ProCurve 2530-48G - false traffic counters

[Anfanger]

I experience false traffic counters on ProCurve switches:

-HP J9853A 2530-48G-PoE+-2SFP+

-Software revision YA.16.01.0012

E.g. there is Total Rx (bps) : 5,071,064 & Total Tx (bps) : 1,940,464 on the uplink port of the switch while on the uplink switch has different values: Total Rx (bps) : 576 Total Tx (bps) : 88,032. The second values seem to be correct because of number of flowing packets, which is the same on both switches (around 70pkts/s).

I  also performed port mirroring and Wireshark sniffing. Results - I was receiving ~320kbps from mirrored port, at the same time this port was reporting ~5/2Mbps (Rx/Tx), so I can assume traffic counters of the switch are false.
I tried to look for mentions about this topic through the Internet, but no effects. Do you have any idea what is the reason of such difference in traffic counters? Maybe some firmware bug?

I have this problem on many machines, also with older firmwares.

[parnassus]

Traffic rates values you reported (taken from a show interface interfacename CLI Command), since are rates, represent, IIRC, a 5 minutes weighted average...would be interesting to repeat your test after committing, on both end, a concurrent clear statistics CLI Command against the (physical/logical) port you're diagnosing before arguing about potential issues regarding traffic rates mismatch between Switches interconnected through an uplink port.

In other terms...since you're looking at traffic rates it's important to set the initial time at which data is collected to produce those rates...otherwise results you're looking are true (data collection happened and a data rate is calculated) but non consistent with what you want to measure in a specific amount of time for a specific port.

Another way to check is, on both ends, to use the show interface port-utilization CLI Command to view real time rates (Tx and Rx in kbps, pps and % Utilization of total port bandwidth).

[Anfanger]

I cleared counters before tests.

I have already tried show interface port-utilization command, the output was similar.

On the other hand:

-the difference is too big  (5M vs 300K) to be a measurment error while the traffic is relatively constant (still 300K, mostly DNS, ARP, DHCP querries and so on).

-port mirrorring with wireshark gives me also 5min average traffic stats.

I am sure counters are wrong, now I need to troubleshoot the reason.

[parnassus]

Would be intersting to know what Switch you have at other end?
Have you already verified if there are newer software versions (YA/YB.16.02 Branch has 16.02.0021, YA/YB.16.03 Branch has 16.03.0005 and YA/YB.16.04 Branch has 16.04.0008) that solve a similar issue (if any) since your 16.01 stops actually exactly at 16.01.0012?

[Anfanger]

At the other end I have HP J9851A Switch 5412Rzl2 with software revision KB.16.02.0009.

I will read release notes of newer versions and make some tests, then I will let you know if it helps.

[Anfanger · Accepted solution]

YA_16_04_0008 resolved the issue.

[parnassus]

What CR exactly fixed your issue? ...or was just the generic update to latest software release of YA.16.04.0008 from YA.16.01.0012 that unexpectedly (and so surprisingly) solved your problem?

I admit I wasn't able to find a specific CR Fix related to wrong counters or wrong rates on YA.16.04.0008 Release Notes...