Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?


Open Issues

Learn about open issues in this release for the PTX10001-36MR, PTX10003, PTX10004, and PTX10008.

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

Authentication and Access Control

  • Remote authentication through console is not supported. Local authentication through console and remote authentication through management interface is supported. This is the expected behaviour on Junos OS Evolved going forward. PR1565251

Class of Service (CoS)

  • After multiple activate/deactivate of class-of-service configuration, and simultaneously running show class-of-service command can result in COSD core file. COSD recovers after the crash. PR1580573


  • Sometimes BUM traffic coming through evpn-mpls tunnel gets dropped or duplicated when going out of aggregated Ethernet interface after tunnel termination when aggregated Ethernet members are spanned across multiple Packet Forwarding Engines. This happens only when aggregated Ethernet has member on root Packet Forwarding Engine in BUM replication path and an additional member on another PPFE. PR1578314

General Routing

  • The convergence time is of the order of 3508 ms for OSPF or IS-IS (50,000 routes scale, which is equivalent to 14,253 routes per second). In case of BGP, it is of the order of around 17,000 routes per second. PR1379961

  • On Junos OS Evolved platforms running BGP with configuration statement route-record or sampling, the rpd process might crash. PR1493963

  • On a PTX10008 device, six SIBs are required to carry line-rate traffic, with no fabric redundancy. Even when ingress traffic rate is such that five SIBs are sufficient to carry ingress traffic (for example, traffic is less than 1280 Gbps), ungraceful SIB failures result in transient loss of traffic, until system failure handling is triggered. We recommend that you take the fabric cards offline by using the request chassis sib offline command before removing the SIBs for maintenance. PR1497212

  • Layer 2 circuit ping that uses CC type 1, also referred to as "PWE3 Control Word with 0001b as first nibble," is not supported. PR1503724

  • On a large scale routing on PTX10008, when an aggregated Ethernet interface flaps, during the next-hop convergence, you might see syslog filled with Possible out of order deleting of AftNode error messages. PR1513474

  • When upgrading from Junos OS Evolved 20.3R1 to a future release which has an FPC type which is incompatible for restart upgrade, the user is prompted in the CLI whether to offline the FPC or abort the upgrade. The CLI command waits for one hour for the user to respond. In case the user does not respond in an hour, the CLI command request system software add var | tmp | ptx.iso restart times out. Subsequent CLI software add |delete |rollback commands fail with the software upgrade in progress. The workaround for this is to restart orchestratord, the application responsible for managing CLI upgrade operations. Restarting orchestratord has no impact on the functioning of the system. The command is request system application app orchestratord node primary re restart. PR1516384

  • Using the following sensor path: /mpls/signalling-protocols/segment-routing/aggregate-sid-counters/aggregate -sid-counter/mpls-label, mpls-label values do not get reaped out when configured for SR-sid ingress sensors. PR1516811

  • The FPC generates a core file that gets stuck in /var/lib/ftp/in/. Sometimes when core files from multiple FPCs are generated, the core files from one FPC can get left behind in /var/lib/ftp/in and are not processed by the core-mgr. If the show system core-dumps shows some older cores in /var/lib/ftp/in/fpc-hostname and core files newer than those appear in /var/core/node-name, it indicates this issue was hit. PR1522404

  • On any Junos OS Evolved system with SSD, in rare scenarios, with multiple abrupt and/or ungraceful power-cycles or reboots, the filesystem might encounter error or it might not get initialized resulting in user login failure. This issue does not occur on a graceful shutdown of the system. PR1523238

  • On platform running Junos OS Evolved with dual Routing Engine support, when a new backup Routing Engine is inserted into the system, the request system software sync CLI command could not synchronize the software from the primary Routing Engine to the backup Routing Engine. PR1529924

  • The SR-TE label with BSID traffic gets dropped in PTX10003. Please use the following configuration statement as a workaround:


  • Each filter contains one or more filters terms. Each filter term might contain a counter. If traffic hits the filter term then the counter (packets/bytes) increments. If a user tries to modify the unsupported scaled filter then counters (packets/bytes) in filter terms are cleaned up which means the counters do not retain the original values before the modification. PR1530597

  • Any route or nexthops updates going during switch over process can appear out-order back on the primary. This can result into rpd core. PR1531653

  • When multicast traffic is running over an aggregated Ethernet bundle, which spans multiple FPCs, if one of the FPCs is brought offline, the traffic on that FPC is impacted. It is expected to see some traffic drops. PR1531859

  • In a scaled setup, it is possible that during GRES operation, the rpd on the primary Routing Engine m receive an ifstates update (DEL and CHANGE) and these are not expected during rpd re-syncing. In such a case the rpd can generate a core file and might recover automatically after sometime. PR1537947

  • During PHP, exp rewrite does not happen on the exposed MPLS payload. PR1538918

  • Software does not support GRES, when FPCs are triggered for restart. GRES with FPC triggered for offline or online is supported. PR1539685

  • Available space in /data on RE0 (primary) and RE1 (backup) can become asymmetric due to the difference in size of the db files. This can cause the software upgrade to fail if there are more than 10 images on the system. Also if /data is only full in RE1 (backup) and not on RE0 (primary) then force option may not able to clean up the old software images to make space for new image to get installed. As a workaround, use the CLI command request system software delete to remove images from the system until there is adequate space on RE0 and RE1 to install the new software image. PR1556992

  • On a PTX10008 running Junos OS Evolved, the show interface command combined with refresh configuration statement becomes slow after a clear interface statistics command. PR1557843

  • Always use the show system switchover command on the backup Routing Engine to determine if the standby is ready to take over. PR1557461

  • We have provided a configuration statement to control zero suppression behavior for streaming telemetry. The default behavior is still going to be suppressing zeroes so there will be no backward compatibility issues.

    The new configuration statement is: set services analytics zero-suppression no-zero-suppression. Committing this configuration disconnects the existing collectors and users are required to establish streaming telemetry connections again from their collectors.PR1559882

  • The asynchronous write APIs to CDA are fire-and-forget calls. If CDA clients make asynchronous write API calls to CDA server, the calls at rate higher than the CDA server can process them, this leads to memory usage buildup in GRPC infrastructure, eventually leading to memory allocation failures in GRPC infrastructure code and process abort. PR1565427

  • The command sysmanctl app stop distributord node re0 trigger is to be avoided in PTX10008 with NSR enabled as this may lead to flap of BGP sessions. PR1570574

  • Per-BT-chip sampling rate limit reduced from 35000/seconds to 7000/seconds. PR1571929
  • This issue is seen on PTX10008 with EVPN MPLS configuration, when there is an aggregated Ethernet interface associated with evpn-extended vlan. The aggregated Ethernet has multiple member links on different FPCs. It is observed that on some FPCs IFBD state for the aggregated Ethernet is out of sync with app controller for some reason. When the flood-nh update comes, it does not find the IFBD and registers for late binding. The IFBD object present in the controller, sends the flood-nh update again and this cycle keeps on repeating. This cycle keeps the Packet Forwarding Engine process busy and the Packet Forwarding Engine is unable to handle requests from CLI, and it causes the CLI timeout issue with error communication failure with /fpc0/evo-aftmand-bt/. Also as flood-nh programming is unable to finish due to missing aggregated Ethernet IFBD, this causes issues with L2 flooding within the bridge domain. The chance to see this issue is relatively low. PR1571929

  • Changing Decap only tunnel destination address configuration after tunnel is up is not handled and previously configured tunnel destination address is used for decapsulation. Once system enters this state any further changes to tunnel configuration is not handled. PR1575724

  • The Junos OS Evolved jkeys match the Junos OS jkeys util-asic-to-host-packet-rate -> util-wan-and-host-inject-pps-rate. PR1578134

  • If a switchover is performed while an FPC is in the offline or online state, the FPC moves to the current state on the primary. PR1580726

  • On the HwD app restart, certain propertyE objects might not be recreated. PR1580735

  • In certain scenarios, shapers applied on a 10G interface might drop the traffic more than the configured max-rate. PR1580795

  • Ping does not work when OSPF configuration is applied. PR1581130

  • The config-sync service might appear offline on the backup Routing Engine in the case of a USB upgrade. PR1582717

  • The sflow ingress sampling does not work for user IPv6 traffic with aggregated Ethernet ECMP case at last hop router with UHP LSP. PR1582960

  • The OSPF-hello DDOS statistics pktCnt lists as 0. PR1584458


  • When using a source IP address as the management address of the box to ping a network address on a peer, the response for the ICMP ping from the peer can end up on the management interface of the box, which is dropped by the Linux kernel as the RPF check is set to strict by default on the Linux kernel used on Junos OS Evolved. The Linux kernel expects the path to the peer to be on the WAN side and so drops the packet when it is received on the management interface from the peer when the RPF check for the management interface is set to strict. PR1498255

  • The standard "issue" message is not displayed for the initial login attempt, but if that login attempt fails for any reason and the user is re-prompted, then the message is displayed. This seems to be an issue common to PTX10004|PTX10008|PTX10001-36MR. PR1528996

Interfaces and Chassis

  • When routing instance primary IFF does not have an address, ping packets do not have source filled. This causes the ping packet reply to be not received. As a workaround, configure the lo0 address. PR1547622

  • LACP force up option does not work. PR1553997

  • There is no functional impact on resiliencyd. This core file is generated from the test command using cli-pfe. Once cored resiliencyd recovers, it behaves normally as expected. PR1578822

  • The format specifier of alarms is changed from string to boolean for following leafs. lane_tx_loss_of_signal_alarm lane_rx_loss_of_signal_alarm lane_tx_laser_disabled_alarm. PR1580113

  • The output and receiver power should be displayed with "- Inf" string if it 0mW as the following: Laser receiver power : 0.000 mW / - Inf dBm. This is not working in 21.1R1. PR1582541

  • The ifmand core file seen at IFAManager::handleIFACModify.

    • Dual Routing Engine system configured with 2 management interfaces having addresses with primary-only configuration.

    • Address primary-only flag is removed in later commit.

    • Ifmand crashes and recovers.

    • Issue is only applicable in 21.1R1 release and 20.3.

    • PR1583681

Layer 2 Features

  • It was observed rarely that issuing a request system zeroize does not trigger Zero Touch Provisioning (ZTP). A simple workaround is to re-initiate ZTP. PR1529246


  • The in-progress change of weighted ECMP to ECMP and vice versa does not take full effect for all routes using the LSPs until all the LSPs currently signaled with non-zero bandwidth have gone through at least one of the MBB/clear/disable events. The traffic distribution is not as expected until all the LSPs finish the transition. This is because without any of those triggers, the LSP does not automatically update its next hop based on weighted ECMP configuration. The next-hop content of each participation LSP affects the traffic distribution among ECMP paths. The weighted ECMP distribution ratio is not as expected until the adjustment of all LSPs is finished. PR1501860

  • On PTX10008 with scaled LSPs (10,000 ingress,40, 000 transit), the show mpls lsp statistics command might take up to 55 minutes.PR1551765

User Interface and Configuration

  • On-box validation of the large YANG files on NETCONF through request-yang-validate RPC might not work as the validation process might reach out of memory state while loading such YANG files. PR1416972

  • When a user tries to deactivate the MPLS-related configuration, the commit fails on the backup Routing Engine. PR1519367

  • In the dual Routing Engine scenario, the backup Routing Engine does not sync the configuration change while deleting a deactivating interface or routing-instances from the primary. So, after the operation, the deactivated interface or routing-instances still exist on the backup Routing Engine. PR1552467

  • If an inet filter is configured with or without the family statement, then changing its configuration to without or with the family statement must not be done. PR1556424

  • If an inet filter is configured with or without a family statement, then changing its configuration to remove or add a family statement might cause the filter process (firewalld) to restart unexpectedly. PR1556426


  • When Layer 2 VPN is configured with warm-standby, then upon Routing Engine switchover the new primary generates a core file. PR1584642