Y Axis scaling past 100%

We have seen a couple of similar issues with spikes in Tx/Rx charts, and we know there is a bug in the calculation, alhough we were not able to find it yet. If you want to help us troubleshooting this issue, please open a support ticket and reference this thread.

If you divide two numbers and the result is > 100% of the theoretical interface speed then something is wrong.  I'm pretty sure you don't need for me to open a ticket to help you with basic division.

Good point However, we are not able to reproduce the issue in our environment, so we can only work with data and logs we get from our customers. The more data and logs we have, the faster we can find the issue.

An update to this thread:

We have identified the cause of this problem. Spikes occur when our software polls a device more frequently than the device updates its internal SNMP tables. Some devices update their internal tables faster than once a second while others update slower than once every 10 seconds by default. However, users can customize their device's internal update times so we can't predict every devices' internal update cycle. Complicating this is the fact that a device may slow down its internal updates when its usage gets high.

Addressing the spike means determining why a counter value is the same as it was at the time of the last poll. Is it because no data came through or is it because it was not updated yet. Assuming intent would lead to other bugs. We are currently testing an approach that should address this acceptably but we are going to be running it over many environments over a long period of time to make sure we don't introduce any new issues. If we like the results it will be included in a future release of Toolset.

For now, if you are seeing frequent spikes you should make your polling interval longer. If you are seeing rare spikes we have verified that the values on each side of the spike are correct and the spike is due to a polled value that was not internally updated since the last poll. It is also possible that a device is just reporting a spike.

Here are some links to topics that are related to this thread:

www.fineconnection.com/How_to_set_the_net-snmp_agent_update_or_counter_refresh_interval

www.paessler.com/.../113

www.loriotpro.com/.../how_to_script-sci.php

--HTH

Steve

I'm aware that the device I'm polling can return identical values.  The manufacturer knows about this snmp update issue and is working on addressing it.

Just because the device returns the same value twice shouldn't mean that your tool freaks out and graphs it as a spike.  Just graph it as 0.  0 is what the difference is between the two polling intervals.

I'm glad you identified the bug and are working on a fix.  Please post when it's ready to be released.

Tom