Windows 2003 hard drive space accuracy/missing polls

[matguy]

I've setup

Cacti v0.8.6g
PHP 5.0.5
MySQL 4.1
RRDtool 1.2.11
Net-SNMP 5.2.1.2

My cpu graphs are decent and look accurate other than a just a few blank spots and the network traffic graphs look perfectly amazing. The graphs for HD space have some issues though. They seem to show data from sporatic polling results as well as innacurate data. The used space isn't reporting accurately at all when it does report, and I'm pretty sure the physical drives on the servers don't dynamically change their total size often but it seems that the total size of the hd's also report inaccurately when at all.

As you can see here on 3 drives on one of the servers here:


The real hd sizes are as follows:
C: 32GB
D: 36GB
E: 500GB

If it was at least accurate on the ones it does report I could try to plot points that are reported, but they're so far off I can't even do that. I am experiencing this same issue on about 10 servers I'm polling. The polling doesn't seem to run too long, usually right around 30 seconds and it's set to poll every 5 minutes and kill the process after 4 minutes (standard windows scheduler running the process under the system account.) The C and D drives are local to the server being polled (cacti is not on the same server) as a RAID5 array and the large E drive is on a SAN.

Here is another server in the same configuration showing similar issues:


As you can see, I did something around 8:30 this morning that helped (not sure exactly what) but still didn't quite fix it since the total size is still reporting inaccurately in some places.

Any ideas?

[BSOD2600]

1) install the patches for cacti
2) Are you using the Microsoft SNMP agent or net-snmp on the boxes?
3) Are the servers / cacti machine under a heavy load? This could account for the gaps in the graphs. Also another possibility would be network latency. What are the ping times for each server having problems? Might also want to increase your php timeout values (in php.ini).

[matguy]

So, how do I install the patches on a windows machine? All I saw were linux packages and instructions.

I'm using net-snmp on the cacti polling box, regular Microsoft SNMP on the machines being polled.

None of them are running particularly high load. Both examples here are dual proc Xeon (P4) machines with at least 4gb of ram. I'm experiencing similar on a new SQL server that isn't production yet, which is a Quad Xeon with 16GB of ram under very little load. Ping times were between <1 and 22ms. They're all in the same rack area running through the same switch. I'll try increasing php timeout.

I changed:

max_execution_time = 90
max_input_time = 120
memory_limit = 16M

[matguy]

I'm also looking through your updated Windows install guide, which looks to be not only more complete than the one in the official install manual, but also actually correct, if you know what I mean

I'll look through it some more tomorrow as there were a lot of things I had to get going through google searching errors as I got them during setup, and who knows if I'm catching them all. Obviously, evidence would point to no. It's quite probable there's a few things in your guide that could help me even now.

[matguy]

Those php.ini changes didn't seem to make any difference. I also tried going back to the microsoft SNMP service on the server, but that didn't seem to make any difference.

[BSOD2600]

1) All other graphs are working on devices, except disk space on Windows boxes?
2) If you run snmpget against a windows box, for the disk space several times in a row, does it always return data?
3) You also increased the snmp timeout in the device(s) themselves, right?
4) What version of snmp are you using?
5) I'm confused... so you have been using the net-snmp snmp agent on the boxes instead of the microsoft one?

[matguy]

1) All other graphs on other devices are working decently: Network bandwidth graphs are perfect, some processor usage graphs show minor gaps. All boxes being polled are Windows servers.

2) I've never run snmpget against any of the boxes. Is this a manual process, or a tool from within Cacti? I apologize, I'm fairly new to SNMP in general.

3) I have not increased the SNMP timeout on the individual devices, I'll work on that this morning.

4) I'm using the SNMP client from the Windows 2003 cd, specified in cacti it has been left at the default version 1.

5) I was using net-snmp on the cacti server and just recently switched it to the Microsoft SNMP agent, but all the bexes being polled are the standard Microsoft SNMP agent.

[BSOD2600]

1) Instructions on how to install the patches are in my guide. Have you got those installed yet? One of the patches fixes some known problems with the partition query.

2) If you click on verbose query next to SNMP - Get Mounted Partitions in the device, do you get a list of all partitions consistantly?

3) snmpget / snmpwalk is a net-snmp tool. So for example to walk the partitions installed on a box, you would run something like: snmpwalk -v 1 -c public 192.168.1.1 .1.3.6.1.2.1.25.2.3.1. I'm sure it will work some of the time, since your graphs reflect it does partially...

[matguy]

I'm figuring out the snmp get testing now.

[matguy]

Didn't see your last post as I was making mine. I'm looking at the patching now as well.

[matguy]

I've installed the patches, once cygwin was installed and the bin folder added to the windows path, the updates went well by simply copy-pasting the code to a batch file and running from the cacti directory, then resetting persmissions on the files when the graph view wouldn't open anymore.

Clicking on a verbose query, I get the same reponse 5 times in a row:

+ Running data query [8].
+ Found type = '6 '[script query].
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ XML file parsed ok.
+ Executing script for list of indexes 'c:\php\php-win.exe -q C:\cacti\scripts\ss_host_disk.php oitex2.seattleu.edu 3 1:161:500:public:::MD5::[None] index'
+ Executing script query 'c:\php\php-win.exe -q C:\cacti\scripts\ss_host_disk.php oitex2.seattleu.edu 3 1:161:500:public:::MD5::[None] query index'
+ Executing script query 'c:\php\php-win.exe -q C:\cacti\scripts\ss_host_disk.php oitex2.seattleu.edu 3 1:161:500:public:::MD5::[None] query description'
+ Executing script query 'c:\php\php-win.exe -q C:\cacti\scripts\ss_host_disk.php oitex2.seattleu.edu 3 1:161:500:public:::MD5::[None] query sau'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'
+ Found data query XML file at 'C:/cacti/resource/script_server/host_disk.xml'

The snmpget and walk commands returned fine as well. Lots of info, the same each time.

Since installing the patches about an hour ago, it seems most of the hard drive graphs are coming in well. One was not getting hard drive info at all, but looking in the rrd dir one of the rrd files for that server/device had permissions issues, which I assume came from being created with the poller running as a system process(?) After fixing that it looks like it's starting to fill in on the graph. It's only been about 20 minutes, so there isn't much there yet, but I see some solid activity. All the others are looking great.

Barring any more polling issues, time for speed tuning. Thanks for the help, your guide really should be included, or at least referenced in the official instructions.

[BSOD2600]

I Thanks for the help, your guide really should be included, or at least referenced in the official instructions.

They plan to eventually....

Glad you got it working.

[nw]

Hallo,

I'm using Cacti to monitoring different Linux and Windows Servers. All Linux Servers are displayed right, but the windows servers sometimes shows a graph and sometimes not.
It would be nice if someone can tell me were i get the manuals or the guid for installing the Patches mentioned in this topic

thanks

[matguy]

The patches are for users running the cacti server on a windows machine and using cygwin was how I ran them by simply copy-pasting the commands listed on the patch page in to a quick batch file, but it seems you can also download the individual files and drop them in the correct places.

The guide I used to get my Windows Cacti Server installation running is in BSOD's signature, or click here: http://bsod2600.home.comcast.net/install_windows.html

[nw]

Yes thanks for the guid and the url

but i thought those patches are also availible for Linux. I'm unsing a SuSE Server to run Cacti Server and wenn i'm monitoring Windows Server the graphs are sometimes there and sometimes not.

I have opend a new topic so please post your answeres there, because i think my probleme haven't anything to do with this topic