Graph showing wrong badwidth metrics

eyoung · Post by **eyoung** » Tue Nov 13, 2007 12:22 pm

I have cacti running on a Windows 2003 server and I am graphing the ports for my HP Procurve 3400cl switch. The graphs are showing values of 10's of MB when I know that it should be 100s of MB because the port is 1GB and the amount of traffic from the servers pushing data through those ports add up to 400-650MB. I am not sure if there is a setting that I am missing, so any help would be appreciated.

Also is there a way to change the poller from 5 min. to every minute?

Thanks, I am new to this.

BSOD2600 · Post by **BSOD2600** » Tue Nov 13, 2007 1:36 pm

Windows doesn't support 64bit counters natively, and thusly the 32bit counters are rolling over with the high bandwidth.

Yes, you can install a 1min polling patch http://forums.cacti.net/viewtopic.php?t=16482 for 0.8.6j.

Another option would be to use the Boost plugin.

eyoung · Post by **eyoung** » Tue Nov 13, 2007 2:07 pm

Thanks, I will apply the patch for the polling cycle.

So does your reply mean that there is no way for me to get accurate data regarding the traffic on my switch?

BSOD2600 · Post by **BSOD2600** » Tue Nov 13, 2007 4:18 pm

Oh a switch, I thought you were trying to monitor Windows. Yes, most switches support 64bit counters. Switch to snmpv2 and use the 64bit graph templates. Then if those don't work, try the 1min polling and/or Boost.

eyoung · Post by **eyoung** » Wed Nov 14, 2007 11:42 am

I changed to SNMP V2 and I have tried graphing Bits per second and Bytes per second and I still do not get the correct data. I should be seeing about 800 MB right now and my graph shows just under 10 MB.

I also tried to install the polling patch and my graphs stopped collecting data. When I ran poller.php I got an error message. I changed back to the original files and now it is working again ( with the wrong data).

BSOD2600 · Post by **BSOD2600** » Wed Nov 14, 2007 2:06 pm

Did you recreate the data source and graphs for that switch? If not, you must. simply delete them and tell cacti to re-create them.

What polling patch and what error message?

jay · Post by **jay** » Thu Nov 15, 2007 12:48 pm

You might want to check the Maximum Value within your data template as well. Maybe the figure is not high enough. Can't remember what it should be for Gig and upwards links. Remember seeing a post about it sometime ago although might have been for 10G links.

Jay

eyoung · Post by **eyoung** » Fri Nov 16, 2007 12:44 pm

I followed the above link for changing the polling from 5 minutes to 1 minute and downloaded this file : PollingIntervalsPatchv8.zip. I followed the directions associated with it and my graphs stopped and I was getting error when running poller.php. I changed back to my original files and my graphs started again and poller.php was working.

I will try to change the files again and I will take note of the error message that poller.php is returning.

eyoung · Post by **eyoung** » Mon Nov 19, 2007 12:18 pm

I am using Cacti 0.8.7 ( on Windows 2003), do I need to install any patches or is 1 minute polling supported? From what I have seen in the forum I am not sure that I need to install any patches. However I cannot find any documentation to configure my cacti to poll every minute. Can anyone point me in the right direction for the doc's?

Thanks

BSOD2600 · Post by **BSOD2600** » Mon Nov 19, 2007 4:06 pm

Upgrade to 0.8.7a. Then in cacti's settings there are options for the polling and cron intervals.

eyoung · Post by **eyoung** » Tue Nov 20, 2007 7:36 pm

I have upgraded tp 0.8.7a and I have selected the polling option for every minute, but my graphs still do not show properly. This is a 1GB circuit and I have not been able to have it report in range of 300 to 800MB that it should be reporting, and graphing every minute. When I run poller.php from the command line sometimes I get no response (no error either) and sometimes it continues to poll every minute and out of the five polls it will only show Datasources:2 once. The other times it reports Datasources:0.

BSOD2600 · Post by **BSOD2600** » Tue Nov 20, 2007 8:40 pm

Have you verified your device actually supports 64bit counters? snmpwalk ifHCInOctets for that device.

eyoung · Post by **eyoung** » Wed Nov 21, 2007 10:03 am

Inside of the graph management screen my graph looks different then the graph I placed in the previous message. The above picture was located from my graphs page and the lower picture is from graph management.

BSOD2600 · Post by **BSOD2600** » Wed Nov 21, 2007 1:40 pm

The same underlaying issues of the counters wrapping is present. snmpwalk your device as I stated before.

Post by **TheWitness** » Wed Nov 21, 2007 2:10 pm

If you poll every minute, the data template must have a step of 60. If it has a step of 300, it will only be polled every 5 minutes. That's the way it works.

Now, if you change the step to 60, you will have to discard your old RRDfiles. That is also just the way RRDtool works.

TheWitness

Graph showing wrong badwidth metrics

Graph showing wrong badwidth metrics

Who is online