After Junos upgrade to 23.4R2-S7.4, SNMP counters for logical units is inconsistent.

Delta for ae0.0 is way above the ones for ae0:

17:52:18.844 user@device> show interfaces ae0

17:52:18.913 Physical interface: ae0, Enabled, Physical link is Up

17:52:18.914 Interface index: 640, SNMP ifIndex: 656

17:52:18.914 Description: xxxxx [200G] {ae0}

17:52:18.916 Link-level type: Ethernet, MTU: 9216, Speed: 200Gbps, BPDU Error: None, Ethernet-Switching Error: None, MAC-REWRITE Error: None, Loopback: Disabled,

17:52:18.917 Source filtering: Disabled, Flow control: Disabled, Minimum links needed: 1, Minimum bandwidth needed: 1bps

17:52:18.917 Device flags : Present Running

17:52:18.917 Interface flags: SNMP-Traps Internal: 0x4000

17:52:18.918 Current address: aa:aa:aa:aa:aa:aa, Hardware address: aa:aa:aa:aa:aa:aa

17:52:18.918 Last flapped : 2026-03-20 06:51:18 GMT (1w4d 14:01 ago)

17:52:18.919 Input rate  : 7639672536 bps (1017344 pps)

17:52:18.925 Output rate : 2201729168 bps (458867 pps)

17:52:18.925 

17:52:18.926 Logical interface ae0.0 (Index 544) (SNMP ifIndex 657)

17:52:18.926  Description: xxxxx [200G] {device:ae0.0}

17:52:18.927  Flags: Up SNMP-Traps 0x4004000 Encapsulation: ENET2

17:52:18.927  Statistics  Packets  pps   Bytes   bps

17:52:18.927  Bundle:

17:52:18.928   Input : 2101629256259 1033251 1811841692591904 7801858456

17:52:18.928   Output: 1556498485336  450004 497337937403606 2087866056

17:53:31.714 user@device> show snmp mib get ifHCInOctets.656 | refresh 10

17:53:31.715 Mar 31 20:53:31

17:53:31.729 ---(refreshed at 2026-03-31 20:53:31 GMT)---

17:53:31.772 ifHCInOctets.656 = 889816073137469

17:53:41.777 ---(refreshed at 2026-03-31 20:53:41 GMT)---

17:53:41.810 ifHCInOctets.656 = 889826978328671

17:53:43.079 ---(*more 100%)---[abort]

17:53:43.080 

17:53:43.080 {master:0}

17:54:05.634 user@device> 889826978328671-889816073137469=10.905191202

There are two ways of fetching IFL stats, one via a user space daemon called 'pfed' and the other through the kernel. if the platform has a sysctl variable called 'bulkget' set to true, IFL stats will be fetched via PFED, else through the kernel. Unfortunately, there was a regression introduced where IFL stats can be fetched via the kernel as well as through PFED and we ended up counting stats from both.

The fix is to upgrade to one of the codes mentioned in the PR, there is no workaround. About the S7 release listed, it is for S7.7 not S7.4.

Resolved-in:

evo:26.1R1-EVO evo:26.2R1-EVO junos:23.4R2-S7 junos:23.4R2-S8 junos:23.4X12 junos:23.4X13 junos:23.4X14 junos:23.4X30 junos:23.4X30-D30 junos:24.2R2-S4 junos:24.2X1 junos:24.2X4 junos:24.4R2-S1-C1 junos:24.4R2-S2 junos:24.4R2-S3 junos:24.4R2-S4 junos:24.4X3 junos:24.4X42 junos:25.2R1-S2 junos:25.2R2 junos:26.1R1 junos:26.2R1

https://prsearch.juniper.net/problemreport/PR1914977