Perl Script Stopped Updating

[gascione]

I had a dozen perl scripts monitoring some linux functions along with 30 or more generic SNMP monitors for bandwidth and other things on some servers. I added two more SNMP monitors today and now none of the perl scripts work any longer. If I execute them from the command line they are fine. The RRD files are update but no data appears. I set all the logging to the highest level of debugging and nothing, no errors. I have cleared the poller cache at least a dozen times, rebooted the server, nothing. These scripts were working fine for the past three months. I'm using revision "b", tried to install "c" but I couldn't even get that working. I also found that cactid doesn't update any of the rrd files for any perl scripts written so I'm stuck with cmd.php.

Anyone have any ideas. I'm so tired of the quirks in this!

George

[TheWitness]

There is a major issue in Cacti 0.8.6b that causes the poller to crash. Your best bet is to upgrade to Cacti 0.8.6c to correct your problem. There are no database changes between releases. Therefore, only files are updated.

Another issue, as the poller is crashed, a table will become filled with data for the RRD process to output. It is called poller_output. If this table get's too large, your poller will consistently time out because php does not have enough memory to read the table.

There are a few 0.8.6b patches if you are a little frustrated. Just go to the patches directory and download the appropriate files.

If your poller_output table has too many rows, just execute the following command from mysql:

mysql --user=root --password cacti
> truncate table poller_output;
> quit

TheWitness

[gascione]

There is a major issue in Cacti 0.8.6b that causes the poller to crash. Your best bet is to upgrade to Cacti 0.8.6c to correct your problem. There are no database changes between releases. Therefore, only files are updated.

Another issue, as the poller is crashed, a table will become filled with data for the RRD process to output. It is called poller_output. If this table get's too large, your poller will consistently time out because php does not have enough memory to read the table.

There are a few 0.8.6b patches if you are a little frustrated. Just go to the patches directory and download the appropriate files.

If your poller_output table has too many rows, just execute the following command from mysql:

mysql --user=root --password cacti
> truncate table poller_output;
> quit

TheWitness

Thank you, I will try all of your suggestions first thing this morning. I guess it does appear that I am frustrated, I am. I had this finally working then a simple change blew it away.

George

[gascione]

I tried the fixes you recommended. Upgraded Cacti as well. It didn't help. I finally found out that all of the Data Sources for the Perl scripts changed to Generic SNMP instead of the scripts. They were not intentionally changed, they changed on their own. This is the source of the frustration, I have no idea why they changed. I changed them all back and now data is being logged to the correct RRA.

I am still having a problem with cactid and the scripts. If I use the binary instead of the cmd.php none of the perl script rra's get updated. I switched the poller to the binary and looked at the time/date stamp of all the rra's. After the normal 5 minute update I then reviewed the time/date stamps and all the RRA's were updated except the ones that called the perl scripts.

This just doesn't work for perl scripts. Any thoughts?

George

[rony]

Did you resently import a template?

[TheWitness]

Run cactid 1 time from the command line in debug and post your results.

TheWitness

[gascione]

Did you resently import a template?

No, didn't import a template.

George

[gascione]

Run cactid 1 time from the command line in debug and post your results.

TheWitness

[gascione]

It seems that some of the perl scripts are working fine, the one that isn't is a script that I modified. I took the loadavg script and modified to accept a single parameter, the remote computer host name. I use a wget inside the load avg to query a 1 line PHP script on the remote computer which returns the uptime string from a linux server. It's dirty but works great and literally no overhead.

This script works fine from the command line, and fine when run normally in cacti using cmd.php. It doesn't work with cactid. Could it be the parameter being passed to the script?

George

[TheWitness]

Ok George, if you would have listened the first time, I would not have to be posting so listen:

PLEASE RUN CACTID from the COMMAND LINE, do not run poller, all you need to do is the following:

$mycacti_dir>./cactid

Do it in debug, it will get all the hosts by default.

Thanks,

TheWitness

p.s. Don't mean to be nasty, just tired.

[gascione]

Ok George, if you would have listened the first time, I would not have to be posting so listen:

PLEASE RUN CACTID from the COMMAND LINE, do not run poller, all you need to do is the following:

$mycacti_dir>./cactid

Do it in debug, it will get all the hosts by default.

Thanks,

TheWitness

p.s. Don't mean to be nasty, just tired.

Thanks for the help. Would you rather me send it to you as an attachment? I'ts lenghty.

George

[maxpower]

Just as a side note, I also had this same problem happen. I believe that it actually happened, for me, when upgrading the version of PHP on the server running cacti. Perl scripts run fine from command line, but cacti just stopped graphing them.

If you need to be graphing these things in the meantime, you could set up net-snmp on the localhost to execute the scripts for you and output them to SNMP OIDs. It's a temporary work around, but it's something to help get by. Read the man page on "snmpd.conf" if you need instructions on how to do this. Very quick and easy to do.

[gascione]

Just as a side note, I also had this same problem happen. I believe that it actually happened, for me, when upgrading the version of PHP on the server running cacti. Perl scripts run fine from command line, but cacti just stopped graphing them.

If you need to be graphing these things in the meantime, you could set up net-snmp on the localhost to execute the scripts for you and output them to SNMP OIDs. It's a temporary work around, but it's something to help get by. Read the man page on "snmpd.conf" if you need instructions on how to do this. Very quick and easy to do.

Thanks I will check into the net-snmp work around. I didn't do any upgrades on this machine, basically took a clean machine and installed cacti.

George

[TheWitness]

E-Mail is fine.

TheWitness