Strange "Drops Tx" Behaviour

[zmichl]

Recently I spent some time on investigating into the "Drops Tx" counters on my 4512zl switches. If I do an "sho int" on the switch, there are constant rates of "Drops Tx" on alomost every connected port. All the other error counters like collisions, errors rx are always 0.
Based on this behaviour, I thought there have to be many mismatches between speed- and duplex-settings of the ports and the connected devices. But even after double-checking the correct settings on both sides the drops tx-counter wont stay at 0.
After some time I figured out the following behaviour: If I do an "clear stat all" at cli the counters are all resetted to 0. If I stay logged onto the cli (even for hours) and do repeated "sho int", the counters will stay at 0. But as soon as I log off and on again, the counters will show drops tx again. The strange thing about it is, that the drops tx counter is always at the exact same value for every port as soon as I logon again. Even after repeated logons, the counters will stay at this value.

Does anyone know something about this behaviour, or is experiencing similar problems?

P.S. This thread has been moved from Switches, Hubs, Modems (Legacy ITRC forum) to ProCurve / ProVision-Based. - Hp Forum Moderator

[Trevor Commulynx]

DO you have any clusters running Mulitcast NLB? are you running IGMP? and have you checked a packet capture on one of the ports?

[zmichl]

There is no NLB in the network. But I found this (http://forums13.itrc.hp.com/service/forums/questionanswer.do?threadId=1411018) thread which seems to be the same problem...so it seems to be some kind of firmware bug.

[Trevor Commulynx]

So you do not have any multicast running on the network?

[zmichl]

Indeed there are any multicasting applications on the network. In fact, the switches itself multicasts packtes when running vrrp. But does this influence the Drops Tx? I've done "clear stat glob" and the counters stay at 0 since i did it.

[Ardon]

Hi,

Port Counters can be reset per session indeed but once logged out and back in again you will see the absolute numbers again. This is per design and also documented.

Drops TX by the way indicate congestion.

All you ports show the exact same number you say and stay the same every time you log on. Sounds like there was a time when the network was flooded for whatever reason and which is not happening anymore. Hard to tell what happened in the past based on the info given.