When 100G interfaces are configured on any port, the traffic fails for packet sizes larger than 86 bytes after an interface reconfiguration that involves changing the port speed from 100G to 400G and then back to 100G. As a result, the 100G interface is unable to send packets exceeding 86 bytes to its peer, which lead to traffic disruption.
This issue might be seen if the following conditions are met:
* On MX304 platforms
* Switching from 400G to any other speed ( ex: 100G)
Here are the snippets from a lab node for reference:
PIC 0 REV 36 750-122718 XXXXXX MRATE LMIC 16x100G/4x400G
Xcvr 0 REV 01 740-058732 YYYYYYYY QSFP-100GBASE-LR4 --- 100G QSFP model
Xcvr 14 REV 01 740-096176 ZZZZZZZZ QSFP56-DD-400GBASE-LR4-10 --- 400G QSFP model
Sep 8 14:53:49 router picd: PICD_MODULE_ALARM: port-0/0/0: Transceiver does not support speed mismatch alarm cleared
Sep 8 14:53:49 router picd: PICD_MODULE_ALARM: port-0/1/0: Transceiver does not support speed mismatch alarm cleared
root@router-re0> show system rollback compare 24 23
[edit chassis fpc 0 pic 0 port 0]
- speed 400g;
+ speed 100g;
[edit chassis fpc 0 pic 1 port 0]
root@router-re0> show system rollback compare 23 22
[edit interfaces]
+ et-0/0/0 {
+ description "reserved: backbone: -> et-0-1-11.router2, 100GE, x-connect";
+ }
+ et-0/1/0 {
+ description "reserved: backbone: -> et-0-2-5.router3, 100GE, x-connect";
root@router-re0> ping 192.168.1..133 source 192.168.1..134 size 86 rapid
Sep 11 09:23:23
PING 192.168.1..133 (192.168.1..133): 86 data bytes
!!!!!
--- 192.168.1..133 ping statistics ---
5 packets transmitted, 5 packets received, 0% packet loss
round-trip min/avg/max/stddev = 0.436/0.515/0.707/0.098 ms
root@router-re0> ping 192.168.1..133 source 192.168.1..134 size 87 rapid
Sep 11 09:23:28
PING 192.168.1..133 (192.168.1..133): 87 data bytes
....
5 packets transmitted, 0 packets received, 100% packet loss
root@router-re0> show interfaces terse | match ae9
Sep 11 08:25:14
et-0/0/0.0 up up aenet --> ae9.0
ae9 up up
ae9.0 up up inet 192.168.2..206/30
root@router-re0> ping 192.168.2..207 source 192.168.2..206 rapid
Sep 11 08:25:43
PING 192.168.2..207 (192.168.2..207): 56 data bytes
--- 192.168.2..207 ping statistics ---
round-trip min/avg/max/stddev = 0.516/0.545/0.647/0.051 ms
{master}
root@router-re0> ping 192.168.2..207 source 192.168.2..206 rapid size 87
Sep 11 08:25:50
PING 192.168.2..207 (192.168.2..207): 87 data bytes
The issue occurs on ports that were initially configured at 400G and later changed to a different port speed. Due to a bug in the third-party Marvell SDK, switching from 400G to any other speed mode causes packet truncation for packets exceeding 86 bytes.
Workaround:
Reset the affected PIC via request chassis pic pic-slot 0 fpc-slot 0 offline/online
However, 22.x does not support PIC reset, and FPC 0 reset via request chassis fpc slot 0 restart is needed
The issue is fixed in Junos 23.4 onwards