Known Problems and Limitations

This section identifies the known problems and limitations in this release. For more information about known problems that were discovered at customer sites, you can log in to the JunosE Knowledge Base at https://www2.juniper.net/kb/, enter the defect ID number in the Search by Keyword field, and click Search. Problems that have not been reported by customers are documented only in these Release Notes.

ATM

When 16,000 PPPoA interfaces are configured on an OCx/STMx ATM line module paired with an OC3-4 I/O module in an ERX14xx model, ERX7xx model, or ERX310 router, Ping traffic passing through the line module on the restarting router experiences an outage of 103 seconds, which is beyond the maximum limit, after a unified ISSU from JunosE Release 9.2.0p1-0 to 9.3.0p0-12. This outage does not occur when the same configuration is applied on a Gigabit Ethernet interface. [Defect ID 179794]
When a mirror rule that triggers on username is employed for packet mirroring of dynamic IP subscribers over ATM, removal of the rule does not disable packet mirroring. [Defect ID 175356]

Work-around: Use a mirror rule that triggers on account session ID rather than on username.

BFD

After you have shut down the interface to the next hop (for the route that is used to establish the BFD session), output for the show bfd session command erroneously indicates the shutdown interface as Management Interface (FastEthernet 6/0). [Defect ID 174271]

DHCP

A memory leak is observed on the SRP module when subscriber sessions are flapped in an environment in which 48,000 DHCP proxy client bindings are established. [Defect ID 189488]
When 32,000 DHCPv6 subscribers are brought up over multiple virtual routers at a call setup rate (CSR) of 24 calls per second, the DHCPv6 local server stops responding. This problem occurs due to a deadlock between two instances of the DHCPv6 local server running on the router while the DHCPv6 client entry is being updated. [Defect ID 189282]
When approximately 10 DHCP subscribers are connected over Agent Circuit Identifier-based (ACI) VLAN subinterfaces, a few of the DHCP requests from clients are returned with negative acknowledgment responses from the DHCP local server. This problem occurs only on ES2 10G LMs and ES2 10G ADV LMs, and not on ES2 4G LMs. [Defect ID 189336]
DHCP packets are not forwarded to the DHCP server over dynamically created interfaces when all of the following are true: [Defect ID 180343]

DHCP relay or DHCP relay proxy is configured on the router.
The client-facing interfaces are created dynamically using bridged Ethernet over static ATM PVCs.
The ip auto-detect ip-subscriber command is configured to enable packet detection (packet triggering) and to trigger creation of dynamic subscriber interfaces.

Work-around: To avoid this defect, do all of the following:

Do not use the ip auto-detect ip-subscriber command to enable packet triggering and to create dynamic subscriber interfaces.
Ensure that DHCP external server is configured in the virtual router.
Ensure that the set dhcp relay inhibit-access-route-creation command is configured in the virtual router to prevent DHCP relay from installing host routes by default.

DHCP External Server

When DHCP relay and DHCP external server are configured in the same VR with server-sync enabled, bindings are not created in the DHCP external server when DHCP clients on an ATM bulk configuration interface stack and dynamic VLAN over Ethernet stack sends a renew message. [Defect ID 87087]
The DHCP renew counter and release counter (displayed with the show ip dhcp-external statistics command) are doubled rather than incremented for each renew and release sent. [Defect ID 78802]

DoS Protection

A Telnet session closes when sending ipLocalBGP protocol traffic at a rate in the range 4096–4200 packets per second (pps) with suspicious control flow detection enabled. [Defect ID 81974]

Work-around: When the traffic drops below 4096 pps, open a new Telnet session.

Forwarding

When performing MAC validation to match subscriber demux entries with ARP host entries, the ES2 10G LM does an exact match, rather than a longest prefix match. The subscriber demux entry source address must be a /32 value matching the IP address of an ARP entry in order to validate the MAC address against that ARP entry. [Defect ID 79641]
On an E120 or E320 router that is running JunosE Release 11.2.1, the ES2 10G LM with an ES2-S2 10 GE PR IOA stops egress forwarding because of an egress IXP static RAM (SRAM) parity error. This problem is observed during the removal of the LM from its slot and reinsertion into the same slot when . This behavior also happens when about 6000 subscribers are brought up and user traffic is transmitted. This problem depends on the type of user traffic sent. [Defect ID 189846]
When you attach certain hierarchical policies to subinterfaces as input policies, secondary input policies, and output policies, incoming traffic loss can occur when the number of subinterfaces to which the policies are attached exceeds 4600. [Defect ID 86741]
An ES2 10G LM resets when you attempt to remove a multicast port in a dual-stack environment that contains both IPv4 and IPv6 subscribers. [Defect ID 189867]
When PPPoE over LAG is configured on an interface, and you re-execute the PPPoE-over-LAG configuration before you delete the previous configuration, the ES2 10G LM line module resets. [Defect ID 179639]

Work-around: Before you can re-execute the PPPoE-over-LAG configuration, delete the existing PPPoE-over-LAG configuration.

Specifying S-VLAN ranges that partially overlap does not work. [Defect ID 81918]

For example, the following configuration fails because S-VLAN 22 falls within the previously specified S-VLAN range of 21–23.

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 21 23 401 426

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 21 23 427 712

host1(config-if)#vlan bulk-config BulkCezarCnfg2 svlan-range 22 22 101 110

Work-around: You can do either of the following to avoid this problem.

Specify each S-VLAN within the partially overlapping range as individual S-VLANs, as in the following example:

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 21 21 401 426

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 22 22 401 426

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 23 23 401 426

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 21 21 427 712

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 22 22 427 712

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 23 23 427 712

host1(config-if)#vlan bulk-config BulkCezarCnfg2 svlan-range 22 22 101 110

Use fully overlapping ranges rather than partially overlapping ranges, as in the following example:

host1(config-if)#vlan bulk-config BulkDHCPCnfg1 svlan-range 21 23 401 426

host1(config-if)#lan bulk-config BulkDHCPCnfg1 svlan-range 21 23 427 712

host1(config-if)#vlan bulk-config BulkCezarCnfg2 svlan-range 21 23 101 110

Ethernet statistics are incorrectly displayed for virtual port 8 of the ES2-S1 GE-8 IOA when that module is paired with the ES2 10G LM or the ES2 10G Uplink LM. [Defect ID 174784]
The ES2 10G LM does not support framed routes configured for dynamic subscriber interfaces. [Defect ID 83154]
On the ES2 10G LM, a VLAN ID of 0 assigned to an interface can prevent packets from being properly forwarded. [Defect ID 176125]
For IP and VLAN policies attached to VLAN subinterfaces on ES2 10G LMs and ES2 10G Uplink LMs, the output policy counters for outgoing control and exception packets are incorrectly displayed in the output of the show ip interface and show vlan subinterface commands. These counters are not incremented correctly in the VLAN policy output section of the output of the show vlan subinterface command and in the IP policy output section of the output of the show ip interface command. [Defect ID 190083]
In a scenario in which approximately 2000 GRE tunnels are configured, filter rules are configured in the policy lists to filter traffic on 1000 tunnels, and forward rules are set up in the policy list to forward traffic from 1000 tunnels, 300 Mbps of traffic is initially filtered. A traffic drop is observed for a brief period and traffic filtering resumes shortly thereafter. Gradually, traffic filtering recedes and stops completely, enabling all of the traffic over the GRE tunnels to start flowing. [Defect ID 190203]

ICR

If you saved the running configuration of the router as a script file (.scr) and execute the script to apply the settings on the router, ICR partition configuration commands in the .scr file might fail to add group members to the partition. This problem happens when the subscriber configuration in the .scr file is placed before the ICR partition configuration. However, this problem does not occur if you used a system configuration (.cnf) file to set up the router. [Defect ID 183913]

Work-around: To correct this problem and enable ICR partitions to be created correctly, make sure that you add the ICR partition configuration before the subscriber interface configuration in the .scr file. You can perform this reordering by modifying the .scr file to place the commands that configure subinterfaces for ICR partitions before the commands used for VLAN-based or S-VLAN-based grouping of subscribers.

IGMP

The E Series router IGMPv3 proxy does not operate correctly in the presence of IGMPv2 queriers. [Defect ID 46039/46045]

Work-around: If an IGMPv2 router is present on the network, do not configure version 3 with the ip igmp-proxy version command on that network interface. (Version 2 is the default.)

The default value for the IGMPv3 proxy unsolicited report interval timer should be 1 second rather than 10 seconds (the value for v2). [Defect ID 46040]
IGMPv3 proxy is not supported. [Defect ID 46038]

IS-IS

On a router configured with IS-IS and BFD, using the redundancy force srp command to force an SRP switchover sometimes brings down IS-IS and BFD. [Defect ID 179287]
IS-IS graceful restart (nonstop forwarding) does not work on the broadcast interface when the restarting router is the designated intermediate system (DIS). Graceful restart works properly when the restarting router is not the DIS. [Defect ID 61496]

L2TP

If you perform a unified ISSU operation on an E120 router or an E320 router that contains two pairs of line modules configured for stateful line module switchover and functions as an LNS device, the SRP module resets during the unified ISSU process. This problem occurs when any one of the following conditions are met: [Defect ID 186910]

A certain number of L2TP subscribers are already connected to the router and more subscriber sessions are attempted to be established during the unified ISSU process.
The logged-in L2TP subscribers are logged out and the subscriber sessions are attempted to be reestablished.
After the initialization phase of the unified ISSU process is started and completed, a stateful line module switchover is performed and another unified ISSU process is performed while more subscribers are logging in.

After a unified ISSU completes on a router functioning as an L2TP access concentrator (LAC), traffic outages occur on the L2TP network server (LNS)-facing interface at the LAC in a configuration with 16,000 or 32,000 L2TP sessions over 500 tunnels. [Defect ID 180147]
Approximately 25 percent of the total number of L2TP subscriber sessions are terminated and reestablished after a long time (about 25 minutes for 8000 sessions) when an ATM line module on a router that functions as the LAC device is reloaded. [Defect 187515]
When you perform a stateful SRP switchover procedure on an LNS device that contains an ES2 4G LM with Service IOA (tunnel server module), some of the 16,000 subscriber sessions over 16,000 tunnels that are established are terminated. This problem occurs when OSPF is used as the routing protocol between the LAC and LNS devices in the L2TP tunnel, and with the number of L2TP retransmission attempts configured as 10. [Defect ID 187358]

LDP

Some of the LDP sessions on ATM line modules do not come up when you perform a unified ISSU operation on an ERX router, which is the restarting router, from JunosE Release 10.3.3 to Release 11.0.3. This problem occurs when an ERX router functions as the restarting router and an E120 or E320 router functions as the helper router. [Defect ID 189588]

MLD

MLDv2 proxy is not supported. [Defect ID 46038]
The E Series router does not log a warning when it receives an MLDv1 query but is not configured to use MLDv1 on the interface. [Defect ID 46046]
The E Series router MLDv2 proxy does not operate correctly in the presence of MLDv1 queriers. [Defect ID 46039/46045]

Work-around: If an MLDv1 router is present on the network, configure version 1 with the ipv6 mld-proxy version command on that network interface. (Version 2 is the default.)

The default value for the MLDv2 proxy unsolicited report interval timer should be 1 second rather than 10 seconds (the value for v1). [Defect ID 46040]

MPLS

When you issue a traceroute or trace mpls command to trace the paths of router packets over MPLS interfaces on an ES2 10G LM or ES2 10G Uplink LM, the results include an extra unknown host. [Defect ID 174537]
If LSPs are announced into IS-IS, then the IS-IS routes cannot be used for multicast RPF checks, because LSPs are unidirectional. [Defect ID 28526]

Work-around: Configure static RPF routes with native hops when LSPs are autoroute announced to IGPs.

When the IPv4 explicit null label appears anywhere other than at the bottom of the label stack, TTL expiration for this label is not handled correctly. As a result, the traceroute command does not work correctly for LSPs that have the IPv4 explicit null label anywhere other than at the bottom of the label stack. [Defect ID 76037]
In a scaled environment with a large number of MPLS RSVP-TE tunnels configured, the states of the hello adjacency instances in the State field in the output of the show mpls rsvp hello instance command are displayed as Down for loopback interfaces. The correct behavior is that the RSVP-TE hello adjacencies must always be in the Up state for loopback interfaces. [Defect ID 189565]

Multicast

When you configure more than 10,219 outgoing interfaces (OIFs) on the same ES2 10G LM in a single multicast group, the configuration of the multicast group's OIF membership from the SRP module to the line module exceeds the size of a single message and is sent in fragments.

Because of this fragmentation, the ES2 10G LM generates the following error message: [Defect ID 81768]

pc: 0x9e5c88: -> fatalPanic(void) offset: 0x8

Policy Management

On the E320 router, redirecting a large configuration with thousands of interfaces to a script file can take a long time, perhaps exceeding a half-hour depending on the configuration. [Defect ID 80429]
When an MD-Port-Number value greater than 65,535 is sent to an E120 or E320 router by means of a COA request, the value that is displayed in the UDP header of mirrored packets is the actual value minus 65,536. For example, an MD-Port-Number of 65,540 is displayed in the mirrored packet as 4. [Defect ID 84712]
When you reload the slot holding a GE-2 or GE-HDE line module and you have configured more than about 2000 policies with rate limiting on that module, the drop count becomes more than expected. This unexpected drop count does not occur when you create the same configuration after you reload the router to the factory-default configuration. [Defect ID 175696]
On the E120 and E320 routers, when a mirror rule is deleted after a CoA request is sent with Juniper-LI-Action set to No-Action, the existing mirroring session is not disabled. [Defect ID 84826]
On E320 line modules that support secure policies, the SRP module enables you to configure more than 1022 secure policies per module. [Defect ID 175756]

Work-around: To avoid potential performance issues, we recommend that you do not configure more than 1022 secure policies per module.

When you modify a rate-limit profile in Global Configuration mode after the system is in a scaled state, changes to the rate-limit profile fail owing to lack of adequate policy resources. However, the changed value of the rate-limit profile is displayed in the output of the show rate-limit profile command. [Defect ID 79342]

Work-around: To avoid this problem, do not update the rate-limit profile in Global Configuration mode in a scaled environment.

When you enter the no ip policy-parameter hierarchical parameterName command or no ipv6 policy-parameter hierarchical parameterName command for a hierarchical policy-parameter type in Interface Configuration mode, the explicit reference of the parameter is removed successfully from the interface. However, the Referenced by interfaces field in the output of the show policy-parameter command does not change from the previously configured value to implicit. [Defect ID 183957]

Work-around: To correct this problem, remove the entire interface configuration.

When you perform a stateful SRP switchover with high availability in the enabled state and with approximately 5000 dual-stack subscriber sessions, independent IPv4 sessions, or independent IPv6 sessions established on the router, the following log message is recorded for the policyMgrGeneral system logging category: [Defect ID 186570]

ERROR 09/26/2011 22:30:43 policyMgrGeneral: Error restoring policy attachment for 480926 from MS/NVS

This problem occurs when the router configuration contains GRE tunnels, IPv4 secure policies, and IPv6 secure policies, and packet mirroring is enabled using username as the trigger. This problem might also happen during unified ISSU.

PPP

On a pair of line modules configured for redundancy, if you disable the primary module in the redundancy group, previously established L2TP sessions are not reconnected. This problem occurs with PPP over ATM interfaces and 16,000 L2TP subscriber sessions. However, this problem does not occur with static and dynamic PPPoE over ATM subinterfaces. [Defect ID 187085]

PPPoE

The E Series router erroneously accepts a PADI with a payload length of 0 instead of rejecting it and incrementing the PPPoE Invalid PAD packet length counter. [Defect ID 48356]

QoS

The dynamic shaping rate calculated by the simple shared shaper can vary because of the variation in the enqueue rate of the constituent queues. Even when the offered load is constant, the mechanism that calculates the enqueue rate introduces a slight variation, introducing a slight variation in the calculated dynamic shaping rate. [Defect ID 80938]
The no qos-parameter-define definition command does not delete the specified QoS parameter definition. [Defect ID 176844]

Work-around: Remove the interface and add the desired QoS parameters when you re-create the interface instead of deleting the definition.

The compound shared shaping feature does not work properly on egress forwarding ASIC 2 (EFA2)-based ATM line modules when the shared shaper is queue-controlled as opposed to node-controlled. In a node-controlled configuration, in which you configure the shared-shaping rate on the best-effort scheduler node for the logical interface, integration of the EFA2 and ATM segmentation and reassembly (SAR) schedulers functions properly. However, in a queue-controlled configuration, in which you configure the shared-shaping rate on the best-effort queue for the logical interface, integration of the EFA2 and ATM SAR schedulers does not function properly. [Defect ID 69167]

Work-around: Use node-controlled compound shared shaping configured on the best-effort scheduler node with EFA2-based ATM line modules.

Simple shared shaping does not function correctly when it is used for 32,000 subscribers on an ES2 10G ADV LM. However, when you change the shaper to compound shared shaping, it works properly. Also, simple shared shaping does not function correctly for 16,000 subscribers on an ES2 10G ADV LM. [Defect ID 183512]
When 32,000 subscribers with 128,000 QoS queues are brought up on an ES2 10G or ES2 10G ADV LM, the LM resets if you modify the QoS profile that contains the best-effort IP or VLAN node rule, which references a scheduler profile configured with shared shaping rate, to a scheduler profile configured with legacy shaping rate. [Defect ID 183291]

Work-around: To avoid this problem, apply shared shaping on the best-effort queue, instead of on the best-effort node.

When QoS resources such as failure nodes and statistics bins are exhausted because of insufficient memory available on the line module, the failures are properly logged, but additional log messages are generated every 10 minutes that report zero failures. [Defect ID 85105]
On a router that has both an ES2 10G LM and an ES2 4G LM installed, the byte count reported by the show fabric-queue egress-slot command is incorrect. The reported packet count is correct. [Defect ID 80965]
When you configure an E120 or E320 router with an ES2 10G ADV LM as a LAC on one side of an L2TP tunnel and as a LNS to receive packets from the LAC on the other side of the tunnel, use RADIUS servers for authentication of subscribers on both sides of the tunnel, and attempt to bring up 16,000 subscribers on the L2TP tunnel, the LM that has subscribers on the LAC side of the tunnel resets when approximately 8000 logged-in subscribers are logged out and try to reestablish the connection. [Defect ID 184118]

RSVP-TE

After stateful SRP switchover, forwarding of VPN traffic might not resume if the core interface that carries an MPLS base tunnel with LDP over RSVP-TE flaps (constantly goes up and down). [Defect ID 182019]

SDX Software and SRC Software

If you perform a stateful SRP switchover operation on an E320 Broadband Services Router with an SRP-100 module that acts as an LNS device, the rate-limit profile that is applied from the SRC client does not take effect on the already logged-in subscribers. This problem occurs when all of the following conditions are satisfied:[Defect ID 189540]

Stateful line module switchover is activated on the LNS device.
Rate-limit profile is associated with the output policy applied on the subscriber interfaces for L2TP sessions.
The output policy is configured using the SRC client on the router to the subscribers while they are logging in.

If you modify the rate-limit profile attached as the output policy on the subscriber interface and apply the policy using the SRC client after the subscribers have logged in or after a stateful line module switchover is completed, the change in rate-limit profile takes effect for the subscribers.

Server Card Manager (SCM)

High availability mode transitions to the pending state when you perform the following steps. The high availability state of the system is displayed in the output of the show redundancy detail command.

Configure a shared tunnel-server port on an ES2 4G line module that functions as the primary in a redundancy group of line modules.
Bring up a GRE tunnel on the primary line module.
Perform a line module redundancy operation to switch over from the currently active primary to the standby module.

When the system is in the pending state, the SCM application running on the router becomes unsupported for 5 minutes, and then it returns to the active state. The client field in the output of the show redundancy clients command displays the status of the SCM application. [Defect ID 188489]

Service Manager

After you activate an independent IPv6 service and issue either of the following commands on the default virtual router or any other virtual router, except the one on which the subscriber session is active, no output is displayed in the CLI interface: [Defect ID 181929]

show service-management subscriber-session subscriberName interface interfaceType interfaceSpecifier
show service-management subscriber-session subscriberName interface interfaceType interfaceSpecifier service-session serviceName

This problem also occurs when a subscriber is authenticated using a RADIUS server for a combined IPv4 and IPv6 service in a dual stack.

Work-around: To avoid this problem, use the show service-management owner-session ownerName ownerId command to display subscriber session information based on the session owner, instead of the show service-management subscriber-session subscriberName interface interfaceType command to display details on subscriber sessions.

Activation of service sessions for a subscriber with DHCPv6 over IPv6 bindings using the CoA method that uses RADIUS Change-of-Authorization-Request (CoA-Request) messages and VSAs does not work if the service session was previously activated using the RADIUS login method that uses Access-Accept messages and VSAs. However, this problem does not occur for IP subscriber service sessions. Also, this problem does not occur if service sessions for subscribers with DHCPv6 over IPv6 bindings are activated only using the CoA method. [Defect ID 189403]

SRC Software and SDX Software

When multiple IPv6 interfaces are configured with policies attached from SRC, only some of the IPv6 interfaces have the policies attached. [Defect ID 179498]
Changing the SSCC status (enable/disable) while IPv6 interfaces are configured might cause the SRP to reset. [Defect ID 179537]

Stateful SRP Switchover (High Availability) and IP Tunnels

A packet loss sometimes occurs during stateful SRP switchover when you use the ping command on a router that is configured for OSPF graceful restart, and is connected to a helper router in the OSPF IPv6 broadcast network and another helper router in the OSPF IPv6 backbone area. [Defect ID 181470]

ERX7xx model, ERX14xx model, or ERX310 router:

When you use the ping command with the IPv6 address of the helper router in the multicast area as the destination address and the loopback address of the helper router in the backbone area as the source address, a packet loss of 2 seconds occurs for the first stateful SRP switchover. However, no packet loss occurs for successive stateful SRP switchovers.
When you use the ping command with the IPv6 address of the helper router in the broadcast network as the destination address and no source address when stateful SRP switchover is performed the first time, an identical packet loss occurs. In this case too, no packet loss occurs during subsequent switchovers.

E120 router or E320 router:

When you use the ping command with the IPv6 address of the helper router in the broadcast network as the destination address and the loopback address of the helper router in the backbone area as the source address, no packet loss occurs.
When you use the ping command with the IPv6 address of the helper router in the multicast area as the destination address and no source address, a packet loss of 1–2 seconds sometimes occurs during stateful SRP switchovers.

Subscriber Management

When a dynamic GRE tunnel interface for Mobile IP relocates between SM modules because the original SM reloads, Mobile IP deletes the relocated tunnel interface. [Defect ID 178399]
When a subscriber has subscribed for a service, service session accounting records always contains a default Acct-Terminate-Cause value of 10. This value remains unchanged even after you use the terminate-code command to configure a custom mapping between application terminate reasons and RADIUS Acct-Terminate-Cause attributes. [Defect ID 181043]
Dynamic subscriber interfaces continue to remain in the down or not present operational state in either of the following scenarios: [Defect ID 81269]

If you configured a dynamic interface column, such as a dynamic bridged Ethernet interface, dynamic VLAN interface, or an ATM interface, and when any one of the following conditions is satisfied:

The major interface is bounced (shut down and reenabled)
The major interface is shut down, which cause the dynamic VLAN interfaces to be removed
The physical link goes down and comes back up
The line module is removed and reinserted

If you configured a static interface column and removed the major interface

These scenarios might occur if you administratively issue the shutdown and no shutdown commands on the major interface in which the dynamic interface column is configured.

Work-around: Use the no interface ip ipAddress command to remove the dynamic subscriber interfaces. Although you can use the dhcp delete-binding command to remove the DHCP binding and the dynamic subscriber interfaces, the DHCP client does not detect the binding removal and retains the lease.

System

Memory leak is observed with the SRP-100 module while subscribers are being brought up on a LAC device and the active link between the LAC device and the LNS device in an L2TP tunnel is flapping. This problem occurs when the following steps are performed: [Defect ID 189353]

Two redundant links connect the LAC device to the LNS device in the L2TP tunnel.
DHCPv6 subscribers over PPPoE interfaces connected to a LAC device are attempted to be brought up.
The active link between the LAC and LNS devices flaps continuously 1000 times using the shutdown and no shutdown commands.
Memory-related output information is collected at a base condition where the active link is up again and no subscriber is connected to the router.

When you perform each iteration of the preceding four steps, the amount of free memory on the SRP-100 module decreases and validates a memory leak.

TCP

The SRP module resets in any of the following circumstances on an E320 router that has a line module configured with 5000 ANCP adjacencies: [Defect ID 176916]

When you issue the issu initialization command from the console and then reload the line module from a Telnet session.
When the client that has the 5000 ANCP clients resets or an intermediate switch resets.
When you reload the line module.

Unified ISSU

Unified ISSU is not supported with 8000 bridged Ethernet interfaces on an OC3/STM1 GE/FE ATM line module. [Defect ID 178811/178797/
179547]
ATM line modules might reset after a unified ISSU when you attempt to add memory to a VLAN subinterface in a large bridged Ethernet configuration. [Defect ID 178798]
During the unified ISSU operation, if you modify the router configuration after the initialization phase of the process is completed and before you issue the issu start command to commence the upgrade phase of the unified ISSU process, the unified ISSU procedure completes successfully and the stateful SRP switchover process begins to synchronize between the active and standby SRP modules. When the synchronization process is in progress, the standby SRP module reloads for the second time. After the second reload of the standby SRP module ends, the synchronization process also ends properly.

Although the standby SRP module reloads for the second time when it is synchronized with the upgraded release, normal router operations, such as handling of subscriber sessions and forwarding of traffic, remain unaffected. [Defect ID 185517]

JunosE 12.2.1 Release Notes