On Junos PTX10008 and PTX10016 platforms, when the speed configuration on the interface is not matching with the speed of the optics present in the port, it causes memory corruption because of this FPC will crash and restart. Traffic loss till the FPC restarts after the ukern core.
FPC crash on PTX 10008 and PTX10016 platforms.
Following logs can be observed:
ppmd[30896]: PPMD_PFE_SHUTDWN: PPMD: Connection Shutdown/Closed with PFE: fpc2
Coredump could be seen:
root@router> show system core-dumps no-forwarding
-rw-r--r-- 1 root wheel 198763695 /var/tmp/fpc-2.J-UKERN.core.0.tgz
The following logs can also be seen:
t6e_qsfp_check_misconfig:T6E(5/0) Mismatch port 7 speed 100G optics 4X10G SR t6e_tvp_qsfp_optic_periodic T6E(5/0): Invalid Combo of Port Speed and plugged-in Optics Type, not initializing the optics: 7 t6e_tvp_qsfp_ifd_link_down T6E(5/0): Disable the LASER, xcvr:7
Memory corruption happens because internal error handling was not done properly when there is a speed configuration mismatch between interface and optics. The issue is also seen when an optic is inserted into a port that does not support the optic speed, even though the port speed is correctly configured with respect to the optic. Example: If a QSFP+-4X10G-SR optic is inserted in a port that only supports 100G If there is a mismatch in the speed configuration with respect interface supported speed level, or when an optic is inserted into a port that does not support the optic speed, even though the port speed is correctly configured with respect to the optic, the logs below will be seen: t6e_qsfp_check_misconfig:T6E(5/0) Mismatch port 7 speed 100G optics 4X10G SR t6e_tvp_qsfp_optic_periodic T6E(5/0): Invalid Combo of Port Speed and plugged-in Optics Type, not initializing the optics: 7 t6e_tvp_qsfp_ifd_link_down T6E(5/0): Disable the LASER, xcvr:7 The workaround is: Fix all the speed configuration mismatches. Set the appropriate speed with respect to optics. Ensure that optics match the supported speed(s) of the port that they are inserted in Permanent fix is via PR1734703: https://prsearch.juniper.net/problemreport/PR1734703