RRD miscalculates interfaces pushing more than 100Mb/s
Description
RRD graphs genereated by opennms 1.2.5 miscalculate an interface once it exceeds 100Mb/s and show the usage are much lower than it actually is.
Environment
Operating System: Linux
Platform: PC
Acceptance / Success Criteria
None
Attachments
2
Lucidchart Diagrams
Activity
Show:
DJ Gregor October 3, 2008 at 8:53 PM
This is probably a 32 bit counter wrapping problem, not an OpenNMS problem.
If you enable SNMPv2 collection on the problematic hosts, you should be able to get 64 bit high-capacity (HC) counters and you should see a new graph using the high capacity counters. In current (1.5.x) versions of OpenNMS, the graph for the HC counters should automatically suppress the non-HC counter graph.
Darrell Hyde January 5, 2006 at 2:08 PM
Created an attachment (id=289) MRTG graph of same OC3 interface
This is an MRTG graph of the interface from roughly the same time. It correctly calculates the link usage.
Darrell Hyde January 5, 2006 at 2:06 PM
Created an attachment (id=288) RRD Graph of OC3 interface
This graph shows the OC3 interface on which I'm experiencing this problem. The graph shows low usage between 1100 and 1800, when in actuality the interface was nearly maxed. Around 1800 traffic across the interface dropped to less than 100 Mb/s and the graph began again reporting correctly.
RRD graphs genereated by opennms 1.2.5 miscalculate an interface once it exceeds
100Mb/s and show the usage are much lower than it actually is.