Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

Known Issues

 

This section lists the known issues in hardware and software for the QFX Series switches in Junos OS Release 17.4R2.

For the most complete and latest information about known Junos OS problems, use the Juniper Networks online Junos Problem Report Search application.

EVPN

  • In a scaled setup, if MAC move is triggered more than 4 times, the MAC move detection might not be reliable. PR1284315

  • CNH (chained-composite-next-hop) is must for EVPN pure type 5 with VXLAN encapsulation. Without this Packet Forwarding Engine wouldn’t program the tunnel next hop. You have to explicit set it on QFX5110. set routing-options forwarding-table chained-composite-next-hop ingress evpn. QFX10000 it is applied as part of default configuration. user@router> show configuration routing-options forwarding-table | display inheritance defaults. PR1303246

  • In an EVPN collapsed Layer 2 or Layer 3 multihomed gateway topology, when traffic is sent from an IP fabric towards EVPN, some traffic loss is seen. If the number of hosts behind the EVPN gateways is increased, the traffic loss becomes higher. This issue is seen with the QFX10000. PR1311773

  • ARP gets deleted and relearned during the first ARP refresh with EVPN-VXLAN multihomed CE, so traffic drops and recovers for the first ARP refresh. PR1327062

  • On QFX platforms (QFX5K/10K), VTEP's MAC address is not learned in the ethernet switching table though they are present in the EVPN database. PR1371995

Layer 2 Features

  • When an FPC encounters a memory exhaustion condition, the FPC restarts unexpectedly with the PPMAN: failed decoding IDL msg - retval -2 type 5 encode_len 208 length 208 data 0x344ff1b0 message. PR1321117

MPLS

  • LDP to BGP stitching with an eBGP indirect next hop having an implicit null label does not work. It does work when BGP indirect next hop has a real label. As a workaround, perform the following: (1) Ensure the peer advertises a real label by adding another router between the egress and ingress PE devices. (2) Use IBGP, which gets resolved over LDP or RSVP-TE LSPs. This will ensure that the BGP indirect next hop has a real label. PR1254702

  • On optimize timer expiry, when the ted version number match indicates a CSPF has already run for the path, if an optimization has not yet been done with that version, it will be run despite the version number match. (Having a per path optimize-seq-no that is updated with ted seq no only on optimization.) When path-cc-updated is false and CSPF fails for optimization, disable the path just like we do for the ones on avoid colors/invalid ERO, making sure this does not interfere with global repair/local reversion. PR1365653

Platform and Infrastructure

  • While scaling more than 2000 VLAN or IRBs, Layer 3 multicast traffic does not converge to 100 percent and continuous drops are observed after bringing down or up the downstream interface or while an FPC comes online after an FPC restart. PR1161485

  • When per-packet load balancing is removed or deleted, the next-hop index might change. PR1198092

  • On PTX platforms with FPC3, PTX1000 with build-in chassis and QFX10000 platforms, a Flexible PIC Concentrator (FPC) major alarm might be seen if the system detects parity error, and the error messages DLU: ilp memory cache error and DLU: ilp prot1 detected_imem_even error might appear. The alarm might be cleared without intervention. This error may also be accompanied by traffic loss. PR1251154

  • Single-bit and multiple-bit ECC errors are not logged on QFX5110 switches. PR1251917

  • On the QFX10000-12C-DWDM coherent line card, it is possible that sometimes the link flaps when MACsec is enabled on Ethernet interfaces. PR1253703

  • The management process (daemon) might crash if the Openconfig package is installed immediately or within minutes of Network Agent package installation. This is a transient issue and will not impact any functionality. There is no action needed from the user side in response to the crash. As a workaround, install Openconfig before installing Network Agent. PR1265815

  • On QFX5100 switches, static LAG link protection switchover/revert is not working consistently. PR1286471

  • When link protection with the backup port state "down" and LACP are configured, if backup state "down" is removed from the configuration, both ports should be up and the primary port should pass all egress traffic. In some instances, however, traffic might pass through the backup port instead of the primary port. PR1297597

  • Traffic drop occurs on sending traffic over "et" interfaces due to CRC errors. PR1313977

  • Family Ethernet-switching cannot be used when flexible-vlan-tagging is configured. It is unsupported. The behavior is non-deterministic with this configuration and there is a possibility of seeing a dcpfe core file. PR1316236

  • Port 0 of Qfx5100-48t does not come up in a mixed VCF. As a workaround, use the phy diag xe0 dsc command as of now from the BCM shell upon reboot, which brings up the port and stays up continuously until the next reboot. PR1323323

  • The management process (mgd) might panic after modifying aggregated Ethernet interface members under the ethernet-switching vlan stanza. After mgd panic, your remote session is terminated as a result. PR1325736

  • In Streaming Telemetry scenario, if performing "commit full", na-grpd daemon might restart causing disconnection of streaming telemetry. PR1326366

  • On QFX5100 Series platforms, in some cases, the CoS (class of Service) configuration is not properly applied in the Packet Forwarding Engine, leading to an unexpected egress traffic drop on some interfaces. PR1329141

  • On QFX52xx standalone devices with Vxlan configured, user configured Ingress ACL scale limit is 256 terms. PR1331730

  • On QFX5110, the FEC for 100g optics is not being displayed when expected behavior is for FEC to be shown as NONE. On QFX10002 Elit, the FEC for 40g optics is being displayed as NONE when expected behavior is for FEC not to be displayed. On QFX10008 Ultimat, the FEC for 40g optics is being displayed as NONE when expected behavior is for FEC not to be displayed. PR1360948

  • When MCLAG is configured with Force-Up enabled on MCLAG Nodes, LACP admin key should not match with Access/CE device. PR1362346

  • On QFX10000 platform with IRB enabled, traffic might not be forwarded on some of the child members when the member link of the aggregated Ethernet is added or deleted. PR1362653

Routing Protocols

  • For single-hop eBGP session, upon interface down event, do not do GR helper logic. In problem state Peer: 8.3.0.2 AS 100 Local: 8.3.0.1 AS 101 Group: EBGP Routing-Instance: master Forwarding routing-instance: master Type: External State: Active Flags: <> Last State: Idle Last Event: Start Last Error: Cease Import: [ reject ] Options: <Preference PeerAS LocalAS Refresh> Holdtime: 90 Preference: 170 Local AS: 101 Local System AS: 0 Number of flaps: 2 Last flap event: Stop Error: 'Cease' Sent: 1 Recv: 0 NLRI we are holding stale routes for: inet-unicast Time until stale routes are deleted or become long-lived stale: 00:01:54 >>>>>>>>>> Time until end-of-rib is assumed for stale routes: 00:04:54 Table inet.0 RIB State: BGP restart is complete Send state: not advertising Active prefixes: 14 Received prefixes: 21 Accepted prefixes: 15 Suppressed due to damping: 0 Stale prefixes: 21 >>>>>>>>>>>>>>>>>> With the fix: Peer: 8.3.0.2 AS 100 Local: 8.3.0.1 AS 101 Group: EBGP Routing-Instance: master Forwarding routing-instance: master Type: External State: Active Flags: <> Last State: Idle Last Event: Start Last Error: Cease Import: [ reject ] Options: <Preference PeerAS LocalAS Refresh> Holdtime: 90 Preference: 170 Local AS: 101 Local System AS: 0 Number of flaps: 1 Last flap event: Stop Error: 'Cease' Sent: 1 Recv: 0 PR1129271

  • On QFX10000 line platforms, during a route next-hop churn or an earliest deadline first (EDF) job priority changes, memory corruption might occur, leading to processing issues and constant packet drop. PR1243724

  • We strongly recommend using BGP as the protocol for configuring the local-address for each multihop iBGP/eBGP peer configuration. We recommend that local-address be a routeable lo0 address. Using loopback address reduces dependency with interfaces. Note: Multihop is by default enabled for iBGP peers. PR1323557

  • On a scaled setup, when the host table is full and the host entries are installed in an LPM table, OSPF sessions might take more time to come up. PR1358289