The "match ...fragment" rule matches any fragments not necessarily related to the initial packet, isn't it?
The fragments show in Wireshark as Protocol: IPv4 not UDP and there is no destination port in header
I've modified the classifiers to:
class ip UDP6000
10 match udp 10.170.14.100 10.100.1.80 range 6000 6999 count
class ip UDP7000
10 match udp 10.170.14.100 10.100.1.80 range 7000 7999 count
20 match udp 10.170.14.100 10.100.1.80 fragment count
and to have a comparison also modified the policy to have the same CIR for both port ranges
policy QOS_TEST
10 class ip UDP7000 action pcp 5 action dscp CS5 action cir kbps 1000 cbs 153600 exceed drop
20 class ip UDP6000 action pcp 3 action dscp CS3 action cir kbps 1000 cbs 153600 exceed drop
and the result is even funnier:
The command at source : iperf3 -c 10.100.1.80 -u -b 10M -p 7201
Result on the target:
Accepted connection from 10.170.14.100, port 49447
[ 5] local 10.100.1.80 port 7201 connected to 10.170.14.100 port 57537
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-1.00 sec 216 KBytes 1.77 Mbits/sec 0.023 ms 97/124 (78%)
[ 5] 1.00-2.00 sec 80.0 KBytes 655 Kbits/sec 0.017 ms 143/153 (93%)
[ 5] 2.00-3.00 sec 80.0 KBytes 655 Kbits/sec 0.023 ms 143/153 (93%)
[ 5] 3.00-4.00 sec 80.0 KBytes 655 Kbits/sec 0.016 ms 141/151 (93%)
[ 5] 4.00-5.00 sec 80.0 KBytes 655 Kbits/sec 0.012 ms 144/154 (94%)
[ 5] 5.00-6.00 sec 80.0 KBytes 655 Kbits/sec 0.017 ms 142/152 (93%)
[ 5] 6.00-7.00 sec 80.0 KBytes 655 Kbits/sec 0.013 ms 143/153 (93%)
[ 5] 7.00-8.00 sec 80.0 KBytes 655 Kbits/sec 0.018 ms 142/152 (93%)
[ 5] 8.00-9.00 sec 80.0 KBytes 655 Kbits/sec 0.026 ms 144/154 (94%)
[ 5] 9.00-10.00 sec 80.0 KBytes 655 Kbits/sec 0.053 ms 142/152 (93%)
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-10.04 sec 936 KBytes 764 Kbits/sec 0.053 ms 1381/1498 (92%) receiver
The command at source :: iperf3 -c 10.100.1.80 -u -b 10M -p 6201
Result on the target
Accepted connection from 10.170.14.100, port 51106
[ 5] local 10.100.1.80 port 6201 connected to 10.170.14.100 port 49348
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-1.00 sec 264 KBytes 2.16 Mbits/sec 0.044 ms 92/125 (74%)
[ 5] 1.00-2.00 sec 96.0 KBytes 787 Kbits/sec 0.044 ms 141/153 (92%)
[ 5] 2.00-3.00 sec 104 KBytes 852 Kbits/sec 0.044 ms 140/153 (92%)
[ 5] 3.00-4.00 sec 96.0 KBytes 787 Kbits/sec 0.039 ms 139/151 (92%)
[ 5] 4.00-5.00 sec 88.0 KBytes 721 Kbits/sec 0.043 ms 143/154 (93%)
[ 5] 5.00-6.00 sec 80.0 KBytes 655 Kbits/sec 0.042 ms 142/152 (93%)
[ 5] 6.00-7.00 sec 88.0 KBytes 721 Kbits/sec 0.046 ms 141/152 (93%)
[ 5] 7.00-8.00 sec 72.0 KBytes 590 Kbits/sec 0.049 ms 113/122 (93%)
[ 5] 8.00-9.00 sec 0.00 Bytes 0.00 bits/sec 0.049 ms 0/0 (0%) <------- !!!!
[ 5] 9.00-10.00 sec 0.00 Bytes 0.00 bits/sec 0.049 ms 0/0 (0%) <------- !!!!
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-10.02 sec 888 KBytes 726 Kbits/sec 0.049 ms 1051/1162 (90%) receiver
Counters:
# show policy hitcounts QOS_TEST interface lag51
Statistics for Policy QOS_TEST:
Interface lag51 (out):
Matched Packets Configuration
10 class ip UDP7000 action pcp 5 action dscp CS5 action cir kbps 1000 cbs 153600 exceed drop [ 0 kbps conform ]
1513 10 match udp 10.170.14.100 10.100.1.80 range 7000 7999 count
15160 20 match udp 10.170.14.100 10.100.1.80 fragment count
20 class ip UDP6000 action pcp 3 action dscp CS3 action cir kbps 1000 cbs 153600 exceed drop [ 0 kbps conform ]
1521 10 match udp 10.170.14.100 10.100.1.80 range 6000 6999 count
packet capture on the source looks like this

------------------------------
-- tommyd
------------------------------
Original Message:
Sent: Apr 10, 2024 11:33 PM
From: ariyap
Subject: 6300 QoS plicy and fragmentation
I think you need to explicitly permit fragmentation in your "class ip UDP6000" with something like this
match udp 10.170.14.100 any fragment
------------------------------
If my post was useful accept solution and/or give kudos.
Any opinions expressed here are solely my own and not necessarily that of HPE or Aruba.
------------------------------