Huge Apache log file - Cause of Cacti

[Guest]

I've got a real bad problem with cacti. From time to time, (2 times a month?) my Apache error log is full of "Libpng error: stream error".

Fount 2 people with the exact same problem, however, I can't reproduce it.

http://article.gmane.org/gmane.linux.sentinix/562

old one here:
http://www.linuxarkivet.se/mlists/rrd-u ... 00133.html

My log file was 1.3 GB, then the partition ran out of diskspace... once it was 2.0 GB, the max file size, apache died then.

I'm running Debian 3.0 stable, cacti 0.8.5a, zlib 1.1.4.
Thx!

[Guest]

Well, happened about 3times after i created this topic. Hmpff

[PapaFerret]

And here I thought it was just my box doing that.

It started on my development Sentinix box 2 weeks ago testing out the latest CVS for Cacti. I've also noticed what seems to be a memory leak since then (watching it drop using 'top') but that may have been there before.

Since attempting to update Cacti I'm lucky if I can kept httpd up for over 8 hours.

[TheWitness]

PapaFerret,

Are you saying you upgraded to the Beta version of Cacti without first understanding that the buyer of said Alpha/Beta software should beware? Users should not be upgrading to Alpha/Beta versions without first accepting the risks.

At a minimum, please provide the following information relative to your deployment.

OS Version,
PHP Version,
MySQL Version,
RRD Version,
Net-SNMP Version

Please keep in mind also that when you download a product packaged by a third party, it likely may have modified the original product to suit certain integration needs. Therefore, always contact the integrator prior to upgrading.

We have already identified problems with certain versions of PHP that cause servers to go into race conditions with the Cacti 0.8.6 product and those versions will be noted in the release notes.

What I am trying to say is that things like this happen, especially with Alpha/Beta products. Note: GNU means NO Warranty.

All beatings aside, your TESTING information is very important to the development of Cacti. As it is still a Beta product, If you are willing to work with me at resolving the root of the Beta issue, that would be great. You can reach me at larryjadams@comcast.net.

Regards,

TheWitness

[Guest]

LOL...whoa there...put down the stick and back away...

...please?...

...lol...

I was just chiming in that my 'development' box seems to be having the exact same same issue since...

The Sentinix disto is nice because you go from the start of the install of the distro to logging into Cacti in about 20 minutes. All of the preliminary set-up with the configuration of mySQL and Cacti (as well as Nagios, Snort, Nessus, PostFix, etc) is done by the install.

Unfortunately the version of Cacti that comes with it is .83a. I set-up a development box for testing the latest CVS so I could kick the tires and get used to all the changes for when it *is* released.

Thank you for the offer to assist with the troubleshooting, but don't worry helping me at the moment (just keep working on getting the latest out the door... ). Good to know about the issue with PHP though...I'll get you the versions I'm running and hopefully the first poster can do the same so we can determine the common factors (I wasn't asking for help myself...really).

Besides, I'm in training in Troy all this week and won't be back in Southfield until next week, and then I'll be tied up with a project launch for about 2 weeks...durn real-life job.

thanks,

--john
--jendahl [at] odinspear.com

[TheWitness]

Didn't notice you were from the Detroit area. How bout that. Were practically neighbors.

TheWitness

[PapaFerret]

I noticed that about a month ago, but I've been on the 'lurk and learn' program.

Here's the requested versions from my system (Sentinix 0.70.5):

GNU/Linux 2.4.21 (OpenMosix)
Apache 1.3.29
PHP 4.3.4
MySQL 4.0.16
Net-SNMP 5.0.9
RRDTool 1.0.45

And if you need another testing box, I've got one I can give you admin access to in my den (that's what testing boxes are for after all...)

[TheWitness]

Ok,

First of all John, PHP 4.3.4 caused (past tense) a race condition in poller.php. However, the update that was committed on Monday morning corrected that. As far as RRDTOOL is concerned, Ian thinks that the version you are using has problems. I have been running 1.0.48 without issue for some time. The race condition is likely what hosed your apache server.

In PHP 4.3.4, the usleep function did not work. Bad thing it turns out.

Upgrade both PHP and RRDTOOL and you should be back in business. You should also feed this information back to that group so that they can alert their users to the problem in PHP 4.3.4 if that's what they are currently distributing.

Also, please note we are still in development with CACTI 0.8.6 although very close. I just added a new feature tonight. Therefore, keep checking the CVS or redownloading as you test.

TheWitness

[PapaFerret]

I'll try that, thanks.

Despite the first name I wasn't the original poster in the gmane newsgroup for nagios, but I had found that same post while researching the issue myself.

I'll post an advisory in both the Nagios and Sentinix newsgroup/lists and let you know what happens upgrading.

[pierre-oord]

I still have this problem. I don't know when, but it happends, and I get sick of Apache crashing each time.

Running Debian stable, but with some testing packages and kernel 2.6.

If I typ "man zlib" on the bottom I find "Version 1.2.1.1", so it's newer and without buffer overflows, like someone said above in this old threat.

I really hope someone has a fix!

Running:
- RRDtool 1.0.48
- Apache 1.3.31
- PHP 5.01
- MySQL 4.0.20

This error occured also on older PHP versions. I really hope this error can be Fixed, I think cacti is perfect then

(Oh, and thank you all for creating a tool like this!)

[TheWitness]

pierre-oord,

Would you be willing to work live with me on this problem? Send an e-mail to larryjadams@gm.com.

TheWitness