Previous Releases

8.2R3

The following issues have been resolved since JUNOS Release 8.2R2.4. The identifier following the description is the tracking number in our bug database.

Platform and Infrastructure

• On M-series and J-series routing platforms, when you include the vrf-table-label statement at the [edit routing-instances routing-instance-name] hierarchy level, the incoming traffic is considered to come from the internal label-switched interface (LSI) associated with the VRF instance. The original incoming logical interface is unknown, so the traffic is not accounted for by the original incoming logical unit. Furthermore, the LSI is an internal interface and has no accounting support. [PR/53148: This issue has been resolved.]
• If you enable the copy-PLP and filter-class features simultaneously with an ATM interface, an unusable topology might be installed in the Packet Forwarding Engine and control traffic might not be handled properly. As a workaround, disable one of the features. [PR/62511: This issue has been resolved.]
• When you issue the show mpls lsp detail command and there is a lot of data to display, the routing protocol process (rpd) might consume 100 percent of the CPU. [PR/75975: This issue has been resolved.]
• When you configure port mirroring, the Packet Forwarding Engine might not handle memory properly, which causes it to lose connection with the Routing Engine and then reboot. [PR/76668: This issue has been resolved.]
• On the TX Matrix platform, when there is a link failure on the preferred path between the LCC and the SCC, the connection does not gracefully switch to the alternate path, resulting in traffic loss. [PR/94629: This issue has been resolved.]
• The show pfe fpc and show pfe feb commands might fail to produce output and instead display the message "error: Unrecognized command". [PR/97014: This issue has been resolved.]
• If you are running VPLS, you must not use the indirect-next-hop statement at the [edit routing-options forwarding-table] hierarchy level. [PR/97070: This issue has been resolved.]
• When a combination of cRTP and LFI packets are sent over a MultiServices 100 PIC at high rates (greater than 350 kpps total), forwarding of traffic over that PIC might stop. [PR/97304: This issue has been resolved.]
• The output of the show route forwarding destination command is limited to 16 indexed next hops, instead of 16 unique indexed next hops. [PR/98381: This issue has been resolved.]
• With large numbers of next hops, the EFPC might overflow its stack, crash, and restart. [PR/98562: This issue has been resolved.]
• The kernel might log DEBUG level messages similar to "/kernel: _aggregate_nh: Sending weight of 1 for index 2"; this is cosmetic only and has no operational impact. [PR/98640: This issue has been resolved.]
• The default Internet Control Message Protocol (ICMP) rate limit has been increased from 50 packets per second to 250 packets per second. No new statements are required for the new default to take effect on the M120 router. [PR/99708: This issue has been resolved.]
• If you enable the Virtual Router Redundancy Protocol (VRRP) in a routing matrix, the T640 routing nodes might generate core files and stop functioning. [PR/104673: This issue has been resolved.]
• Under heavy, bursty traffic conditions, an FPC Type 4 with an OC-768 PIC might continuously RED drop almost all traffic going through one queue. Only an FPC reboot can recover from this condition. [PR/228909: This issue has been resolved.]

User Interface and Configuration

• If a connection from a client to the router is not established through the CLI (is a JUNOScript session, for example), the router never closes the connection if it becomes idle or closes unexpectedly on the client side. [PR/11689: This issue has been resolved.]
• If you delete and then reconfigure an applied configuration group that does not contain any configuration data, the routing protocol process (rpd) might generate a core file and stop operating. [PR/63677: This issue has been resolved.]
• According to RFC 3986, port numbers can be optional in SCP transfer syntax. The JUNOS CLI accepts the syntax with or without using a colon to separate the hostname from the port number, but the UNIX shell requires a colon. The SCP transfer accepts the syntax only if a colon is not used. This fix enables routers to handle zero digit port numbers. [PR/66375: This issue has been resolved.]
• When mutiple users simultaneously use JUNOScope to perform device interactions on the same device, more than one management (mgd) session might be created on the device. Of the mgd sessions created, one is used by JUNOScope for pooling purposes. The remaining mgd sessions are not automatically closed even after interaction with the device is completed. [PR/72451: This issue has been resolved.]
• When either or both of the allow-command and deny-command statements are included for a login class at the [edit system login class class-name] hierarchy level, users who belong to the class might experience long delays when trying to connect to the router (for example, to start a CLI, JUNOScript, or NETCONF session). [PR/95042: This issue has been resolved.]
• If you include the traceoptions statement in a configuration that also includes the apply-groups statement, the JUNOS software process responsible for generating the trace might generate a core file and stop functioning. The process might continue to malfunction when restarted. [PR/101491: This issue has been resolved.]
• During reboot of the router, the previously active configuration is automatically committed. If the commit operation does not completely succeed and you then issue the show configuration command, the JUNOS management process (mgd) generates a core file and stops functioning. [PR/102442: This issue has been resolved.]
• JUNOScope user password can be minimum of 6 characters and maximum of 128 characters [PR/105151]
• If you configure a user login class containing deny-commands, and the regular expression that is used to define which commands are denied includes a top-level statement or command (such as configure), the user might still be able to gain access to the statement or command. [PR/228617: This issue has been resolved.]
• When apply-groups are referenced in a routing instance, the configuration commit operation fails and causes foreign file propagation (ffp), and the management process (mdg) core dumps. [PR/233470: This issue has been resolved.]

Interfaces and Chassis

• When an Ethernet PIC operating in cut-through mode receives a packet without a destination MAC address, the output of the show interfaces interface-name extensive command might include incorrect values. This event correctly increments the value in the Input DA rejects output field, but incorrectly also increments the value in the L3 incompletes field. [PR/56750: This issue has been resolved.]
• In vlan-ccc mode, the multicast source MAC bit filtering allows packets with the multicast source MAC bit to be set and reset when vlan-ccc mode is disabled. [PR/67401: This issue has been resolved.]
• For Channelized STM1 IQ PICs installed in M320 routers, if you activate or deactivate an E1 channel in a link services intelligent queuing interface (LSQ) or redundant LSQ bundle, and then issue the show interfaces extensive command for other E1 channel bundles on the same PIC, the Missing sequence number field might increment in the output of the command. [PR/77592: This issue has been resolved.]
• When the amount of memory required by interface information reaches the default data size limit of 22 MB, memory leak issues might occur. . [PR/77880: This issue has been resolved.]
• If you modify a time slot on a member link of a redundant link services (rlsq-) interface bundle when the Routing Engine is unusually busy (because of route flaps, for example), that particular member link might stop forwarding traffic. [PR/78591: This issue has been resolved.]
• On Gigabit Ethernet IQ PIC interfaces, the show interface extensive command output displays incorrect values for the Total and Unicast packet counters. [PR/81079: This issue has been resolved.]
• On Adaptive Services (AS) PICs, if the PIC receives a large amount of traffic, CPU utilization is high, and there are many memory transactions, some components in the PIC might become stuck and stop processing frames. [PR/94287: This issue has been resolved.]
• ATM-over-SHDSL interfaces on J-series Services Routers sometimes discard incoming MPLS packets. [PR/94353: This issue has been resolved.]
• On M7i routers, when you disable the integrated Gigabit Ethernet interface, you might see a ping round-trip time (RTT) delay of about 50 msec. As a workaround, do not disable the interface, but deactivate the configuration for the integrated interface. [PR/95572: This issue has been resolved.]
• If you take a PIC offline and bring it back online, the chassis process (chassisd) might leak memory. [PR/95964: This issue has been resolved.]
• If you install a first-generation SONET/SDH PIC in an Enhanced II FPC, the FPC might stop operating. [PR/96309: This issue has been resolved.]
• The all-chassis option is only included with the request chassis routing-engine master switch command. This option should also be available with the request chassis routing-engine master acquire and request chassis routing-engine master release commands. [PR/96502: This issue has been resolved.]
• When the configuration for APS unidirectional mode is deactivated or deleted, the system might never switch to bidirectional mode, even if that mode is explicitly configured. As a workaround, deactivate APS on both the working and protect circuits and then reactivate it. [PR/96801: This issue has been resolved.]
• When a configuration change is committed, the device configuration process (dcd) can leak memory. Eventually, the process might exceed the maximum memory size allowed and restart unexpectedly. [PR/97295: This issue has been resolved.]
• On a J4350 Services Router with a Gigabit Ethernet interface, when you toggle the VRRP mastership state, the interface stops forwarding traffic. [PR/97420: This issue has been resolved.]
• For link services intelligent queuing (LSQ) interfaces configured on AS PICs installed in M40e routers, when a graceful Routing Engine switchover (GRES) event occurs, the AS PIC might reboot, PPP keepalives might be lost, and traffic loss might result. [PR/98391: This issue has been resolved.]
• On Gigabit Ethernet interfaces, when the routing platform processes a configuration containing the reject-the-rest toggle option, the device control process (dcd) might dump core. [PR/98972: This issue has been resolved.]
• During high congestion, 10-Gigabit Ethernet XENPAK PICs might trigger some very confusing messages as a result of packet loss. The fix replaces the error system log messages with counters that indicate the error count for txxg and rxxg interrupt registers. [PR/100022: This issue has been resolved.]
• Immediately after you boot the system or restart the chassis process (chassisd), buffer pool utilization MIB object jnxOperatingBuffer might be erroneously reported as –1. [PR/100170: This issue has been resolved.]
• For link services intelligent queuing (LSQ) interfaces configured on AS PICs, if you configure an FRF.16 bundle that contains the minimum-links statement, and then deactivate and reactivate one of the interfaces that is part of the bundle, the other active interfaces might not rejoin the bundle when the modified interface is reactivated and the bundle comes back up. This issue only happens if there is an explicit configuration change. [PR/101109: This issue has been resolved.]
• If an LSQ bundle goes down because of minimum-link member settings, the QoS bandwidth for the bundle is not applied correctly after the bundle comes back up again.. [PR/101119: This issue has been resolved.]
• When graceful Routing Engine switchover is configured and a switchover occurs, the kernel might reset and generate a core file. [PR/101359: This issue has been resolved.]
• When the AS PIC resets, it sometimes does not generate a core file as expected. [PR/101488: This issue has been resolved.]
• When you delete or deactivate an interface on a channelized IQ PIC, the PIC might stop operating and generate a core file. [PR/102420: This issue has been resolved.]
• In certain circumstances, a channelized IQ PIC might generate a core file and stop operating. [PR/103268: This issue has been resolved.]
• When you issue the show interfaces extensive command for an interface on an M-series router that uses ethernet-ccc encapsulation, the values in the Input rate field of the output might be incorrect. [PR/103527: This issue has been resolved.]
• On SONET/SDH and ATM interfaces, in some cases interface monitoring can cause a bad page fault because of a race condition. [PR/103589: This issue has been resolved.]
• If a member link in an aggregated Ethernet interface becomes inoperative, multicast packets might not be routed through other member links. [PR/105136: This issue has been resolved.]
• The delay between VRRP advertisements might be longer than the value specified by the fast-interval statement at the [edit interfaces interface-name unit logical-unit-number family inet address address vrrp-group group-number] hierarchy level. [PR/105436: This issue has been resolved.]
• If you configure an interface to be stacked VLAN initially and then change it to flexible VLAN, the management process (dcd) does not register the new change. As a workaround, deactivate and reactivate the interface. [PR/105626: This issue has been resolved.]
• On the 10-Gigabit Ethernet IQ2 PIC, the default weighted random early detection (WRED) profile is incorrect, causing incorrect RED handling on queue 0. [PR/228742: This issue has been resolved.]
• When you issue the monitor interface command for an OC768 PIC running at line rate, the byte count and bps rate values are longer than the space allotted for them, which causes the delta value to appear on the line below instead of the same line. [PR/229494: This issue has been resolved.]
• On Channelized OC-12 intelligent queuing (IQ) interfaces, the output-traffic-control-profile statement at the [edit class-of-service interfaces interface-name] hierarchy level is not supported. [PR/229880: This issue has been resolved.]
• When scheduler maps are applied to more than one interface, the interface process (dcd) might not release memory that it has allocated. [PR/230770: This issue has been resolved.]

Services Applications

• Because of interoperability issues or network failures or retries, a router might receive an IPCP CONF-REQ message even after IPCP is Up. When this happens, the PPP stack transitions from IPCP up to IPCP down. When the CONF-ACK message is finally received, a second IPCP up event occurs. [PR/73334: This issue has been resolved.]
• On Adaptive Services PICs, when you apply class-of-service (CoS) marking and reclassification for the SIP application, voice and video packets might be marked incorrectly in terms of direction, initiator, and responder. [PR/77249: This issue has been resolved.]
• When you configure the DLSW protocol, the lt-0/0/0 interface, and other interfaces and then commit the configuration, DLSW might not become established between peers. As a workaround, restart the DLSW process on both peers to establish the DLSW connection. [PR/79493: This issue has been resolved.]
• If initiator and response packets for TCP flows arrive at nearly the same time, the Adaptive Services PIC might generate a core file and stop functioning. [PR/94388: This issue has been resolved.]
• If a large number of flows expire in a short period of time, the Adaptive Services PIC might generate a core file and stop functioning. [PR/101117: This issue has been resolved.]
• When you configure a Session Initiation Protocol (SIP) application-level gateway (ALG) on an AS PIC, the PIC might stop operating and generate a core file. [PR/103973: This issue has been resolved.]
• A reverse route for a remote proxy network that is shared by multiple dynamic end point tunnels might be deleted if any of the tunnels stops functioning. [PR/105656: This issue has been resolved.]

General Routing

• When you include a policy with a from interface qualifier in a configuration, the routing process might restart unexpectedly. [PR/96752: This issue has been resolved.]

Routing Protocols

• When a Fast Ethernet interface is configured to accept multicast traffic and its associated PIC is taken offline and brought back online, the interface might not accept multicast traffic. The message "Could not enable multicast for INET on interface fe-x/x/x" might be written to the system log in this case. [PR/77351: This issue has been resolved.]
• The remove-private command does not remove the private AS from the AS path. The private AS is left in the AS path of the routes advertised to peers. [PR/78141: This issue has been resolved.]
• If you configure identical static and backup-router prefixes, the routing platform might incorrectly trigger a warning message. [PR/78615: This issue has been resolved.]
• In rare cases, processing an upstream assert message for an (S,G) node that was previously marked as deleted might cause the routing protocol process (rpd) to reset unexpectedly. [PR/80253: This issue has been resolved.]
• After Routing Engine graceful switchover, the OSPF MD5 crypto sequence number on the restarting router might be lower than the previous one used. If this is the case, the neighbor should reject all packets, and the adjacency will time out. If the OSPF MD5 crypto sequence number generated on the restarting router is higher than previously sent, the neighbor router will accept those packets. Only OSPF MD5 crypto sequence numbers lower than the ones previously used are not allowed. [PR/97050: This issue has been resolved.]
• If class of service-based forwarding is configured and the router fails over to the backup Routing Engine, the routing process (rpd) might restart. [PR/98686: This issue has been resolved.]
• When you include the always-compare-med statement at the [edit protocols bgp path-selection] hierarchy level, the routing protocol process (rpd) might generate a core file (but it continues to run). [PR/99220: This issue has been resolved.]
• For events such as graceful Routing Engine switchover or graceful restart where a previous state must be reconstructed, L3VPNs configured with vrf-table-label caused the label management library to lose track of already assigned labels. The JUNOS label management library for vrf-table-label was expecting a label value to be unassigned when in fact the label value was assigned. . [PR/99339: This issue has been resolved.]
• If you enable unicast reverse path forwarding (uRPF) when a next-table next-hop route is present, the routing protocol process (rpd) might restart repeatedly. [PR/99488: This issue has been resolved.]
• When a BGP routing table is removed, the routing protocol process (rpd) might stop operating and generate a core file. [PR/99777: This issue has been resolved.]
• When the interface for reverse path forwarding (RPF) goes down, PIM immediately removes the associated multicast route entry, even if unicast routing has not yet converged. In a configuration where multicast traffic is always sent on the backup interface, the immediate removal can cause a delay in route installation after unicast routing converges. [PR/102306: This issue has been resolved.]
• When a bootstrap router message (BSM) is received, the Protocol Independent Multicast (PIM) software does not verify that the message came from the appropriate reverse path forwarding (RPF) neighbor. The result is unnecessary forwarding of some BSMs. [PR/103376: This issue has been resolved.]
• When a routing table is updated with information from RIP, the routing protocol process (rpd) might generate a core file and stop operating. [PR/105027: This issue has been resolved.]
• When there is an LSA ID conflict in two summary LSAs for the generating router, RPD might crash. [PR/230127: This issue has been resolved.]
• If a policy with a community delete action deletes all communities, the routing process will leak memory. [PR/230207: This issue has been resolved.]
• When multiple MPLS LSPs are configured, the mplsTunnelReoptimized trap is triggered on every expiration of the value set for the 'optimize-timer' statement at the [edit protocols mpls label-switched-path path-name] hierarchy level. In addition, the no-trap statement set at the [edit protocols mpls log-updown] hierarchy level does not block the traps defined in RFC 3812, Multiprotocol Label Switching (MPLS) Traffic Engineering (TE) Management Information Base (MIB). [PR/231678: This issue has been resolved.]

MPLS Applications

• In some cases graceful restart does not work correctly with ingress RSVP LSPs. [PR/80061: This issue has been resolved.]
• Convergence time for VPN routes on M-series or T-series routers might take longer with fast reroute once the primary LSP is only one hop and LDP tunneling is configured. There are no delays with multihop LSPs, with or without LDP tunneling configured. As a workaround configure explicit null for LDP. This workaround does not apply to the M120 router. [PR/96730: This issue has been resolved.]
• When configuring LDP graceful restart, you can enter various values for these statements:
  • [edit protocols ldp graceful-restart recovery-time]
  • [edit protocols ldp graceful-restart maximum-recovery-time]
  • [edit routing-options graceful-restart restart-duration]

  If you configure values for the recovery-time and maximum-recovery-time statements that lead to a recovery time of 0 seconds being advertised to a peer router, LDP graceful restart will not function. [PR/97713: This issue has been resolved.]

• When an M120 router or an MX-series Ethernet Services Router is operating as a penultimate node for a transit bypass tunnel, link protection might not function properly. As a workaround, include the 'explicit-null' statement at the [edit protocols mpls] hierarchy level on the egress node of the bypass tunnel so that the router is no longer operating as a penultimate node. [PR/101833: This issue has been resolved.]
• When an LDP neighbor's IP address changes, information about the previous address persists in the newly established LDP session along with the new address. When the previous address times out, the FEC for the neighbor's loopback address might be mistakenly removed. As a workaround, clear the LDP session when the address changes. [PR/103001: This issue has been resolved.]
• An error in the Constrained Shortest Path First (CSPF) software might cause the routing protocol process (rpd) to generate a core file and stop operating. [PR/103777: This issue has been resolved.]
• After the show mpls lsp extensive command is issued, the routing process might log messages reporting that there was a scheduler slip, and BGP sessions might be disrupted. [PR/104179: This issue has been resolved.]
• MPLS graceful restart might not function in the main routing protocol process (rpd) if one or more logical routers are configured. [PR/228195: This issue has been resolved.]
• The router might begin to receive LDP hellos and start to establish an LDP adjacency even if the interface is not fully operational. [PR/228298: This issue has been resolved.]
• If an LDP adjacency times out while a router is acting as helper for another router that is undergoing LDP graceful restart, the graceful restart might not complete successfully. [PR/229848: This issue has been resolved.]
• When RSVP PATH messages are sent, the session name includes extra characters. In accordance with RFC 3209, RSVP-TE: Extensions to RSVP for LSP Tunnels, the session name should instead be padded with trailing NULL characters. [PR/232926: This issue has been resolved.]
• RSVP PATH messages are not bundled when RSVP aggregation is negotiated on a point-to-point link. [PR/234545: This issue has been resolved.]

VPNs

• If a route target is not imported by any instance, then a vrf-table-label may not be advertised. [PR/98322: This issue has been resolved.]

Class of Service

• Unusually high memory allocation might occur for link services (ls) interface queues because the queues' buffer allocation is not based on bundle bandwidth. As a workaround, apply a scheduler map to the multilink bundle, with buffer sizes configured to be a fraction of the original buffer sizes. The maximum buffer size allocated to the multilink bundle interface does not affect other interfaces as long as the sum of buffer sizes of the queues on the multilink interface equals a fraction of the maximum buffer size. [PR/70562: This issue has been resolved.]
• On J-series Services Routers, when you modify the value for the shaping-rate statement at the [edit class-of-service adaptive-shapers adaptive-shaper-name trigger] hierarchy level and then issue the commit command, the change does not take effect on the interface to which the adaptive-shaper adaptive-shaper-name statement is applied. As a workaround, deactivate and then reactivate the interface. Issue the deactivate class-of-service interfaces interface-name unit unit-number adaptive-shaper command, followed by the activate class-of-service interfaces interface-name unit unit-number adaptive-shaper command. [PR/82243: This issue has been resolved.]
• When the policer action involves setting the loss priority or the forwarding class, the policer counter is not incremented. [PR/96341: This issue has been resolved.]
• If you modify the buffer size in a class-of-service scheduler, the FPC might reset. [PR/99780: This issue has been resolved.]
• If you include the scheduler-map statement at the [edit class-of-service interfaces aex] hierarchy level and a member link of the aggregated Ethernet interface goes down or comes back up, traffic forwarding on the routing platform halts for approximately 400 milliseconds. [PR/102200: This issue has been resolved.]

Forwarding and Sampling

• On all platforms, certain combinations of load override and rollback operations might cause firewall filters configured for the VPLS, CCC, and TCC families not to work correctly. To restore correct performance, deactivate and reactivate each interface to which the filters are applied. [PR/95140: This issue has been resolved.]
• If you create an input firewall filter chain with multiple filters referencing the same policer (term-specific policer, three-color-policer, or prefix-action template), the firewall filter compiler might stop operating. As a workaround, include the filter-specific statement or apply the policer to a single firewall filter containing multiple terms. [PR/96040: This issue has been resolved.]
• On M10i and MX960 routing platforms, filters are not replicated to the backup Routing Engine, with the result that the filter disappears after a graceful Routing Engine switchover (GRES) event. As a workaround, issue the commit synchronize command. [PR/98113: This issue has been resolved.]
• When you configure an interface in a logical router and apply a firewall filter, the software allows you to delete the firewall filter from the configuration while the interface still references it. When you commit the configuration, the system returns a warning message: "Referenced filter 'filter-name' of family 'family-type' is not defined," but allows the commit to succeed. [PR/98442 : This issue has been resolved.]
• On T-series platforms only, for traffic generated by the Routing Engine, forwarding-class filters do not work properly because the QoS field is not initialized correctly in the kernel. [PR/98943: This issue has been resolved.]

Network Management

• If there is a communication failure when the routing protocol process (rpd) is processing an SNMP request from the SNMP process (snmpd), rpd might dump core. [PR/69969: This issue has been resolved.]
• Taking a PIC offline and then back online causes the chassis process (chassisd) to leak memory. [PR/97479: This issue has been resolved.]
• When a large number of trap packets are sent in a short interval and the packets are kept in the trap queue with more than two trap destinations configured, the SNMP process (snmpd) runs out of memory. [PR/98306: This issue has been resolved.]
• On J4350 and J6350 routers, the engine-id use-mac-address statement at the [edit snmp] hierarchy level is not supported. [PR/98711: This issue has been resolved.]
• When the bcm0 interface flaps on a TX Matrix platform master unit, the SNMP process (snmpd) connection between the T640 routing nodes and the master unit does not reset properly. [PR/100593: This issue has been resolved.]
• When forwarding MPLS LSP traps with a nondefault context (for example, from a nondefault logical router or routing instance), the SNMP process (snmpd) might dump core. [PR/100851: This issue has been resolved.]
• The minimum value for the size statement at the [edit snmp traceoptions file] hierarchy level is 10000, which is less than the minimum value of 10240 that applies to tracing configuration at other hierarchy levels. Attempts to commit a configuration with the value 10000 wrongly fail, because the verification software applies the standard minimum of 10240. [PR/101560: This issue has been resolved.]
• The Management Information Base (MIB) process (mib2d) sometimes generates a core file and logs the message "SNMP_GET_ERROR2: jnxDCUsEntry get-next failed for dcu index/name: index1 <index1> index2 <index2> (Operation timed out)." [PR/228471: This issue has been resolved.]

8.2R2

The following issues have been resolved since JUNOS Release 8.2R1.7. The identifier following the description is the tracking number in our bug database.

Platform and Infrastructure

• Route records for Routing Engine-based sampling are exported to the AS PIC, even when the records are not needed by the AS PIC. This action consumes valuable memory resources on the AS PIC. [PR/75550: This issue has been resolved.]
• The NTP daemon selects the best source address to reach its peer, overriding the configured source address. [PR/94290: This issue has been resolved.]
• On routing platforms that use IPv6 addresses, duplicate address detection (DAD) fails after some operations (including graceful Routing Engine switchover). As a result, ping requests are sent from an invalid source address and fail. [PR/94946: This issue has been resolved.]
• When an FPC reports a hardware error for a PIC, a FRU failure trap might not be generated. [PR/95006: This issue has been resolved.]
• On M120 Internet routers, large VLAN configurations might cause the router to repeatedly dump core. [PR/95012: This issue has been resolved.]
• When a TCP session is initiated from a local endpoint that is defined in a VPN routing and forwarding (VRF) table, the session can fail when the path maximum transmission unit (MTU) changes to a value small enough to make packet fragmentation necessary. [PR/95379: This issue has been resolved.]
• On M10i routers, when the system is upgraded to the JUNOS 8.2R1 release, memory buffers might become full and cause a core dump. [PR/95671: This issue has been resolved.]

User Interface and Configuration

• When mutiple users simultaneously use JUNOScope to perform device interactions on the same device, more than one management (mgd) session might be created on the device. Of the mgd sessions created, one is used by JUNOScope for pooling purpose. The remaining mgd sessions are not automatically closed even after interaction with the device is completed. [PR/72451: This issue has been resolved.]
• RADIUS/TACACS+ authenticated users who are mapped to a local template user might not be able to archive files using the CLI. [PR/96266: This issue has been resolved.]
• When you configure commit script management at the [edit groups group-name system commit] hierarchy level and commit the configuration, the CLI session terminates. As a workaround, configure commit script management at the [edit system commit] hierarchy level. [PR/96917: This issue has been resolved.]

Interfaces and Chassis

• When an M40e router chassis keeps sending a “failed reading voltage” error message, it might indicate that an incorrect power entry module (PEM) is installed. [PR/70868: This issue has been resolved.]
• When you enter an name with more than 80 characters as the interface-name variable used with the show interfaces descriptions interface-name command, the system produces a core dump. [PR/72155: This issue has been resolved.]
• On non-Enhanced FPCs with STM16 PICs on M160 routing platforms, if you configure the MTU size to the maximum value of 9192 bytes, packet forwarding might stop. As a workaround, reset the MTU size to the default of 4474 bytes. [PR/76390: This issue has been resolved.]
• When you configure a local gateway for an IPSec tunnel within a routing instance, and then change the routing instance type from VRF to virtual router and commit the configuration, the IP address of the local gateway might disappear from the interface configuration and the IPSec tunnel might not become established. [PR/76894: This issue has been resolved.]
• In a few cases, when member links of an MLPPP bundle are deactivated and then reactivated, output from the show interfaces interface-name statistics detail command might display unreliable statistics. [PR/78696: This issue has been resolved.]
• You cannot configure the bandwidth on DS, T1, and E1 interfaces. [PR/81827: This issue has been resolved.]
• On a dual-port Channelized T1/E1 PIM with interfaces channelized into DS0 interfaces, all links may not be available after all 24 timeslots on one ct1 port are mapped to 24 timeslots on another ct1 interface on the same PIM. [PR/81976: This issue has been resolved.]
• When MPLS packets are received on SHDSL (ATM) interfaces on a J-series router, they might be dropped. [PR/94353: This issue has been resolved.]
• When a redundant link services (rlsq-) interface is modified, if an internal query process is being run at the same time as the modification, the kernel might process might stop. [PR/94508]
• While using MLPPP, the loop-detected device flag remains active for the local interface, even after the remote loop is removed from the end router. [PR/94800: This issue has been resolved.]
• On M320 or T-series routing platforms, logical interfaces on Channelized T3 PICs that are configured with DLCIs of less than 64 in frame-relay-ccc or frame-relay-tcc encapsulations will not pass traffic. [PR/95358: This issue has been resolved.]
• On ATM2 PICs, if you initially configure both the atm-options linear-red-profile statement and the sonet-options aps statement and then change a portion of the linear-red-profile configuration later, the interface associated with the linear-red-profile might flap even if there is no change in the associated linear-red-profile properties. [PR/96621: This issue has been resolved.]
• When you remove a spare Switch Interface Board (SIB) without first taking the SIB offline, the working SIB in slot 1 erroneously changes to Check status in the output from the show chassis sibs command. [PR/96959: This issue has been resolved.]
• the JUNOS software truncates the endpoint discriminator option block received from the peer while sending an LCP config- reject message. This is the case when the endpoint discriminator is an IEEE 802.1 MAC address. [PR/97169: This issue has been resolved.]

Services Applications

• Some probe types require particular options to be configured using specific attributes. For example, the udp-ping-timestamp statement requires a minimum data size of 12. The minimum data size for TCP probe packets is 1. Using a smaller value (or none) for data size results in a commit error. [PR/69765: This issue has been resolved.]
• If traffic is heavy in a tunnel that has numerous concurrent Layer 2 Tunneling Protocol (L2TP) sessions, the AS PIC might discard some Layer Control Protocol (LCP) echo messages. [PR/72334: This issue has been resolved.]
• In active FTP mode on a services PIC, FTP application-level gateways (ALGs) refuse connection attempts that originate at port numbers other than 20. [PR/82232: This issue has been resolved.]
• If you type Ctrl+c to interrupt the request security pki ca-certificate enroll ca-profile ca-profile-name command, the key management process (kmd) might terminate and create a core file. [PR/94713: This issue has been resolved.]
• When the DF bit is set in the IP header, CRTP does not compress the packet. [PR/95396: This issue has been resolved.]

General Routing

• When load balancing is configured, an install-nexthop policy action might not work completely. [PR/72457: This issue has been resolved.]

Routing Protocols

• In certain cases, on multiprotocol BGP sessions configured with a family route-target (AFI/SAFI 1/132), receipt of the very first route-target NLRI might cause the routing process (rpd) to dump core. [PR/78876: This issue has been resolved.]
• If you configure BGP path selection with BGP groups that have no neighbor setting, the routing process (rpd) dumps core. [PR/95558: This issue has been resolved.]
• In IS-IS XML output, when the JUNOS routing software encounters sub-TLV-related XML tags not enclosed within the IS-reachability XML tag, it discards this information. [PR/95711: This issue has been resolved.]
• When you commit only the routing-options section of a configuration, changes in the confederation member list are not forwarded to BGP. [PR/96277: This issue has been resolved.]

VPNs

• On M10i routers configured with VPNs, the MIB variables in vpnRTTable, vpnPwTable, and vpnIfTable might provide incorrect values. [PR/75649: This issue has been resolved.]
• A routing instance can fail to be deleted properly within the routing process if another routing instance is configured with the same vrf-target. The workaround is to remove that vrf-target from all routing instances. Modifying the route distinguisher in the routing instance configuration can cause prefixes added to the routing table for a particular VRF to not be propagated to the forwarding table.The workaround is to change the name of the routing instance. [PR/80933: This issue has been resolved.]
• A flag mismatch prevented BGP from flashing Layer 2 VPN routes to VPLS instances.[PR/82517: This issue has been resolved.]
• In earlier releases of JUNOS 8.0 and 8.1, when you issued the show snmp mib walk jnxVpnIfTable command, the output provided only the first entry of each instance. [PR/95924: This issue has been resolved.]

Class of Service

• On ATM 2 PICs, when a logical interface is disabled, there is no channel table entry for it. However, the class-of-service process (cosd) still tries to configure a classifier for the channel corresponding to that disabled logical interface and the operation fails because the channel entry does not exist. [PR/71081: This issue has been resolved.]
• Insufficient memory was allocated for forwarding-class configuration data. As a result, trying to access information might have exceeded the allocation and caused the CoS process (cosd) to terminate unexpectedly. [PR/95404: This issue has been resolved.]

Forwarding and Sampling

• When host sampling is configured, some rarely occurring conditions can cause the Packet Forwarding Engine to stop functioning. [PR/74173: This issue has been resolved.]
• When you configure source class usage and change the configuration for a policer, the changes might not take effect. [PR/78754: This issue has been resolved.]
• Flow-route filters that are configured with terms matching TCP fragments might not match the TCP fragments. [PR/81235: This issue has been resolved.]

Network Management

• If you rapidly poll ipCidrRouteNumber OID .1.3.6.1.2.1.4.24.3.0 for multiple virtual routers, you might see incorrect values for all but the first table. [PR/95217: This issue has been resolved.]
• When you set the SNMP engine-id statement to use-mac-address, the engine ID is not visible after you reboot the router. As a workaround, manually restart the SNMP process (snmpd). [PR/97356: This issue has been resolved.]

System Log

• The logged text for the CHASSISD_FASIC_INPUT_DROP system log tag refers to the “Packeting Forwarding Engine” instead of the “Packet Forwarding Engine”. [PR/77189: This issue has been resolved.]

8.2R1

The following issues have been resolved since JUNOS Release 8.1R1. The identifier following the description is the tracking number in our bug database.

Platform and Infrastructure

• When you include the explicit-null statement at the [edit protocols mpls], [edit protocols ldp], or [edit protocols bgp group group-name neighbor ip-address family inet labeled-unicast] hierarchy level, the default time-to-live (TTL) value might not be copied correctly from an MPLS null label to an IP packet header. [PR/44814: This issue has been resolved.]
• If you configure the log-out-on-disconnect statement at the [edit system ports] hierarchy level and the console statement at the [edit system syslog] hierarchy level, and you remove and replace the console cable, the console port might lock. [PR/69459: This issue has been resolved.]
• If you configure the routing platform as a DHCP relay agent, it might incorrectly broadcast the BootReply message (DHCP offer) to the client, regardless of whether the client's Broadcast flag bit is set or not. [PR/74799: This issue has been resolved.]
• If you telnet to the loopback address of a router configured for graceful Routing Engine switchover and switch mastership to the backup Routing Engine, the route to the telnetted host is cloned to the Backup Routing Engine and points to a reject next hop. When mastership switches back to the primary Routing Engine, this reject next hop remains in the routing table and might affect connectivity to the host. [PR/76996: This issue has been resolved.]
• When the Self-Monitoring Analysis and Reporting Technology System (SMART) process (smartd) performs an automatic power check every 20 seconds to determine if a disk drive is in standby mode, this check might cause some Routing Engine hard drives to become unresponsive. To resolve this issue, the automatic power checking function has been disabled by default. [PR/77503: This issue has been resolved.]
• For M160 routers, if you configure IS-IS and graceful Routing Engine switchover (GRES), and then take the first Switching and Forwarding Module (SFM) in the chassis offline, IS-IS neighbor adjacencies might transition down and up. [PR/77771: This issue has been resolved.]
• When you configure IPv6 on an interface and the interface comes up late in the duplicate address detection (DAD) retry cycle, the IPv6 address might remain tentative. [PR/78212: This issue has been resolved.]
• M7i and M10i routers containing a Compact Forwarding Engine Board (CFEB) with 256 MB of DRAM might not operate properly with JUNOS Release 8.1R1. [PR/78877: This issue has been resolved.]
• When you configure real-time performance monitoring (RPM) for BGP, the output of the show services rpm probe-results command might not display all of the information for all routing instances. [PR/78888: This issue has been resolved.]
• On J20 GGSN routers, if you delete the instance-type statement at the [edit routing-instances instance-name] hierarchy level and commit the configuration, the routing protocol process (rpd) might dump core. [PR/79264: This issue has been resolved.]
• For Adaptive Services and MultiServices PICs, you might not be able to transport IPv6 packets larger than 1158 bytes through an IPSec tunnel using default values. As a workaround, increase the maximum transmission unit (MTU) of the IPSec tunnel. [PR/79442: This issue has been resolved.]
• In a virtual private LAN service (VPLS) network, if you use an aggregated Ethernet interface to connect a provider edge (PE) router to a provider core (P) router, VPLS traffic might leak between VPLS instances. Additionally, if you add or remove member links from the aggregate bundle, traffic might be forwarded with the wrong VPN label and be placed into the wrong VPLS instance. [PR/79702: This issue has been resolved.]
• After you issue the request chassis routing-engine diagnostics command with the compact-flash, hard-disk or all option, the memory utilization of the Routing Engine increases significantly. The amount of increase depends on the total size of the files located on the compact flash and the hard disk. [PR/80095: This issue has been resolved.]
• The routing platform currently does not support indexed next-hop load balancing for IPv4 to MPLS next hops. [PR/80261: This issue has been resolved.]
• On MultiServices PICs, if you restart the routing platform when link services intelligent queuing (LSQ) interfaces are carrying a heavy traffic load, the PIC might stop functioning. As a workaround, stop the traffic, restart the router, and then resume the traffic. Similarly, if you configure Network Address Translation (NAT) on the MultiServices PIC when there are more than 30,000 flows per second, the PIC might stop operating. [PR/80940: This issue has been resolved.]

User Interface and Configuration

• If the software cannot find a referenced policy, the routing protocol process (rpd) might dump core. [PR/67098: This issue has been resolved.]
• If you issue the load update command and there are duplicate elements in a BGP import or export policy, the system might resequence those elements in the configuration. [PR/76909: This issue has been resolved.]
• When you use JUNOScript to change or access configuration attributes while the JUNOS software is performing an automatic rollback, the management process (mgd) might dump core. [PR/78791: This issue has been resolved.]
• If a comment is present at the end of an XML configuration file, the commit process might fail, in which case the system issues an “internal communication failure” message. [PR/80464: This issue has been resolved.]
• If you issue the replace pattern command to modify a statement that offers a choice of several values (such as a route-filter in a policy statement) and the new pattern is the same as the old pattern, the management process (mgd) might dump core. [PR/81347: This issue has been resolved.]

Interfaces and Chassis

• When an interface goes down, the resulting system log message might not contain enough information concerning the cause (such as specific alarms for SONET/SDH interfaces). [PR/22776: This issue has been resolved.]
• For link services intelligent queuing (LSQ) interfaces, when you deactivate and reactivate a remote LSQ interface, the show interfaces lsq-fpc/pic/port extensive command might display erroneous counter values for the LSQ bundle. [PR/54855: This issue has been resolved.]
• When a multicast next hop is added and deleted, the packet and byte rate counters are not zeroed out. These counters continue to show the old values. [PR/68074: This issue has been resolved.]
• For Gigabit Ethernet and 4-port Fast Ethernet interfaces installed in J6350 and J4350 Services Routers, MAC address filtering is not supported. [PR/70278: This issue has been resolved.]
• For 2-port Channelized T1/E1 PIMs in J-series Services Routers, BERT is not supported for JUNOS Release 8.1R1. [PR/71435: This issue has been resolved.]
• On M-series and T-series routing platforms, when a link participating in a multilink bundle transitions down and up (for example, when a physical link fails), the multilink bundle might become unresponsive and affect the traffic flowing on all the links in the bundle. [PR/72151: This issue has been resolved.]
• For J4350 and J6350 Services Routers, if you use dynamic end points to establish a large number of IPSec tunnels in a single service set, some tunnels do not become established and some might transition down and up. [PR/76475: This issue has been resolved.]
• On J-series Services Routers, if you modify your configuration to move a manual IPSec tunnel from a routing instance into the main router, traffic might not travel over the tunnel in the main router. [PR/76643: This issue has been resolved.]
• For redundant Link Services Intelligent Queuing (LSQ) bundles configured on Adaptive Services II PICs, if you test a local loopback or a remote loopback on one of the links in the bundle, throughput for all bundles terminating on the same PIC might be impacted and packets might be dropped. [PR/76650: This issue has been resolved.]
• On M120 routers, when an FPC is no longer mapped to a Forwarding Engine Board (FEB) because the FEB was taken offline or removed, the associated interfaces appear to be available, when in fact they are not available and should be marked as unusable. [PR/76846: This issue has been resolved.]
• On Fast Ethernet interfaces installed in J-series Services Routers, if the interface receives traffic with a destination MAC address set to broadcast (ff:ff:ff:ff:ff:ff), it might handle this traffic as keepalive traffic and limit the throughput to ~2500 packets per second. [PR/77376: This issue has been resolved.]
• For intelligent queuing (IQ) interfaces, the output of the show interfaces extensive command incorrectly displays the class-of-service (CoS) queue transmit rates based on the total bandwidth of the interface rather than the shaping rate of the interface. [PR/77483: This issue has been resolved.]
• If you commit a configuration that produces a warning message, the XML display output for that warning might use an <xnm:error> tag instead of the expected <xnm:warning> tag. [PR/78338: This issue has been resolved.]
• For AS PICs configured with redundant link services intelligent queuing (LSQ) interfaces, even when you do not configure aggregated Ethernet interfaces, the Link Aggregation Control Protocol process (lacpd) might consume excessive amounts of CPU time and memory. [PR/78739: This issue has been resolved.]
• For AS PICs or MultiServices PICs configured with redundant link services intelligent queuing (LSQ) interfaces, when the redundant PICs are congested with a high traffic volume, the rlsq- interface might enter a “both down” state. As a workaround, take both PICs offline and bring them back online to clear this state. [PR/78742: This issue has been resolved.]
• For link services intelligent queuing (LSQ) interfaces configured on Adaptive Services and MultiServices PICs, when you reboot the PIC, multilink class counters for active classes might not be cleared and the output of the show interfaces lsq-fpc/pic/port command might display incorrect statistics for those counters. [PR/79080: This issue has been resolved.]
• When you configure the routing platform as data terminal equipment (DTE) for a Frame Relay Local Management Interface (LMI) and a circuit goes down and comes back up (flaps), some data-link connection identifiers (DLCIs) configured on the circuit might remain in the DOWN state. As a workaround, disable and re-enable these DLCIs. [PR/79208: This issue has been resolved.]
• On a 10-port Gigabit Ethernet PIC, if you configure extended-vlan-vpls encapsulation on multiple logical interfaces for one of the ports and then configure a second port with a different encapsulation type, the VPLS enable port might reconfigure the MAC filter and drop the VPLS connections. As a workaround, deactivate the port containing the VPLS logical interfaces and then re-enable it. [PR/81069: This issue has been resolved.]

Services Applications

• When you configure Network Address Translation (NAT) pool addresses with the same prefix as the address for an interface on which a service set has been applied, and you configure NAT in the output direction, the routing protocol process (rpd) might not respond to ARP requests for the NAT addresses. [PR/67901: This issue has been resolved.]
• For link services intelligent queuing (LSQ) interfaces on AS PICs or MultiServices PICs, when you configure an LSQ interface as an independent interface and also as a redundant LSQ primary or secondary interface, the IP Control Protocol process is not enabled. As a workaround, do not configure an LSQ interface individually when the same interface must be part of a redundant LSQ bundle. [PR/73496: This issue has been resolved.]
• For J-series Services Routers, if you configure a real-time performance monitoring (RPM) probe and include the hardware-timestamp statement at the [edit services rpm probe owner-name test test-name] hierarchy level, the round-trip time for a UDP ping does not improve. [PR/75467: This issue has been resolved.]
• When you configure a stateful firewall for an AS PIC or MultiServices PIC and issue the show services stateful-firewall statistics command, the output might display a larger value for the “Reject counters” field than the number of packets actually rejected by the stateful firewall rule. [PR/75593: This issue has been resolved.]
• On Adaptive Services PICs configured with a stateful firewall, a packet with overlapping TCP fragments might cause the PIC to stop operating. [PR/79443: This issue has been resolved.]
• For Adaptive Services PICs configured for Layer 2 Tunneling Protocol (L2TP), when a bundle is deleted and there are outstanding fragments to be transmitted in a Multilink PPP (MLPPP) session, the PIC might stop operating. [PR/80540: This issue has been resolved.]

Routing Protocols

• In a fully meshed BGP network using route reflectors and route target filtering, if there are multiple routes for the same route target or if you configure the advertise-default statement and there is at least one received default route target route, VPN routes might not be sent to all BGP neighbors who requested them. [PR/74517: This issue has been resolved.]
• The output from the show route label-switched-path command might not display all routes that are using an LSP as a next hop. [PR/78875: This issue has been resolved.]
• If there are multiple BGP routes to the same prefix which are received from different neighbor autonomous systems and which differ only in their IGP cost to the next hop, BGP might select the wrong route as the best route. [PR/78975: This issue has been resolved.]
• When you configure Virtual Router Redundancy Protocol (VRRP) on an interface in a routing instance, and you include the accept-data statement at the [edit interfaces interface-name unit logical-unit-number family inet address ip-address vrrp-group group-number] hierarchy level, and if no router ID is configured at the [edit routing-instances instance-name routing-options] hierarchy level in the routing instance, the VRRP virtual IP address might be selected as the router ID for the routing instance. [PR/79825: This issue has been resolved.]

MPLS Applications

• If the traffic rate for an ingress MPLS LSP exceeds 2 Gbps and you issue the monitor label-switched-path command, the “Bits per second (bps)” field in the output might display a negative value. [PR/77924: This issue has been resolved.]
• When you configure the graceful-restart statement at the [edit routing-options] hierarchy level and the routing protocol process (rpd) restarts twice in a short time period, the routing protocol process might dump core. [PR/78539: This issue has been resolved.]
• When an RSVP session goes down and comes back up (flaps), LSP counters might not be reset to zero. [PR/79295: This issue has been resolved.]
• If you issue an MPLS traceroute command and the immediate downstream transit node in the LSP path is an M320 or T-series routing platform with an Enhanced FPC, the traceroute might not work if the transit node uses equal cost multipath (ECMP). [PR/80089: This issue has been resolved.]

VPNs

• If you configure two or more MPLS LSPs, create a VPLS instance, and configure a policy to send the VPLS traffic over one of the LSPs, and if the policy contains the from rib mpls.0 match condition and is applied at the [edit routing-options forwarding-table export] hierarchy level, the policy might not match on the VPLS routes. [PR/80585: This issue has been resolved.]

Class of Service

• On M320 and T640 routers with Gigabit Ethernet IQ2 PICs, egress class-of-service queue drops might be not be commensurate with transmit rates. As a workaround, configure the delay-buffer-rate statement at the [edit class-of-service traffic-control-profiles profile-name] hierarchy level, or configure the q-pic-large-buffer statement at the [edit chassis fpc slot-number pic pic-number] hierarchy level. [PR/76994: This issue has been resolved.]
• For redundant link services intelligent queuing (LSQ) interfaces configured on AS PICs, if you issue the show interfaces queue command for the multilink PPP (MLPPP) bundles, the output does not show random early detection (RED) drops broken down into individual drop profiles (such as low and high). [PR/77059: This issue has been resolved.]

Forwarding and Sampling

• If you configure Routing Engine-based sampling on a Monitoring Services or Adaptive Services II PIC, the output of a cflowd record might show the destination port field for Internet Control Message Protocol (ICMP) as 0. As a workaround, configure PIC-based sampling. [PR/70906: This issue has been resolved.]
• When port is used with either source-port or destination-port in the same term of a match condition in a firewall filter, the system does not generate an error when a commit is issued. [PR/76674: This issue has been resolved.]
• When a forwarding next hop is changed from the Ethernet Management interface (fxp0) to a transit interface, an indirect next hop is created for each route that shares the same next hop. As a workaround, do not use the Ethernet Management interface as a forwarding next hop for transit traffic. [PR/80139: This issue has been resolved.]
• On T-series routing platforms, if you include the from traffic-class match condition in an IPv6 firewall filter, the filter might not work as expected. [PR/80359: This issue has been resolved.]

Network Management

• On AS PICs, if you change or rename a service set, the output of the show snmp mib walk jnxSpSvcSet command does not display the correct service set name. As a workaround, issue the restart adpative-services command before reissuing the show snmp mib walk jnxSpSvcSet command. [PR/78130: This issue has been resolved.]
• The values returned for the ifChassisFpc MIB might not match the corresponding index values used in the Juniper Networks enterprise chassis MIB (for example, jnxContentsDescr). [PR/78429: This issue has been resolved.]
• When a large number of trap packets are sent in a short interval and the packets are kept in the trap queue, the SNMP process (snmpd) runs out of memory. [PR/78727: This issue has been resolved.]

Copyright © 2007, Juniper Networks, Inc. All Rights Reserved. Trademark Notice.