Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

 
ON THIS PAGE
 

Open Issues

Learn about open issues in this release for QFX Series switches.

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

EVPN

  • QFX5210-64C :: After Deactivate and Activate underlay BGP configuration observing MH Load Balance Traffic failed. PR1878641

  • If there is Multicast-router-interface marked on an IFL part of Snooping, if remote IGMP querier ceases or IFL is disabled then the Multicast-router-interface state on IGMP/MLD Snooping will be timed out by default in 255 secs. Till then all Multicast traffic will be sent out of that interface. This is as per RFC. The issue can be mitigated by configuring the query-interval to tune the other querier timeout to lower value so that Multicast-router-interface marking is cleared soon and the Type-3 is advertised with IGMP-Snooping in Mcast community. PR1909989

General Routing

  • There exists a hidden CLI upgrade option to do "clean-install". This has been enabled in 23.4R2-S3 and Mainline from 25.1R1 onwards. Using CLI upgrade with this option will do a "nist" compliant secure-erase for SATA disks. This method of CLI upgrade needs to be used with caution since this will wipe clean all configs/logs/files on the SATA FS and re-install the image. PR1847058

  • There is an inconsistency in the way auto channelization works. Based on the peer status it may split up unexpectedly. So for configuration which requires a specific speed/channelization we recommend setting the port manually. PR1852964

  • The pfed process crashes after downgrading Junos MX/EX/QFX platforms below the 22.3 releases. It doesn't cause any traffic and service impact. There will be loss of some statistics. PR1854637

  • On QFX10K2-60C switches, if traffic is coming from MPLS and going towards EVPN VXLAN tunnel (problem with the next-hop via ARP/IPV6 NDP to reach final destination IP) then traffic drop will be seen. PR1861501

  • On QFX10002-60C platforms when ECMP (Equal Cost Multi-path) is required to get multiple-path to multiple hosts connected, and with ipv6 address configured as a destination address, the destination MAC rewrite process fails, due to the unilist nexthop for ipv6 destination is getting overwritten. This leads the IPv6 anycast packets from/to the host, to get dropped over the ECMP routes. PR1865354

  • On rebooting DCI (Data Center Interconnect) Gateway device, while the device is coming up, multicast traffic drop is observed in the highly scaled config. The traffic drop is observed in the following condition. The rebooted device was the EVPN DF on I-ESI and after coming up, the DF election is triggered and the device is elected as EVPN DF on I-ESI. After coming up and elected as DF, the device builds the multicast routes afresh. This results in traffic drop. PR1872219

  • On Junos platforms utilizing the JNP10008-SF (aka SIB8), systems experience CRC errors on fabric links between the SIB and the FPC (Flexible PIC Concentrator). These errors are attributed to electrical noise on an internal power rail within the SIB. This condition will lead to multiple FPC-to-SIB link failures, potentially affecting traffic forwarding and overall fabric stability. PR1882584

  • On QFX10K8/16 , IFL memory is not freed on non-local interface of FPC during configuration changes (eg: IFL delete/deactivate) for L3IFL/L2IFL on AE/Scalar interfaces. Fix is present in common code and risky. It is a day one issue and the per IFL leak is minimal (0.00016% of total Kernel heap size). PR1884163

  • On all QFX-5K devices where EVPN-VXLAN is configured. Some MAC(Media Access Control) Routes were left pointing to stale SVLBNH(Shared Virtual LAN Bridging Next Hop). This showed up in PFE (Packet Forwarding Engine) as traffic "traffic directed to CPU" and not getting forwarded in Data Plane (dropped). PR1887225

  • On Junos QFX5100 and EX4600 platforms with releases 21.4R3-S3, 21.4R3-S10, and 21.4R3-S11, the SNMP trap generated after reinserting a PEM on the master switch incorrectly reports the jnxOperatingState as 6 (down) instead of the expected value 2 (running). This behaviour is consistently reproducible across multiple versions and persists even after performing a mastership switchover. PR1888543

  • The IPv6 multicast packet gets flooded by the DF Spine to NDF and ESI leafs, loops back to the Spine, and causes the switch to see the same MAC on different ports, resulting in a traffic loop and MAC flapping. PR1916949