Single/First CPU and Disk Graphs are broken

[andymack]

Hi there,

I have an issue with both my CPU and Disk graphs for my Windows servers not reporting - specifically on Index 0 for CPU and Index 1 of Physical Disks. And just so I am clear, all remaining indexes are working as expected for CPU and disk.

My log file shows the following line for each of the machines affected:

Code: Select all

02/21/2009 08:35:26 PM - SPINE: Poller[0] Host[153] DS[1872] SS[2] WARNING: Result from SERVER not valid. Partial Result: ... 

If I run the logging in Debugging mode, I receive the following line:

Code: Select all

02/21/2009 08:45:39 PM - SPINE: Poller[0] Host[105] DS[1872] SS[2] SERVER: C:\Inetpub\wwwroot\cacti\scripts\ss_host_cpu.php ss_host_cpu SERVER_NAME 105 2:161:1000:25:40:public:::::: get usage 0, output: U

When I run the same command outside of Cacti, I receive a real value. Looking inside of the MySQL database (poller_item table), I looked for where the command was being looked up. This is what that line has:

Code: Select all

C:\Inetpub\wwwroot\cacti\scripts\ss_host_cpu.php ss_host_cpu SERVER_NAME 105 2:161:1000:25:40:public:::::: get usage 0

I am very puzzled needless to say and have managed to pain stakenly clean these rows in that table that have those extra characters in it, but creating new CPU or disk graphs results in the same thing re-occuring for those new graphs.

If you need further information to offer a suggestion, please let me know. Below is a system summary:

• Cacti Version -- 0.8.7d
  Cacti OS -- Windows 2003 EE 64-bit
  SNMP Version -- NET-SNMP version: 5.4.2
  RRDTool Version -- RRDTool 1.2.x
  PHP Version -- 5.2.6
  MySQL Version -- 5.0.x

I am thinking there must be something messed up with the current template, but I just upgraded to 0.8.7d and am only running one custom script for collecting UPS statistics and this was happening before hand for months now running with 0.8.7b. My Cacti was originally installed using the Windows installer package from the Cacti forums and was not configured a piece at a time.

Thanks for any ideas!

[andymack]

Has no one else experienced this problem? Or did I not include enough information?

[TheWitness]

You have to be very careful as to how you upgrade the "scripts" and "resource" directories. The latest version of Cacti included fixes to these templates. During the upgrade we repopulate the poller cache. So, you should verify that you have the latest files and then re-populate the poller cache or simply edit and then save each host affected.

TheWitness

[andymack]

Thanks for the note. I have tried renaming both the "scripts" and "resource" directories to "scripts.old" and "resource.old". Then I copied the new scripts and resource directories from the Cacti 0.8.7d zip file as a replacement. Then I went to "rebuild poller cache" and had it run. The web page shows that the entries are correct, but the graph is still not showing the correct data and the database is showing the additional characters again.

I had tried backing up the entire Cacti DB using the MySQL admin tools and found that there seemed to be other tables that include these bad characters. As a result of an earlier test I had scrubbed the entire database of those characters, but they re-appeared over time.

This was also happening on a previous release of Cacti (0.8.7.b), though I am not sure if it was the exact same thing as I never checked the database for bad characters.

[TheWitness]

Rerun the queries. Other than that I would have to do a GotoMeeting or something. Right now it's almost dinner time and we are going out.

TheWitness

[andymack]

What I will try first here is to flush out the bad characters from the database again now that I have replaced those two directories and see if the problem comes back. I will let you know how that goes.

Enjoy supper.

[andymack]

After further testing, the values slowly started to come back into the system causing those specific graphs to slowly fall off yet again. My first thought is that there could be a bad PHP file somewhere along the line. Would it be worth trying to replace all the PHP files with the latest build of 0.8.7d? Thanks!

[andymack]

Another interesting thing I am seeing is this:

Code: Select all

C:\Inetpub\wwwroot\cacti\cli>php poller_reindex_hosts.php
ERROR: You must supply input parameters

Don't know if that is at all related to this or not.

[andymack]

Well everything seems to be working once again. Here is what I did:

• 1. Backed up and removed live cacti folder.
  2. Downloaded Cacti 0.8.7d and placed files into Cacti folder.
  3. Replaced CactiD with v0.8.7a (was running 0.8.7c).
  4. Cleaned Database of all remaining bad data.
  5. Restored RRD files.

After all of this, everything seems to have worked. I don't know where things broke down, but at least it is all up and running. Thought anyone who comes across this would like to know.