DFM 3.0.2 curious HighDiscardRate on cat2950

Unanswered Question
Jun 27th, 2008

there are a few reports of curious HighDiscardRate events on the forum and a customer is seeing this event on most of his WS-C2950G-24-EI. The event is appearing in different time ranges on the 2 uplink ports of the device.

Currently my assumption is that it is caused by a wrap around of one of the counters that is used to calculate the discard rate.

I found some interesting facts:

- the InputPacketDiscardRate and the InputPacketRate do have bogus values of 1.7xxxE7

- I found 197 entries of this kind in the FH database for 'InputPacketDiscardRate' but only 4 different numeric values (from which 169 entries do have the same value)

- I do not find any curious OutputPacketDiscardRate but do also see a wrap around of the counters in the MIB (which could be a random issue)

I assume DFM still uses the 32-bit counter instead of the 64-bit counters for the gig Interfaces and does not have any mechanism to detect the wrap around of the counter.

I added the active and cleard DFM events as well as an snmpwalk of the ifTable (forgot to walk the ifMib for the high capacity counters...)

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Joe Clarke Sat, 06/28/2008 - 16:51

This bogus high utilization (or high discard rate) is a relatively common problem. Some causes have been device bugs where 64-bit counters wrap at 32-bit limits. Others seem to be obvious bugs with DFM, but EMC/Smarts is still investigating. The best thing to get for this is a sniffer trace showing at least two polling cycles when this bogus event occurs. Once we rule out a device issue, then you can talk to the TAC to gather other data for EMC.

Martin Ermel Mon, 06/30/2008 - 06:03

I will ask the customer to get the trace. But as there is no schema visible I assume it will take a while to get the requested information....


This Discussion