9

We have put some 3750Gs in a stack for our DC access switches. We are concerned that these won't be able to handler higher data loads as traffic increases due to their limited buffers. What statistic/s and/or specific SNMP OID's should we be monitoring to find out how these buffers are handling the load?

FYI We are using PRTG as our tool of choice.

Improve this question
2
  • 1
    FYI there's nothing wrong with approving an edit to a question. :-)
    – John Jensen
    Commented May 26, 2013 at 18:04
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can post and accept your own answer.
    – Ron Maupin ♦
    Commented Jan 3, 2021 at 5:53

2 Answers 2

Reset to default
9

As far as I know data you want is not available on SNMP.

3750#sh mls qos interface FastEthernet0/1 statistics 
...
  output queues dropped: 
queue: threshold1 threshold2 threshold3
-----------------------------------------
 queue 0:            0            0            0 
 queue 1:       100989            0            0 
 queue 2:            0            0            0 
 queue 3:            0            0            0 
...

What you'd like to know is ifIndex, queue, threshold and drop counter. I'm not aware of populated MIB/OID where these values can be polled from.

Like John Jensen explained, outDiscard is only thing you can get, but it aggregates all of this, so you won't know if it's BE, AF, EF, NC or what, which is dropping. You probably wouldn't care about BE drops, but you'd care about EF drops.

There are two OIDs where these aggregate egress drops are stored, if your ifIndex is 10001, you'd find them here (symbolic and numeric presentation):

IF-MIB::ifOutDiscards.10001
.1.3.6.1.2.1.2.2.1.19.10001
EtherLike-MIB::dot3StatsDeferredTransmissions.10001
.1.3.6.1.2.1.10.7.2.1.7.10001

3750/3560 are not very good switches for application which can microburst, i.e. if your egress is 1GE and ingresses are 1GE too, two very low average rate ingress ports can easily congest the egress port, causing drops. To maximize the available buffers (and minimize microburst drops) follow this document.

Improve this answer
0
1

Assuming you're wanting to monitor this data via SNMP (I'm not familiar with PRTG), your best bet is to monitor for:

  • input queue drops
  • output queue drops
  • inDiscards
  • outDiscards

Keep in mind that if you're monitoring input queue drops, you'll see these happen across multiple subsets of ports, because (IIRC - someone correct me if I'm wrong) these map to the port ASIC's and the buffers are shared across port groupings (ASICs).

Here's a link to a Cisco doc that has some other good info:

http://www.cisco.com/en/US/tech/tk648/tk362/technologies_q_and_a_item09186a00800b69ac.shtml

Improve this answer
3
  • Yeah SNMP is the go. Anyone familiar with PRTG know any good SNMP templates?
    – Tim
    Commented May 26, 2013 at 5:41
  • You may want to amend your question and ask for PRTG templates if that's ultimately what you're after.
    – John Jensen
    Commented May 26, 2013 at 5:51
  • Nah, I can get the SNMP values ok, was just a side though to see if anyone already had them to save the the work.
    – Tim
    Commented May 26, 2013 at 5:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.