team interface packet counter

[OKyHb]

After switching to team driver we have problem with packet counters statistics. Like this:

2: enp3s0f0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master team0 state UP mode DEFAULT qlen 1000
    RX: bytes  packets  errors  dropped overrun mcast
    4376837523188 6318951869 0       0       0       6020187
    TX: bytes  packets  errors  dropped carrier collsns
    4666802536740 3362998670 0       0       0       0

4: enp3s0f1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master team0 state UP mode DEFAULT qlen 1000
    RX: bytes  packets  errors  dropped overrun mcast
    4397082118769 6206241089 0       0       0       13589800
    TX: bytes  packets  errors  dropped carrier collsns
    15371589631976 11724374918 0       0       0       0

9: team0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT
    RX: bytes  packets  errors  dropped overrun mcast
    8418024056841 8013411841 0       4       0       19609988
    TX: bytes  packets  errors  dropped carrier collsns
    19745880035170 9675699255 0       8       0       0

The total number of bytes on the physical interfaces is almost the same as for team0. But number of packets is very different.

What is the cause of the difference?

[gf-mse]

(speculating) different packet size, reassembling the packets, etc

ps. we have seen a 10-fold ( ~500T / 50T ) difference in the number of bytes as well ( accumulated over ~2.4 years uptime with teamd 1.23 ) -- but this, of course, is not necessarily related

however, frankly, I would love to see some possible explanations -- just to get an idea for where this might be coming from