Application Support for Stateful SRP Switchover

Applications are either supported or unsupported by stateful SRP switchover.

Table 8 indicates which applications support or do not support stateful SRP switchover.

Application Support

Table 8: Application Support for Stateful SRP Switchover

Application

Supported

Unsupported

Notes

Physical Layer Protocols

DS1

DS3

HDLC

SONET/SDH

SONET/SDH VT

Link-Layer Protocols

ATM

Static and dynamic interfaces, with the exception of ATM subscribers, are supported.

In this case, ATM subscribers refers to a technology on the E Series router where the ATM layer does authentication (that is, not PPP or IP subscriber manager).

ATM 1483 bulk configuration of dynamic interfaces

Bridged Ethernet

Cisco HDLC

Ethernet (with and without VLANs)

Frame Relay

PPP

PPPoE

Transparent bridging

Unicast Routing

Access Routes

BFD

During a stateful SRP switchover, the BFD transmit interval is set to 1000 ms with a detection multiplier of 3. These values result in a liveness detection interval of 3000 ms. This longer interval helps prevent a BFD timeout during the switchover. BFD negotiates the interval with the remote peer before applying the temporary change. The BFD timers revert back to the configured values after 15 minutes (the maximum duration for graceful restart completion).

BGP

Supported for IPv4 only when the graceful restart extension is enabled. Does not support graceful restart for IPv6 address families.

FTP

Static recovery support only.

IP

IPv6

IPv6 neighbor discovery

IPv6 neighbor discovery characteristics are replayed during switchover. Line modules do not flush neighbor discovery information during the switchover.

IPsec Transport

IPsec Tunnels

Completed IKE phase 1 and phase 2 negotiations supported only.

IS-IS

Supported only when the graceful restart extension is enabled.

IS-ISv6

Supported only when the graceful restart extension is enabled.

OSPFv2

 

Supported only when the graceful restart extension is enabled.

OSPFv3

Supported only when the graceful restart extension is enabled.

RIP

Static recovery support only.

Static Routes (IP and IPv6)

After all high-priority interfaces have been replayed from NVS and mirrored storage, static routes are replayed from NVS, followed by replay of low-priority interfaces from NVS and mirrored storage. This behavior enables static routes that are dependent on high-priority interfaces to be resolved quickly and installed in the IP routing table.

Telnet

Static recovery support only.

IPv4 Multicast Routing

Multicast Routing

Stateful SRP switchover. During switchover, the system mirrors the multicast queue so that IP can use the same queue without needing to recreate a different connection. The multicast queues are also preserved during the switchover and graceful restart period to ensure that multicast data continues to be forwarded using the previously learned multicast forwarding state.

DVMRP

Static recovery support only. DVMRP gives the restart complete indication to the IP routing table after getting a peer update (60-second timeout).

IGMP

IC IGMP deletes its interface and membership state on SRP failover (controller down). As part of SRP warm start, IGMP interfaces are reconfigured from NVS and dynamic IGMP interfaces are reconfigured from mirrored storage. IGMP hosts are queried as IP interfaces come back up, the join state is re-established, and SC IGMP state is created.

After the maximum query response time (across all interfaces) expires to allow hosts to re-establish join state, IGMP notifies MGTM that graceful restart is complete.

MGTM

On SRP failover, old mroutes are retained on the line module to preserve multicast forwarding; cache-misses to the SRP are disabled. When MGTM warm starts on the SRP, it reads the NVS configuration and enables multicast routing. When IGMP, DVMRP, and PIM have completed graceful restart and the IP route table multicast-view has completed graceful restart, old mroutes are deleted from the line module and cache-misses to the SRP are enabled. This triggers re-creation of mroutes and establishes the current multicast forwarding state.

Although cache-misses to the SRP module are disabled, forwarding is preserved for old multicast joins to downstream routers and hosts. However, forwarding for new multicast joins requested by downstream routers and hosts after SRP module switchover is not provided until cache-misses are re-enabled.

PIM

Static recovery support only. For warm start, PIM interfaces are reconfigured from NVS. A Hello message with a new Generation ID is issued as IP interfaces come up. A neighbor that receives this Hello determines that the upstream neighbor has lost state and needs to be refreshed. A VR-global configurable graceful restart timer is required for PIM to time out the re-establishment of the join state for sparse-mode interfaces. After this timer expires, PIM notifies MGTM that graceful restart is complete.

IPv6 Multicast Routing 

Multicast Routing

Stateful SRP switchover. During switchover, the system mirrors the multicast queue so that IP can use the same queue without needing to recreate a different connection. The multicast queues are also preserved during the switchover and graceful restart period to ensure that multicast data continues to be forwarded using the previously learned multicast forwarding state.

MGTM

On SRP failover, old mroutes are retained on the line module to preserve multicast forwarding; cache-misses to the SRP are disabled. When MGTM warm starts on the SRP, it reads the NVS configuration and enables multicast routing. When MLD and PIM have completed graceful restart and the IPv6 route table multicast-view has completed graceful restart, old mroutes are deleted from the line module and cache-misses to the SRP are enabled. This triggers re-creation of mroutes and establishes the current multicast forwarding state.

Although cache-misses to the SRP module are disabled, forwarding is preserved for old multicast joins to downstream routers and hosts. However, forwarding for new multicast joins requested by downstream routers and hosts after SRP module switchover is not provided until cache-misses are re-enabled.

MLD

IC MLD deletes its interface and membership state on SRP failover (controller down). As part of SRP warm start, MLD interfaces are reconfigured from NVS and dynamic IMLD interfaces are reconfigured from mirrored storage. MLD hosts are queried as IPv6 interfaces come back, the join state is re-established, and SC MLD state is created. After the maximum query response time (across all interfaces) expires to allow hosts to re-establish join state, MLD notifies MGTMv6 that graceful restart is complete.

PIM

Static recovery support only. For warm start, PIM interfaces are reconfigured from NVS and a Hello message with a new Generation ID is issued as IPv6 interfaces come up. A neighbor that receives this Hello determines that the upstream neighbor has lost state and needs to be refreshed. A VR-global configurable graceful restart timer is required for PIM to time out the re-establishment of the join state for sparse-mode interfaces. After this timer expires, PIM notifies MGTM that graceful restart is complete.

Multiprotocol Label Switching

MPLS

MPLS is HA-unsafe during a graceful restart. It is HA-unsafe until all the configured MPLS signaling protocols have completed their graceful restart procedures and any stale forwarding elements have been flushed from the line modules.

If you force an SRP switchover while MPLS is HA-unsafe, the SRP module switches but the SRP module and the line modules undergo a cold restart.

If the primary SRP module resets while MPLS is HA-unsafe, the router undergoes a cold restart.

MPLS over IPv6 supports HA. This functionality enables BGP to support graceful restart for IPv6 labeled addresses.

BGP signaling

LDP signaling

To provide uninterrupted service during an SRP switchover in a scaled configuration, such as one with 32,000 Martini circuits, set the LDP graceful restart reconnect time to the maximum 300 seconds and set the LDP graceful restart recovery timer to the maximum 600 seconds. This requirement is true for all SRP switchovers, including those in the context of a unified in-service software upgrade.

LDP signaling does not support HA for IPv6.

RSVP signaling

Local cross-connects between layer 2 interfaces using MPLS

Policies and QoS

Policies

QoS

Static recovery support only.

Remote Access

AAA

DHCP External Server and Packet Trigger

Following a switchover, the DHCP lease (that is, time remaining) is recalculated based on when the lease started. When the release timer for a client expires, the client is deleted and the access route is removed, along with the dynamic subscriber interface if it was created. If the client requests a new lease, DHCP external server resynchronizes with the new lease time.

DHCP Packet Capture

DHCP Proxy Client

When the standby SRP module takes over as the primary after a stateful SRP switchover operation, it continues to handle DHCP lease renewal requests from existing clients based on their states and processes state transitions without any disruption. For more information, see "Preservation of DHCP Proxy Client Bindings During Stateful SRP Switchover"

DHCP Relay Proxy

 

DHCP Relay Server

Before HA support, clients identified by the DHCP relay server were maintained on a switchover (their state was stored to NVS); DHCP relay server always had some level of HA support.

Currently, following a switchover, the DHCP lease (that is, time remaining) is reset. When the release timer for a client expires, the client requests a new lease. The E Series router DHCP relay server then synchronizes with the new state.

DHCPv4 Local Server

DHCPv6 Local Server

DHCPv6 now supports stateful SRP switchover (high availability).

After SRP warm switchover, the router restores the client bindings from the mirrored DHCPv6 information as it does for other applications that support stateful SRP switchover.

L2TP

L2TP Dialout

IPv4 Local Address Pools

The internal local address server state supports only static recovery. However, the AAA application reallocates active addresses on a switchover. The resulting effect is the IPv4 local address server having full HA support.

IPv6 Local Address Pools

When the IPv6 local pools are configured, you can perform an HA switchover without cold booting the router because the configuration is now HA safe. The prefix assigned to the subscriber, before and after the warm restart, remains the same. The In Use prefix count also remains the same before and after the warm restart.

RADIUS Client

Similar to local address server, AAA recovers disrupted RADIUS communication on a switchover. The resulting effect is the RADIUS client having full HA support.

RADIUS Dynamic-Request Server

Static recovery support only.

RADIUS Initiated Disconnect

 

RADIUS Relay Server

RADIUS Route-Download Server

Service Manager

SRC Client

TACACS+

Static recovery support only.

Miscellaneous

DNS

DNSv6

If DNSv6 is configured, no warning or error is displayed during a warm start. DNSv6 is subsequently configured from NVS as it is after a cold reboot.

J-Flow (IP flow statistics)

Line Module Redundancy

Network Address Translation

NTP

Resource Threshold Monitor

Response Time Reporter

Route Policy

Static recovery support only.

Subscriber Interfaces

IPv4 only. Subscriber interfaces are not applicable to IPv6.

Tunnels (GRE and DVMRP)

VRRP

Static recovery support only.

Caution: When IP tunnels are configured on an HA-enabled router and the Service Module (SM) carrying these tunnels is reloaded, HA transitions to the pending state. HA remains in the pending state for 5 minutes after the successful reloading of the SM. This amount of time allows for IP tunnel relocation and for the tunnels to become operational again on the SM. If an SRP switchover occurs while HA is in the pending state, the router performs a cold restart.

Related Documentation