        S      0:01 php /var/www/html/cacti/poller.php
19485 ?        Ss     0:00 /bin/bash -c php /var/www/html/cacti/poller.php > /dev/null 2>&1
19488 ?        S      0:01 php /var/www/html/cacti/poller.php
19552 pts/2    S+     0:00 grep cacti

This seems very odd...shouldn't there only be one running under sh? Does this help at all? I'm not sure where to check now. My crontab entry looks like this:

*/5 * * * * cacti php /var/www/html/cacti/poller.php > /dev/null 2>&1

I've got the db file if you still want it but I think there are other issues...

[/code]

[TheWitness]

Send it along. You may also want to review your crontab for multiple entries, which I am sure you have already done.

TheWitness

[rony]

Have you checked the users cron?

Code: Select all

crontab -l -u cacti

You might have 2 entries and not even know it.

[jarosoup]

Okay, I think this is fixed now. I've again verified that there is only 1 entry in crontab, but I've discoverd a file named cacti in /etc/cron.d with the same crontab line that is in /etc/crontab. I'm not sure how this got there, but I moved it and now all seems well - no more duplicates and ps ax | grep cacti comes up empty. I'll keep an eye on this tho for now. Thanks for all of the help folks

[jarosoup]

As I haven't dealt much with cron jobs in Linux, this may not be correct...but, when installing the rpm for 0.8.6c-fc3 it places a file called cacti in /etc/cron.d that contains the same cron line as listed in the install manual. So, I'm assuming that this step (adding the cron line to /etc/crontab) is no longer needed or that in /etc/cron.d shouldn't be there? Am I missing something here?