PCM+ 2.1 measuring super speeds.

[Anders_35]

After upgrading PCM to 2.1 we started getting some strange traffic measurement. The best we have so far is 7264% :)

I have installed update 6, and I have /deleted/ and rediscovered my switches.
So, it should have updated the port speed settings now, shouldn't it?

The ports are auto-neg, some running at 100mbps, some at 1gpbs.

My gut feeling says there's something about the auto-neg that's messing it up (there is always a problem with auto-neg...), but apart from that I don't have a clue as to what is going on.

[Mohieddin Kharnoub]

Hi

I think this may help :
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1045994

Good Luck !!!

[Anders_35]

Thanks, but I've already tried deleting the devices in question. Unfortunately, I still have the same problem.

It also seems as if the problem is not constant. Most of the time it's reporting correctly for the same ports, as far as I can see.

Cheers,
Anders :)

[Anders_35]

After rebooting the PCM server, the problem hasn't occured, so at least we are up to 24hrs in the clear. We'll leave it at that, unless the problem shows up again...

[Mohieddin Kharnoub]

So what you have done is:

Deleting the Devices, then rediscoverd again.
Then after the Network Map discovered, you rebooted the PCM server , and this solves your problem ...

[Anders_35]

Ouch.... there is a message missing here... I closed the thread, thinking it was OK, but an hour ago I reopened.
I thought I had posted some new text as well:


I got another utilization message after all, just after I though it was all right.

The port in question was reported as having 5215% utilization, however the load on that port has not been over 1% for several hours.
It's a port on a 3424, auto-neg to 100-full.

[Steve Britt]

Anders,

Are the ports that you see this behavior on (using PCM 2.1 autoupdate 6) consistently the same ports? If so, which device types and firmware revisions do they correspond to? Do you see it more frequently on high-speed ports? Does it happen more frequently on busier ports than those that are not busy? On a port like the 100Mbps port you mentioned last, does it happen at a fairly regular interval (roughly every 24 hours from the sound of that last posting)?

Are you sampling or polling traffic data on these ports? Whichever means of data acquisition you're using now, could you choose a port that you see the problem recur on and try the other method of acquisition to see if the problem persists (e.g. if you're sampling now try polling, and vice versa)? I'm going to just guesstimate that you're using sampling as that's where most of these issues seem to crop up... I've not seen anything like this in our lab environment since AU6 was released, but I'd sure love to reproduce and fix it. Any further detail you could provide, along the lines of the preceding questions, that would help us isolate and reproduce the problem would be invaluable.

Regards,

SVB

[Steve Britt]

Anders,

Sorry - forgot to ask if utilization is the only fishy metric you're seeing on these ports or if frame, broadcast, or multicast rates also spike at the same time. Not that they're necessarily crossing any thresholds like utilization but are the shape of their Top 5 graphs the same as that of utilization (with an increase in the same minute as the erroneous utilization) or do they stay smooth, as it sounds like you would expect them to? The answer to this question will help narrow the list of possible causes ...

Thanks,

SVB

[Anders_35]

Thanks, Steve.

>Are the ports [..] on consistently the
>same ports?

Yes, but we only have a small number of ports in user on these switches (about 25 in total), and I've now seen this on half of them.
Today I saw two ports that I haven't seen before, so it could be a general problem.

The switches are 3424 (J4905A) and 3448's (J4906A) running M.08.87.

I've seen on both 100mb and 1 gig edge ports, not on our 10 gig ports or other switch-to-switch links.


>Does it happen more frequently on busier ports

No. In fact, it's quite the opposite. I've only seen it on ports that have very low traffic volume (<1% most of the time).
Some ports are quite busy most of the day, but I have not seen a single error for any of those.

The errors occur at seemingly random intervals througout the day.
Load doesn't seem to affect it, I tried hammering some of the the ports with traffic for a long time yesterday, without getting anything.

We are sampling, but I'll try the polling like you suggested.
I'll also add some 2626-switches to traffic monitor, to see if the problem pops up there, too.

Utilization is the only metric that goes awry, AFAIK. A typical Top 5 graph is attached.

Thanks again for helping out,
Anders :)

[Hector Manzo]

Anders,

I work in the ProCurve support group and would like to work with you directly on the traffic problem you are experiencing.

I've tried duplicating this problem with multiple versions of switch software but have been unsuccessful.

Could you please provide me your email address and we can start working directly on this issue?

Thanks,
Hector