The article describes the LACP message received on the peer QFX10002 when a PECHIP HMC interrupt was generated on a local QFX10002 device.
Sep 19 04:15:09 QFX2 kernel: %KERN-3: xe-0/0/54:1: received pdu - length mismatch for lacp : len 456, pdu 124
Topology:
QFX1 ------- QFX2
The HMC error was generated on QFX1 and the LACP log message was seen on QFX2.
2023-09-19T03:42:52.396988+00:00 localhost dcpfe[6877]: pechip_cmerror_set_error:3113: Level: Minor, cmerror_code: 0x2101a1 (id=417), recover_err: 0 (counter: 0), fh_msg: 0x0
2023-09-19T03:45:50.718096+00:00 localhost dcpfe[6877]: pechip_cmerror_set_error:3113: Level: Minor, cmerror_code: 0x2101a1 (id=417), recover_err: 0 (counter: 0), fh_msg: 0x0
user@QFX1> request pfe execute target fpc0 command "show pechip interrupt" | no-more
DEV 3
pe.hmcif.link[2].intr.status
hmc_err(0) : 1
pe.hmcif.link[3].intr.status
As defined in RFC, the maximum size of the allowed LACP frame is 128 bytes, including the FCS that is of additional 4 bytes. Hence, a PDU of size 124 is the maximum allowed. In the above scenario, QFX2 received LACP frames from QFX1 greater than 124 bytes and thus the log message is expected to indicate the receipt of such packets and will not accept such frames. These are not error logs but informatory messages regarding a corrupted packet received on its interface.HMC provides for packet buffering and these HMC errors on QFX1 causes the corruption of these LACP packets egressing the device. These errors may affect traffic and have an impact on the traffic forwarding.To recover from this, we would need to do a graceful hypervisor reboot or power drain of the device. If the errors still persist, open a JTAC case.