This article discusses issues related to the QFX10008 and QFX10016 Junos platforms, specifically focusing on data parity errors or hardware failures in the line card memory, which can cause the FPC to enter a faulty state.

On QFX10008 and QFX10016 Junos platforms due to data parity or related hardware failure in linecard memory which caused FPC to go in bad state. It took over 2-3 minutes for system to identify this state and switchover traffic to other working interfaces.

This caused traffic black holing for 2-3min. Traffic restored once system identifies the state.

This issue can be identified with "Err Detect Register" and "Data Cache Parity Error" error logs.

Logs: 

Mar 1 14:28:12 hostname-re0 fpc2 fpc2 kernel: Caused by (from MCSR=20080000): Data Cache Parity Error

<>

Mar 1 14:30:30 hostname-re0 fpc2 fpc2 dcpfe: PANIC PANIC PANIC PANIC PANIC PANIC

<>

Mar 1 14:30:38 hostname-re0 chassisd[7371]: CHASSISD_IPC_CONNECTION_DROPPED: Dropped IPC connection for FPC 2

Juniper Engineering is developing a software fix to immediately drop the IPC connection for an FPC when a "Data Cache Parity Error" is logged. This will eliminate the current delay of 2 minutes and 26 seconds observed in such cases.

The fix is available through PR1795339. 

As a temporary workaround, inline BFD can be configured which can timeout much quickly in such failures resulting in traffic switchovers.

2024-03-20 : Article Created

2025-01-22 : Article edited to reflect additional data from the external PR referenced.