I have a strange problem with the Dude.
For some months now the traffic graphs are very weird.
I cannot explain it well so I am attaching 2 screenshots.
In case of a Linux/Windows device with standard Net-SNMP/Windows SNMP the graphs are with Peaks of X seconds/minutes (on a local device it’s every 10 seconds, on an internet device it’s every 5 minutes) even though the real traffic is nowhere near that.
In case of a Mikrotik RouterOS device, then the graphs are perfect.
I am running Dude 4 Beta 3 but I’ve seen this problem since beta 2 (as I said this is going on for some months now)
Do you have any suggestions as to what is causing this behaviour?
Prior to beta2 the graphs where shown correctly.
600kbps is about 0bps, I would be curious what copy or batch job is moving a fairly large file every 30 seconds, 10Mbps. Is the dude collection actually causing that spike? You must be collecting the entire MIB if that is the case… that is what it looks like. Check your probes and make sure they are reasonable.
There are no spikes of traffic whatsoever.
This kind of graphs is recorded for ALL devices (except mikrotik devices).
I have graphs from cacti of the same devices and the traffic is smooth as the second graph I attached.
No I do not collect the entire MIB, and this traffic is not real. It does not exist. Only Dude registers it that way.
Cacti does not show those spikes because they don’t exist.
It’s a default Dude installation with a few devices using SNMP v1 to collect data.
Nothing custom.
What probes have to do with anything?
As far as I am concerned probes do not measure traffic via SNMP which in this case it might is the source of the problem.
As I said, this behavior started after upgrading to beta2 and this behavior is recorded on all kinds of devices except mikrotik routeros.
No one else has this problem?
I have 5-6 dude installations the last 6 months in different datacenters, with different devices and ALL of them measure the traffic in that manner.
Yes that seems totally odd. Sorry I wasn’t trying to point fingers… I meant the dude must be creating that spike by reading too much of the MIB.
Cacti can’t record those spikes since it only polls every 5 minutes. The Dude records every 30 seconds.
Everything is collected by SNMP so I am not sure what you are trying to getting at. Since those spikes are every 30 seconds it must be the Dude collection but the amount collected seems like it would be a bunch of stuff. Therefore I think it is likely one of the things being collected is not configured correctly. Since you say it is default and didn’t happen before it sounds like the default settings in the dude are suspect.
same problem here too. All UBNT devices have problems in their graphs as spikes shown by the OP, but in Cacti seems just fine. RouterOS devices are smooth in Dude although the polling period is the same for all devices.
I am not sure on this, but the Polling interval I think does not apply to the graphs (?)
The graphs collect data every 10 seconds and the Polling interval is 30 seconds (on my installation - which I suppose is the default since I never changed any polling values).
Besides that, on the Polling tab on the Dude Server Configuration window it refers to Service Polling and to Probes.
It does not say anything about graphs (unless of course it’s just a bad interface design and it’s not that clear)
Also if it was a polling interval issue why would the problem only appear on every device except Mikrotik devices (on which the graphs as I posted are just fine)?
And why is it that on Dude 3.x the graphs are working OK with the same settings, and since Dude 4.x the graphs are broken for non-mikrotik devices?
Can anyone from Mikrotik comment on the issue?
It can’t be only me having this problem. I tried on on numerous servers (Win 2003, 2008, XP, 7, 32/64bit) and the behavior is same everywhere!
I forgot to mention that the link label with the real-time traffic indicator (I don’t know how it’s called) shows the real traffic.
So for example it might show 1mbit upload 200kbit download which is real traffic and properly measured, but on the graph it will put something like 10mbit upload and 2mbit download (roughly x10 times the real data).
Sorry Chaos, I thought it was actual traffic. I agree with you it must be some bug of some sort. Just do us a favor and change polling interval to 1 minute to see if the problem changes with the polling interval. I don’t have this problem but I only monitor about 20 servers. I do monitor about 300 network devices and have over 1000 probes with polling set to 1 minute.
Also there is a neat java application that will poll the interface counters ever 1 second if you like called “AsItHappens” you can use it to prove the dude is crazy. I don’t know what is going on here.
I can confirm that this issue is still active today 23.12.2014 for Ubiquity devices as posted by original poster. Increasing or decreasing pooling interval does not affect graphs generated by SNMP. Traffic pulled from Ubiquity devices SNMP looks as periodic spikes.
