Known Limitations

If a packet with unknown inner ether-type is received at the device over an EVPN-MPLS tunnel, then such packet is dropped. PR1564431

Restarting the fabspoked-pfe application for the line card restarts the line card. PR1486023

For a faulty link, after autoheal is triggered and before the action is completed, if fabspoked-fchip restarts, the link stays in fault state. PR1508915

Excess-rate configuration in port schedulers might not be completely honored in certain scenarios. In such scenarios, with explicit excess-rate configuration, the actual excess-rate achieved might still be more in proportion to the configured transmit-rate.. PR1528124

Double-fault scenarios are not handled by link auto-heal feature and fabric links remain down if the Routing Engine switchover is attempted while auto-heal recovery is in progress. PR1529599

When a scheduler-map binding is removed from an interface, then the default scheduler-map is bound to the interface. If default scheduler-map is an oversubscribed scheduler map for the interface, then that map does not be apply to this interface and all the "interface queue" counters for this interface shows statistics as 0. PR1539052

PTX10008: By default IPv6 addressing is configured at the /64 subnet by default irrespective of the subnet configured on the DHCP server side. PR1539839

On all Junos OS Evolved platforms when the next-hop is added or changed to the Packet Forwarding Engine and the same next-hop is also forwarding nexthop of an indirect route, if ingress Packet Forwarding Engine is fast and egress Packet Forwarding Engine is slow (MPC and DPC cards for example or MPC cards with different versions), then this results in packet loss as ingress Packet Forwarding Engine being faster would have seen new FNH and also the indirect change. But egress Packet Forwarding Engine being slower would not have consumed indirect change yet. PR1547432

On Junos OS Evolved PTX10008 platforms, if multiple SIBs are in offline state and GRES is performed immediately, SIBs may be stuck in offline state for sometime. PR1554423

During fabric link bringup, fabspoked-fchip[0-5] app restart, sib offline, or sib online, /re0/fabspoked-fchip[0-5] might not respond to CLI commands (show chassis sibs). This is indicated by the following output on CLI

error: communication failure with /re0/fabspoked-fchip[0-5]/.

The producer app is either down or unresponsive. Run the show system processes node node | grep app command to check if the app is running. Run the show system application app to check the state of the app. If there was a change in primary role recently, wait for the switchover to complete for the app to be online. Run request chassis routing-engine master switch check to check the status of the switchover. If the command is reissued after training or ifs detraining reaches a steady state, the valid output shows up in the CLI. PR1570605

The PTP FPGA is kept in reset during BIOS boot. During Linux boot, the PTP FPGA is taken out of reset and pcie-tree is re-enumerated. Hence you would be seeing the Link-up/down during this sequence. PR1572061

Sometimes during ZTP auto-speed detection, ports in 4x25G mode might not come up if there is difference in FEC across the nodes. PR1572666

Ungraceful FPC restart followed by an immediate master switchover is a double fault scenario. The final state of system is unpredictable. The user has to wait until the system stabilizes after ungraceful FPC restart before triggering events like primary switchover. PR1576549

Double-fault scenarios are not handled by link auto heal feature and fabric links remain down if the Routing Engine switchover is attempted while auto-heal recovery is in progress. PR1578615

During NSR switchover, there is a potential of loss of a small number of packets addressed to the host or Routing Engine. This may cause a LDP sessions to flap since BFD sends only a single keepalive and loss of this keepalive causes the BFD to be declared down. PR1580355

UDP encapsulated MPLS packets with explicit null label received on FTI tunnel gets dropped after UDP decapsulation. After UDP tunnel header decapsulation MPLS payload with explicit null label cannot be forwarded as it requires popping MPLS explicit null label and lookup of MPLS inner payload which is not supported in BT ASIC based products without looping back the MPLS payload for additional lookup. We support only scenarios where we decapsulate the tunnel header and forward the packets based on the exposed MPLS Label. PR1580641

PTX10003-60C or PTX10003-160C interface queue and voq does not report drops when the low priority queue is slightly oversubscribed ( seeing page timeouts). PR1581490

If we offline multiple sibs and halt the primary Routing Engine, the SIBs can be stuck in the offline state for 15 minutes, before it goes to offline state. PR1584712

Change in flow label configuration is handled by deleting and re-adding the Layer 2 VPN route. Therefore, traffic loss is seen during the configuration change. PR1585059

Ungraceful power off or removal of SIBs can cause PCIE errors for SIB devices and the system might report the alarm PCI Uncorrected error on dev 0000:00:03.0.PR1590286

Additional triggers like fabricHub or Fabspoked-fchip daemon restarts while FPC offline or online is in transition, results in traffic loss after the FPC is online. PR1596818

On Junos OS Evolved PTX Series platforms, the firewall prefix-action-stats is not supported, so if a user tries to use the show firewall prefix-action-stats CLI then the respective XML tags <firewall-prefix-action-information> is not be visible. PR1443325