I have come across a problem after recently installing the Dude 3.0rc2.

After enabling SNMP on an element, the Dude consumed 2.1GBytes of traffic in a single month polling just two devices. Now this element does have quite a few interfaces, but the "Polling" of the element was set to once every 10 minutes.

Using Wireshark, I found the Dude was snmpwalking the MIB about every 2 seconds.

There does not seem any way to control the frequency the Dude was collecting SNMP information.

Also the Dude does a full walk of the interfaces MIB (amoung others) and for this device it is quite large - there are some 2500 interfaces as the device as an FTTH OLT. The element also has some 2000 FDB entries and several hundred ARP entries.

There are several aspects to this that I would like addressed, or answered:

a) the SNMP polling should be either associated with the Element Polling setting (which is not the case) or have a seperate value that can be set, or even to turn off the SNMP statistics collection.

b) The SNMP walk should have some mechanism to control what is walked and what is collected.

c) If I enable the SNMP my SNMP based probes work, but a lot of useless information is collected. If I turn SNMP off, then I lose the probes which I have built for charting and alarms, but my bandwidth is significantly reduced

Can you post a couple of your probes to give us an idea what they look like?

I have found where the SNP data is collected.

It is under charts-> data sources.

Whereever an interface has been found through SNMP polling, it has been added as builtin.

However clicking on tha data source which is builtin does not permit the polling frequency from being changed. I would like to be able to change the polling frequency to say 5 minutes

We also have the same problem for pooling devices with really big routing table.
For example, we have Cisco 6500 Series Switch with 4 full BGP view.
Then dude is trying to get some data from device, processor of the Cisco is getting up to 95% because of this. He is trying to get complete routing table and this is ~1.000.000 of routes.

We would propose to make this type of pooling as option.

Hello,

is there a solution to manage the time between the snmp polls? We have also a Cisco 6500 router where a complete snmpwalk takes around 2minutes! Is there a possibility to manage only polling the interfaces?

Best regards,
Patrick

Oh, I thought this was a probe and not the default collection of snmp stats. I had never looked to see if that default collection was excessive, it sounds like it is. This is a question for Normis.

the documentation of dude is unfortunately not so detailed to read such spezific details.
It would be great to calibrate the snmp queries like the polling intervall. Additionally it would be a great feature to enable/disable specific query types (such as: only Interface, or IP, or Route and so on!) to avoid a complete snmpwalk all the time.

I think this feature would be also interesting for other users of Dude

Best regards,
Patrick

Sorry for intruding but we are discussing a similar topic in a different thread..
http://forum.mikrotik.com/viewtopic.php?f=8&t=50492

Gsandul has suggested using snmp view to prevent the Dude from polling too much data (like - the internet routing table) from Cisco hardware which can kill the CPU. We also have a couple of 7600's (which in fact are 6500's with a different marketing department) and I am fighting with CPU / excessive polling, too.

I will quote his answer here:

This is done using snmp viev
sample cisco config could be

Code: Select all

snmp-server view noifread iso included
snmp-server view noifread interfaces excluded
snmp-server view noifread iproute excluded

snmp-server community TheSuperSecret view noifread ro 1

also you should read this
http://www.cisco.com/en/US/tech/tk648/t ... 48e6.shtml

Andreas

ok, thanks. I will switch to the other article

We also have the same problem for pooling devices with really big routing table.
For example, we have Cisco 6500 Series Switch with 4 full BGP view.
Then dude is trying to get some data from device, processor of the Cisco is getting up to 95% because of this. He is trying to get complete routing table and this is ~1.000.000 of routes.

We would propose to make this type of pooling as option.

+1

We have the same problem with snmpd & The Dude.
And crash with memory violation error.