Cisco MGX 8800 Series Switches

Table Of Contents

Release Notes for Cisco MGX Route Processor Module (RPM-XF) IOS Release 12.3(11)T3 for PXM45-based Switches, Release 5.1.00

Contents

New Features

Features Introduced with 12.3(11)T3

Features Introduced with 12.3(7)T3

Features Introduced with 12.3(2)T6

Features Introduced with 12.3(2)T5

Features Introduced with 12.3(2)T4

Feature Introduced with 12.3(2)T2

Features Introduced Prior to 12.3(2)T2

RPM-XF Redundancy Support

Features Not Supported in This Release

Network Management Features

SNMP MIB

RPM-XF Limitations and Restrictions

Notes and Cautions

RPM-XF auto_config File Management

Card Management

Open Anomalies —Release 12.3(11)T3 for RPM-XF

Open Anomalies —Release 12.3(7)T3 for RPM-XF

Resolved Caveats—Release 12.3(11)T3

Resolved Caveats—Release 12.3(7)T3

Resolved Caveats—Release 12.3(2)T6

Resolved Caveats—Release 12.3(2)T5

Resolved Caveats—Release 12.3(2)T4

Resolved Caveats—Release 12.3(2)T2

Compatibility Notes

RPM-XF Boot File and Firmware File Names and Sizes

RPM-XF Compatibility Matrix

MGX RPM-XF Hardware

Cisco IOS Release Compatibility Information

Using XModem to Download Flash to RPM-XF Cards

Historical Information for 12.2.x IOS Baseline

Resolved Caveats—Release 12.2(15)T5

Resolved Caveats in Release 12.2.15T

Resolved Caveats Prior to Release 12.2.15T

Related Documentation

Obtaining Documentation

Cisco.com

Documentation CD-ROM

Ordering Documentation

Documentation Feedback

Obtaining Technical Assistance

Cisco TAC Website

Opening a TAC Case

TAC Case Priority Definitions

Obtaining Additional Publications and Information

Release Notes for Cisco MGX Route Processor Module (RPM-XF) IOS Release 12.3(11)T3 for PXM45-based Switches, Release 5.1.00

---

These release notes are part OL-7059-01 Rev. B0, January 27, 2005

Contents

New Features

The MGX RPM-XF is a next-generation, high performance model of the RPM for the MGX 8850 and MGX 8950 platforms, using PXM45 processor modules. It is a router module based on an RM7000A MIPS processing engine that fits into slots 1-6 and slots 9-16 in the MGX 8850 and slots 1-6 and slots 11-16 in the MGX 8950.

The RPM-XF hardware provides forwarding technology for packet switching capabilities in excess of 2-million pps. The forwarding engine is packet based and is interfaced to the midplane of the system through a combination of switch interface technologies. For more information on the RPM-XF, refer to the Cisco MX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 5.

Features Introduced with 12.3(11)T3

•Copper SFP

•RPM-XF Software: Queueing elements, statistics, FTP elements, CRTP elements, MIBs

•Dynamic Bandwidth

For more information on these new features added to the RPM-XF with Release 12.3(11)T3, refer to the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 5.

Features Introduced with 12.3(7)T3

The following new features were introduced with release 12.3(7)T3:

•MGX-XF-UI/B notched backcard

The MGX-XF-UI/B notched back card is a re-spin of the user interface back card for the RPM-XF. The notch has been added to allow clearance for installation of the RCON APS connector on the Cisco MGX 8850B and MGX 8880 chassis.

•Preferred Routes on RPM-XF

IOS software release 12.3(7)T3 introduces enhanced support for preferred routes on the RPM-XF. Currently the AXSM and other service modules provide the facility to "associate" an already defined preferred route on the PXM to an SPVC mastered on that service module. The following commands have been updated to support Preferred Route association via CLI or SNMP for SPVC/Hybrid/XPVC connections that have an RPM-XF as the master end:

Syntax Description

Router(config-if-swconn)#[no] prefrte <Route ID>

The Route ID is an identifier for the configured preferred route that is being associated with this connection. Preferred routes are maintained in a separate database on the PXM and referenced by the ID. The range is 0 through 65535. Setting the ID to 0 means no preferred route is configured. The default value for preferred route ID is zero (no preferred route attached).

Router(config-if-swconn)#[no] directrte

Setting the Directed Route flag to Yes means that the user wants the connection to be routed only on the specified preferred route. The default value for directed route is No.

Router(config-if-swconn)# prefrte ?

<1 - 65535> Preferred Route ID value

Router(config-if-swconn)# directrte ?

<cr>

---

Note If you try to use the directrte command to specify a directed route for a connection with its preferred route ID set to zero, an error message displays. Both the prefrte and directrte commands must be run on the master end of the connection. If you try to use these commands on the slave end of the connection, an error message displays.

---

Examples

To configure a preferred route ID value of 10 for the connection.

Router(config-if-swconn)# prefrte 10

To configure a preferred route id value of 5 AND set the connection on directed route-

Router(config-if-swconn)# prefrte 5

Router(config-if-swconn)# directrte

To change the directed route flag for the connection to NO-

Router(config-if-swconn)# no directrte

To set the preferred route id to zero AND set the directed route to NO-

Router(config-if-swconn)# no prefrte

Features Introduced with 12.3(2)T6

The following new features were introduced with release 12.3(2)T6:

•eiGRP between CE to PE

•Basic PPP over ATM feature evaluation on various port speeds from 768Kbps up to DS3 with a maximum of T1 bandwidth per flow.

•PPP over ATM with cRTP on various port speeds from 768Kbps up to DS3 with a maximum of T1 bandwidth per flow.

•PPP over ATM with cRTP and QoS enabled on the links.

•Scaling up to 200 cRTP enabled pppoATM links with QoS

Features Introduced with 12.3(2)T5

No new features were introduced with release 12.3(2)T5.

Features Introduced with 12.3(2)T4

Link Fragmentation Interleaving (LFI)

The 12.3(2)T4 release adds support for Link Fragmentation Interleaving (LFI). For more information on the CLI commands that have been introduced or modified to support this feature on the RPM-XF, refer to the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 4.

IOS configuration information can be found at:

http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121newft/121t/121t5/dtlfifra.htm

Increase the Maximum Number of Policy Maps to 2048

Policy maps along with class maps and service policy maps are used to define a traffic policy and attach them to an interfaces. Prior to this release, you were able to create 256 separate policy maps and a total of 256 class maps per policy map. With this release, the maximum number of policy maps has been increased to 2048. Each policy map supports up to 32 class maps per policy map. For more information on the CLI commands that have been introduced or modified to support this feature on the RPM-XF, refer to the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 4.

MPLS VPN Multicast

The frame-based Multicast VPN (MVPN) feature enables the RPM-XF to pass frame-based multicast traffic to VPNs across the ATM core.

Configuration information can be found at:

http://www.cisco.com/univercd/cc/td/doc/product/software/ios122s/122snwft/release/122s14/fs_mvpn.htm

cRTP

The cRTP acronym refers to the Compressed Real-Time Protocol (CRTP) header. This release of the RPM-XF adds the ability to configure this header.

Below are the CLI commands introduced to support this feature:

–ip rtp header-compression - enables RTP header compression for a particular interface.

–no ip rtp header-compression - disables RTP header compression for a particular interface.

–clear ip rtp header-compression <interface> - resets all statistics for the interface to 0.

–show ip rtp header-compression <interface> [detail] - shows all statistics for an interface.

–show policy-map int sw1.x -shows the number of packets which are compressed because of match in policy map.

Configuration information can be found at http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122newft/122t/122t13/fthdrcmp.htm

WRED Drop Counters

The WRED Drop Counters feature adds Class-Based Packet Counters to existing RPM-XF functionality. The counters can be dscp based or precedence based. For more information on the CLI commands that have been introduced or modified to support the WRED Drop Counters feature on the RPM-XF, refer to the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 4.

TMS

The TMS (Traffic Matrix Statistics) feature allows an administrator to gather the number of packets and bytes that travel across the backbone from internal and external sources. These packets and bytes are called traffic matrix statistics. Use the statistics collected to determine how much traffic the backbone handles. The statistics are always collected on the incoming interface. For more information on the CLI commands that have been introduced or modified to support TMS on the RPM-XF, refer to the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 4.

SAR based TM/QoS

Traffic Management, Weighted RED, and Cell-Based Weighted Fair Queuing algorithm processing is now accomplished using the Segmentation and Reassembly engine (SAR).

TCP de-compression

The 12.3(2)T2 release adds support for TCP de-compression as an adjunct to supporting the Compressed Real-Time Protocol (CRTP) header feature on the RPM-XF.

Dual MPLS Partition for RPM-XF (related to LSC redundancy)

The 12.3(2)T2 release adds support for dual MPLS partitions on the RPM-XF.

Configuration information can be found at
http://www.cisco.com/univercd/cc/td/doc/product/wanbu/8850px1e/rel4/scg/rpm.htm

Feature Introduced with 12.3(2)T2

2-Port POS (Packet Over SONET) and 2-Port GIG-E (Gigabit Ethernet) Service Module Back Cards

The 12.3(2)T2 release adds support for two new service module back cards that provide either two GIG-E (Gigabit Ethernet) or two POS (Packet Over SONET) ports. For more information on the fit and function of these 2 new service module back cards as well as the CLI commands that have been introduced or modified to support these 2 new service module back cards on the RPM-XF, refer to the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide, Release 4.

Features Introduced Prior to 12.3(2)T2

MGX 8950 Support for RPM-XF

In Cisco MGX Release 3.0.10 and later, the MGX 8950 supports the MGX RPM-XF card set. The MGX RPM-XF card set can occupy any of the available service module slots in the MGX 8950, which are slots 1-6 and slots 11-16.

EiBGP Load Balancing

To balance load by BGP on multiple paths to destination, traffic is directed on multiple available paths between autonomous systems (AS) by gateway routers.

The following CLI commands are used to implement this feature.

Command	Description
maximum-path <nums>	Configure maximum number of EiBGP parallel routes. For example: bgpbox-zenith-CE1(config)#router bgp 4 bgpbox-zenith-CE1(config-rout)#maximum-paths 3 bgpbox-zenith-CE1(config-rout)#end
show ip bgp	This command has been enhanced to show the multipaths. Each multipath is marked as 'multipath'. The bestpath is marked as 'multipath' and 'bestpath'. The output also has what flavour of multipath is enabled. For example: bgpbox-zenith-CE1#sh ip bgp 141.22.0.0 BGP routing table entry for 141.22.0.0/16, version 18 Paths: (2 available, best #1) Multipath: eBGP Advertised to non peer-group peers: 7.0.76.9 100 5 7.0.76.2 from 7.0.76.2 (100.0.0.2) Origin IGP, localpref 100, valid, external, multipath, best 100 5 7.0.76.9 from 7.0.76.9 (100.0.0.9) Origin IGP, localpref 100, valid, external, multipath

Limitations:

•If there are multiple alternate paths for the peering point, only one of the paths is utilized for a given prefix.

•Only per-flow load balancing is supported. Per-packet load balancing is not supported.

•Supported only in MPLS/VPN networks.

•Load balancing proportional to the link's bandwidth [dmz-linkbw] is not supported. The load balancing will be performed on the available links with equal costs.

•The maximum number of paths that can be used for load sharing is 6. This is the current IOS limitation too.

•Load balance will not work if RDs are same with RR. RDs have to be different if RRs are used.

•CEs in different VPNs using same RDs does not work.

IP Accounting Counters

MGX RPM-XF stores the packet/byte counters based on precedence/DSCP values on a per interface level at input ONLY.

The following CLI have been added or enhanced for this release to implement this feature:

Command	Description
ip accounting ?	pop20-slot6(config-if)#ip accounting ? precedence Count packets by IP precedence on this interface dscp Count packets by DSCP on this interface
ip accounting precedence ?	pop20-slot6(config-if)#ip accounting precedence ? input received packets and bytes
ip accounting dscp ?	pop20-slot6(config-if)#ip accounting dscp ? input received packets and bytes
show int [interface] precedence	pop20-slot5# show int [interface] precedence
show int [interface] dscp	pop20-slot5# show int [interface] dscp
clear counters	pop20-slot5#clear counters

Limitations:

•Counters are maintained ONLY at input per interface.

•There is no count of dropped/transmitted based onDSCP/PREC packets per interface.

Multiple Actions under Police Command

The MGX RPM-XF police command is now similar to the IOS/RPM command. Therefore, multiple exceed and conform actions can be applied on the police command.

The police CLI command has been enhanced as follows:

Prior to this release, the police command had no menus and all parameters were listed on one line, as shown in the following example.

domino80p01-z001#sh policy test1z

.............

police 128000 8000 8000 conform-action transmit exceed-action drop

This allows only one value for conform-action and exceed-action.

The newly enhanced police command functions as shown in the following example.

ipfrtx90r14-01(config-pmap-c)#police 128000 8000 8000

ipfrtx9(config-pmap-c-police)#conform-action transmit

ipfrtx9(config-pmap-c-police)#exceed-action set-dscp 28

ipfrtx9(config-pmap-c-police)#exceed-action set-mpls 2

Note that multiple conform-action and exceed-action parameters can be configured.

QoS Sub Optimal Link Utilization

RPM-XF uses VTMS as a scheduling algorithm. VTMS schedules queues based on the current link utilization in real time. The previous version of the VTMS algorithm was efficient and mapped well in an ASIC or network processor. However, it did not fully utilize the link.

In this release, a fix is made (via a new CLI command) to allow the user to specify the over subscription factor on a queue. The factor is in the range on 1-31 and can be denoted as 2n. An over subscription factor of n = 2 on any queue, essentially means to subscribe that queue by a factor of 4 ( 2n where n (in this example) is 2; so 2 raised to power 2=4).

The new syntax for bandwidth and priority commands are:

[no] bandwidth {<kbps> | percent <percentage> | remaining percent <percentage>} 
[maximize-utilization [<max-shift>]] 

[no] priority {<kbps> | percent <percentage>} [maximize-utilization 

RPM-XF Redundancy Support

RPM-XF 1:N redundancy is used to switch configuration and traffic from one RPM-XF module to another RPM-XF module. Route processing continues with minimal traffic loss even if an RPM-XF fails and there is no operator or direct access to swap the failed card or fix the problem. Currently we support redundancy that ensures Layer 2 state restoration. Layer 3 state is restored via convergence.

---

Note When resetting a shelf with RPM-XFs configured for 1:N redundancy, it is recommended that you bring up the primary slots in active state.

---

The main benefits of 1:N Redundancy are:

•An RPM-XF card with hardware problems can be fixed while the redundant standby card takes over its functionality.

•Software upgrades are easier and can be done with less downtime.

•LAN interface redundancy supported with MAC addresses of primary RPM-XF copied to standby RPM-XF.

•1:N Redundancy support for Gigabit Ethernet interface backcards during front card switchover.

•Y cable redundancy support for POS backcards during front card switchover. (With Y cable, 1:N redundancy is restricted to N = 1).

The following are the general guidelines for redundancy on the RPM--XF:

•Addred is not allowed between RPM-PR and RPM-XF.

•To configure redundancy, the Primary RPM-XF should be in Active state and Secondary RPM-XF card must be in Active/Standby state.

•Removal of the Active RPM-XF back card does not cause switchover to the standby RPM-XF.

•User has to make sure that E:RPM/auto_config_slot# is created before adding redundancy. This may require a login to primary card through the command line and manually adding boot config e:auto_config_slot# followed by a write mem.

•Executing switchcc back-to-back with switchredcd can cause problems. We recommend giving a gap of at least 5 seconds between switchredcd and a switchcc.

•IOS software on a standby card should be the same or higher version than the Active RPM-XF card.

•Booting the card from an image on tftp server is not recommended when the card is in redundancy group. The card should be booted from image in bootflash or PXM disk only.

•Configuring the standby RPM-XF is not recommended.

Features Not Supported in This Release

The following features are not supported in this release.

•LSC Redundancy

•MPLS TE tunnels on ATM Interfaces

•VC Merge

•RPM-PR to RPM-XF upgrade

•OIR of backcards without interfaces in shutdown mode

•Per packet load balancing

•Modem connectivity on Auxiliary port

•ROMMON's Xmodem functionality does not support the Speed option.

Network Management Features

Network management features are detailed in the Release Notes for Cisco WAN Manager 15.1.00 at:

http://cisco.com/univercd/cc/td/doc/product/wanbu/svplus/151/rnotes/index.htm

SNMP MIB

SNMP MGX Release 5.1.00 MIB files are provided with the delivery of this release. The MIB files are in standard ASN.1 format and are located in the same directory within the release bundle on CCO. These files may be compiled with most standards-based MIB compilers. The tar file for MIB contains the file that contains the MIB release notes. This contains only MGX MIBs.

Cisco IOS MIBs are not part of this bundle. They are part of the 12.3(2)T6 CCO release.

RPM-XF Limitations and Restrictions

The RPM-XF limitations and restrictions that apply to this release are as follows:

•E: RPM/auto_config_slot# must be created before adding redundancy. This may require a login through the CLI and manually adding the boot config command followed by a write mem.

•PVPs can not operate at a rate greater than 599039 kbps.

•High speed VC (SCR greater than or equal to 599,039 kbps) would not get full-configured rate for single flow (unique source and destination IP address). This happens because for high speed VCs, the PXF creates two queues and these queues can't be shared for same stream. Sharing two queues for same stream would cause out of sequence packets.

•PXF queue selection algorithm may cause traffic drop for multiple stream going to same destination via multiple paths. When the PXF gets a packet, it selects the output queue based on source and destination IP address. These addresses hash into one of the queues for the selected destination. So if there are multiple paths for the same destination, there is a possibility that multiple streams would hash to one queue, causing some queues to overflow, while others might be under-utilized.

•PXF buffer depletion may occur if packets of the same size (especially packets greater than 640 bytes) are sent to a congested interface.

•Currently VBR-nrt and VBR-rt are treated with same priority system wide.

•RPM-XF PVP only supports UBR.

•PVP in RPM-XF is not OAM managed.

•If out-of-sync SPVC or SPVP exist on RPM-XF, shrinking of PNNI partition would not be permitted.

•A single RPM-XF can only function as either an Edge LSR or as an LSC, but not as both.

•Because RPM-XF only supports UBR, VBR-rt and VBR-nrt, on the PXM, dsppnportrsrc for RPM-XF port will show 0 available resource for CBR, ABR and signaling service types. Also, cnfpnportcac for CBR and ABR will be rejected.

•If RPM-XF is configured as an eLSR, RPM-XF does not support incoming VC-merge LVCs. There is a problem logged against LSC module that it cannot support both VC-merge/non-VC-merge supporting VSI slaves at the same time. So for now, if RPM-XF eLSR is part of a cell based MPLS network (with RPM-PRs or AXSMs in the same node), disable the VC-merge feature on LSC. (Note that VC-merge is enabled on LSC by default).

•RPM-XF eLSR only supports at most two MPLS sub-interfaces. Attempting to configure over the limit will result in an error message.

•Although RPM-XF VSI slave supports connections statistics Get command, only packets and bytes counts are available. Therefore, show xtag cross-connect traffic int xtagatm connection statistic display on LSC are actually packet counts from RPM-XF eLSR.

•OIR of MGX-1GE and MGX-1OC12POS-IR back cards are supported only with interfaces in shutdown state.

•MGX-1GE back card does not have the capability to provide line loopback.

•Flow Control Option is not configurable with MGX-1GE back card.

•MGX-1GE back card does not support SFP security.

•Line loopback and internal loopback cannot be set at the same time for the MGX-1OC12POS-IR back card with AMCC Mux.

•pos ais-shut command is not supported on MGX-1OC12POS-IR back card.

•Traffic rate per flow is at half the interface speed for POS/GigE interfaces in this new release.

•Autonegotiation is not recommended to be toggled on GigE interfaces when traffic is going through. This may result in permanent disruption of traffic.

•For UBR 2 queues always exist which would result in half the Flow rate for each flow because of hashing algorithm hashing into single queue .

The performance limits supported in this release are the following:

•2K ATM SPVC Connection endpoints

•2K IDBs

•4K LVCs

•100 VPCs

•256 Policymap

•100 OSPF neighbors

•6 IOS-based cards in MGX shelf

•500 VRFs: 500

•500 BGP CE Peers

•100 RIP CE sessions

•500 Static CEs

•100,000 VPN Routes per PE

•250K non-VPN Routes per RPM-XF

•50 Xtag interfaces per RPM-XF

•300 OAM enabled connections

For more RPM-XF performance details, contact your sales representative.

Notes and Cautions

The following notes and cautions should be reviewed before using this release.

•Attempting to initiate RPM-XF switchover when write mem is in progress on the active RPM-XF card may lead to the card coming up with a partial configuration. When an addred is executed, an automatic write mem is triggered on the primary RPM-XF. If the primary card fails when the write mem is in progress, the card may come up with a partial configuration. The duration of write mem depends on the configuration size and can take up to 4 minutes to complete.

•There is a new stable "Boot-Hold" state displayed on the PXM45 when dspcds is executed. This state indicates that the RPM-XF is running only boot image. This state is reached when config register is set to 0x1 or when the bootldr cannot find the run-time image, but found the boot image. Enter cc to access the RPM-XF from the PXM45.

•Valid boot image need not be the first file in the boot flash. The RPM-XF will load from any valid boot image from the bootflash:. The run-time image can be the first file in the boot flash and RPM-XF will come up with that image.

•Trying to change PCR value of VP tunnel or changing MTU of switch interface with more than 4K VCs may cause CPU hog.

•If there is a large number of VCs (PVCs or LVCs or both) on RPM-XF card, executing disruptive operations on the main switch interface (int switch1) may cause flapping of protocols that run on these VCs. Examples of disruptive operations are clear int switch1 and modification of PVP parameters. These operations cause deactivation and re-activation of all VCs under the main switch interface. Depending on the number of VCs, the time required to complete such operations may exceed certain protocol timeout limit. Examples of protocols that may be affected are OSPF and TDP/LDP.

•RPM-XF VSI slave tends to output informational warning/trace back messages caused by misconfigurations and CAC failures (onto console/IOS log file). These messages are mostly for information/debugging purpose. When these messages are observed, confirm that connection status is still intact and traffic is still passing successfully.

•Due to PXF scr granularity, the configured scr on IOS pvc CLI may not be the same as the actual scr programmed in the PXF. PXF bandwidth chunk size is 18 kbps; all PXF VC scr will be programmed as multiples of 18 kbps. For instance, if the PVCs were configured with 50 kbps as pcr, 54 kbps would be programmed in PXF. show atm pvc display will show 50 kbps, and VSI Slave will account 50 kbps during CAC. However, 54 kbps is actually being used. So as a result, when bandwidth usage is reaching the maximum value, both VSI Slave and PNNI will continue to allow connection provisioning, because VSI Slave and PNNI available bandwidth shows more than PXF actually has left.

•Saveallcnf (issued on the PXM45/B card) captures configuration data saved by the RPM-XF card (as well as AXSM and PXM45 cards), and saves it on the active PXM45/B card's hard disk. Configure the RPM-XF to store its configuration on the PXM45/B hard disk (E:/RPM) by entering boot config e:auto_config_slot# in the running configuration of the RPM-XF. To ensure that the saved file contains the latest RPM-XF configuration, execute the write mem command on each RPM-XF card prior to the entering saveallcnf command. This also ensures that the RPM-XF files on the active PXM45 hard disk will contain the latest configuration to be saved.

•For ELSR to LSC connectivity, the default control VC used is 32. If PNNI partition exists with VCI 32 as part of its partition range, when an MPLS partition is added, there are two options to handle the situation:

–Add the MPLS controller and define its partition with available range. On eLSR, define control VC from any VCI value within the range defined in partition. The same VC should be defined on LSC on xTag interface.

–Reconfigure PNNI partition to spare the control VC usage both on RPM-XF and AXSM, AXSM/B or AXSM-E APS Management Information.

•Whenever the RPM-XF configuration is changed, enter the write mem command on the RPM-XF to save the configuration. If this is not done, the changed configuration will be lost on an RPM-XF card reboot or RPM-XF switchover, in the case of redundancy.

RPM-XF auto_config File Management

The RPM-XF auto_config_slot# file stores the configuration for the RPM-XF card. The slot# portion of the name should be set to the logical slot number that corresponds to the RPM-XF card. This file can be stored in bootflash or in the E:RPM directory on the PXM45 hard disk. The configuration is also stored in NVRAM using the name startup-config.

When the RPM-XF card is inserted or rebooted, it searches for the configuration file in the following sequence:

1. If there is an auto_config file corresponding to its logical slot on the PXM45 hard disk, the RPM-XF card uses the configuration stored on the hard disk.

2. If boot variable points to configuration stored in the PXM45 hard disk or Bootflash and if the file is not found, the card comes up as Active-F with the default configuration.

3. If there is no auto_config file on the hard disk, then the NVRAM version is used.

---

Note In case of RPM-XF redundancy, the configuration should always be stored in auto_config_slot# file in the E:RPM directory of the PXM45 hard disk. Failure to find the auto_config file will lead to aborting of a user-initiated switchover (switchredcd) and a fatal error will be flagged.

---

Card Management

The following card management notes and cautions should be reviewed before using this release.

•There is a new stable state displayed on the PXM dspcds command—Boot-Hold, which signifies that the RPM-XF is running the boot image only. On the RPM-XF, the prompt will display as boot>

•The run-time IOS image cannot be used as a bootloader to load a different IOS image.

•Change of console speed on the terminal server may cause the card to end up in the ROMMON state. To avoid this, set the config register to 0x2102.

Another workaround is to enter cont on the ROMMON within 2 minutes of going into ROMMON state. This will bring the card to its original stable state.

---

Note It is recommended to always use 9600 baud as the console speed.

---

•The IOS version of the runtime as well as the boot image will be displayed in the dspcd, dsprevs, and dsprevs -s output. The version will be displayed under the heading of IOS version. Revision Control is not available for RPM-XF (like RPM-PR).

---

Note The commands loadrev and setrev do not apply for RPM-XF.

---

RPM-XF Bootflash Precautions

The RPM-XF bootflash is used to store boot image, configuration and run- time files. Erasing the boot image from the Flash will cause the card to not boot.

The RPM-XF boot image, which comes loaded on the Flash, will work for all RPM-XF IOS images. Therefore, there is no reason to delete or move the factory installed boot image.

In order to avoid any unnecessary failures that would require card servicing, do the following:

•Never erase the boot file from the RPM Flash

•Never change the position of the boot file on the RPM Flash

•Use care when "squeezing" the Flash to clean it up.

As long as the boot file remains intact in the first position on the flash, the RPM-XF will boot successfully.

If the bootflash is corrupted, use the tftpdnld procedure described in the Cisco MGX Route Processor Module (RPM-XF) Installation and Configuration Guide or xmodem procedure described in "Using XModem to Download Flash to RPM-XF Cards" later is this document to download a new boot image.

Solving the RPM-XF Bandwidth Issue When Adding a 12th VISM Card

Adding more than 11 VISM cards to an MGX chassis with RPM-XF cards requires enabling the expanded memory option on PXM45/B. The command to enable this option is cnfndparms (option 4). This expanded memory option does not impact chassis performance while allowing more connections.

Open Anomalies —Release 12.3(11)T3 for RPM-XF

Table 2 lists the open caveats in Cisco IOS Release 12.3(11)T3 as of January 25, 2005.

Table 1 Open Anomalies on 12.3(11)T3 for RPM-XF 

Caveat Number	Description
CSCee75243	Description: RPM-XF may reload abnormally during a rapid adding and removal of the service policy map Condition: While adding/removing service policy map RPM-XF router reloaded abnormally with following error / traceback: %GENERAL-3-EREVENT: Policy map is in use -Traceback= 400BAD74 400BB498 400BB6A0 Workaround: Unknown.
CSCef05018	Symptoms: Disconnecting and reconnecting a Gigabit Ethernet cable on an RPM-XF may cause the TCP/IP connection to be lost on the VISM. Conditions: This symptom is observed on an RPM-XF that runs Cisco IOS Release 12.2(11)YP but could also occur in Release 12.3. Workaround: There is no workaround.
CSCef93482	Symptoms: RPM-XF goes into Failed state. Conditions: In Redstone solution, after MPSM switchover, turn on PPP debug negative on RPM-XF Workaround: None.
CSCeg23771	Symptom: All compressed UDP packets from RPM_PR CE are dropped by PE RPM_XF. Conditions: Configure basic frame based MPLS and send UDP traffic from CE to PE. Workaround: Unknown.
CSCeg24938	Symptoms: Continuously execute "clear ip mroute *" causes RPM-XF to get reset Conditions: While RPM-XF is configured as XFL, "clear ip mroute *" command was executed repeatedly which eventually caused the router to get reloaded abnormally. Workaround: Unknown.
CSCeg27043	Symptoms: PIM neighbors continue to flap after you have reloaded microcode. Conditions: This symptom is observed on a Cisco MGX 8850 series RPM-XF that runs Cisco IOS Release 12.3. Workaround: There is no workaround.
CSCeg40721	Symptom: SAR tail drops seen on multicast PEs Conditions: This problem was seen on a MGX8850 node having RPM-XF cards running 12.3(11)T images. we see tail drops on the SAR while sending bursty traffic to multiple multicast destinations. Workaround: Under Investigation.
CSCeg64074	Symptom: Switch Connection go through the following states with the maximum PCR value. inSync, unknown, OnlyOnRPm states. Conditions: Create a Switch connection between any two XF cards with service type vbr-nrt with maximum PCR value. Workaround: None.
CSCsa45189	Symptom: Header compression is not working on an RPM XF card that is configured for SAR based CBWFQ. Conditions: This problem is seen when header compression is configured with PPP configuration on a virtual template. This configuration is used since the PVC size is more than 768kbps and MLPPP does not support COS for PVC's of this size. Workaround: Use smaller PVC with MLPPP and cRTP.

Open Anomalies —Release 12.3(7)T3 for RPM-XF

Table 2 lists the open caveats in Cisco IOS Release 12.3(7)T3 as of August 18, 2004.

Table 2 Open Anomalies on 12.3(7)T3 for RPM-XF 

Caveat Number	Description
CSCec16481	A Cisco device running Internetwork Operating System (IOS) and enabled for the Open Shortest Path First (OSPF) Protocol is vulnerable to a Denial of Service (DoS) attack from a malformed OSPF packet. The OSPF protocol is not enabled by default. The vulnerability is only present in IOS release trains based on 12.0S, 12.2, and 12.3. Releases based on 12.0, 12.1 mainlines and all IOS images prior to 12.0 are not affected. Refer to the Security Advisory for a complete list of affected release trains. Further details and the workarounds to mitigate the effects are explained in the Security Advisory which is available at the following URL: http://www.cisco.com/warp/public/707/cisco-sa-20040818-ospf.shtml.
CSCed05924	Symptom: PXF reloaded abnormally due to software exception %PXF-2-FAULT:T0 SW Exception:CPU[t0r3c1] 0x00000180 at 0x0DE1 LR 0x084B Conditions: MVPN traffic was being passed. No other activity on the card at that time. Workaround: Unknown.
CSCed34575	Symptoms: MPLS packet will be transmitted without fragmentation even if the MPLS packet exceeds the outgoing interface MTU. Conditions: When the MPLS packet is greater than the outgoing interface MTU. Workaround: None.
CSCed39641	Symptoms: SAR rx drops all packets due to no buffer Conditions: RPM-XF configured as PE in frame-based or cell-based MPLS/VPN network. Execute "sh pxf cpu buff leaked 0-5" or "clear interface sw1" command while system is under load, may cause SAR rx failure. Workaround: None.
CSCed48954	Symptoms: Traffic will tail drop on the output of GigabitEthernet interface even when the traffic rate is well below the interface limit. Condition: This problem seems to happen after multiple GigabitEthernet interface flaps. Workaround: Reload the PXF using "microcode reload pxf" command.
CSCed86771	Symptoms: Pull/insert RPM-XF while running call rate, made the card to reload/failed. Conditions: While running 360K BHCA (Busy Hour Call Attempts) with 120 sec CHT. Upon pulling one of RPM-XF card and putting it back into the rack, this RPM-XF card rebooted and went to failed state. Workaround: None.
CSCee36771	Symptoms: A PPPoA interface constantly flaps when passing data with SAR-based CBWFQ enabled. Conditions: This has been observed under the following conditions: - SAR-based CBWFQ is enabled on a PPPoA interface. - the class default is assigned a very small bandwidth (less than 10 percent) - all classes on the VC are congested. Workaround: Assign 10 percent bandwidth to the class-default of the policy map attached to the interface.
CSCee53246	Symptoms: The standby (secondary) Route processor Module does not release the config_file boot variable after the primary redundant RPM-PR card takes over after the card switch over command was executed. Conditions: This intermittent problem was observed after a switchover from secondary RPM-PR card to primary RPM-PR card. Secondary (redundant) card is in Standby state, but the "show bootvar" command still shows that config_file variable is not null. Workaround: Unknown.
CSCee74057	RPM throughput performance degradation in pi5 branch
CSCea84387	Two simultaneous policy map displays cause problems Symptoms: A user session may pause indefinitely, causing a Cisco router to become unresponsive. Conditions: This symptom is observed when multiple simultaneous users enter modular QoS CLI (MQC) commands on the same router via separate vty sessions. Workaround: Allow only one user at a time to enter MQC commands.

Resolved Caveats—Release 12.3(11)T3

Table 4 lists the resolved caveats in Cisco IOS Release 12.3(11)T3 for MGX 5.1.00 as of January 25, 2005.

Table 3 Resolved Caveats in Cisco IOS Release 12.3(11)T3 for MGX 5.1.00 

Caveat Number	Description
CSCee58709	Symptoms: PCI retry counter expired during the boot up of RPM (route processor module) Conditions: This issue was reported against 12.3(0.2)T internal Cisco Release which fails to boot up about 10% of the time with a PCI retry counter expired warning error. Workaround: None
CSCef95597	Symptoms: RPM-XF router reloads abnormally when the ATM encapsulation for a PVC is changed. Conditions: This symptom is observed when an MLP session is configured over a PVC subinterface. Workaround: Shut down the PVC subinterface, change the encapsulation, and bring up the PVC subinterface.
CSCsa40567	Symptoms: The output of the "show rpm iphc cids [src-ip dest-ip src-udp-port dest-udp-port max-cids]" command does not show the CID values. Only zeros are seen in the command output. Conditions: This symptom is observed on a Cisco RPM-XF that runs Cisco IOS Release 12.3(11)T1 when cRTP is configured. Workaround: There is no workaround.
CSCee65241	Symptoms: In a "show policy-map interface" command, all the "police" counters should be read as Layer 2 counters. Conditions: This is observed on a Cisco RPM-XF running IOS release 12.3(11)T2 and lower Workaround: There is no workaround and is just a display issue
CSCef91218	Symptoms: An RPM-XF corrupts the DSCP values of traffic passing through. Conditions: This symptom is observed after an output service policy is applied and fails because of incorrect parameters. The following error message is generated when the policy map is applied and fails: PE1(config-if-atm-vc)#service-policy output in_policy_forout PE1(config-if-atm-vc)#end "set-mpls-exp-imposition-transmit" is not allowed in an output service policy. Service policy installation failed on VC Workaround: Verify the validity of the service policy before you apply the output policy map.
CSCeg16953	Symptoms: DBF tracebacks on XF by RPMXF VSIS process. Conditions: Create a basic setup with RPM-XF and MPSM. Attach a Policy-map with Police to a interface. Change the traffic parameter in the MPSM side so that we hit a low bandwidth of 4Kbps in rpmxf. Workaround: Unknown.
CSCeg65362	Symptom: The "show policy interface <intf> input" command shows more packets are received than the show interface <intf> precedence command. Conditions: The test is carried out between two routers running MLPPP and cRTP. Workaround: None.
CSCsa45197	Symptoms: When you enter the "show policy-map interface [interface-name] [output]" command for a switch subinterface, the drop rate counter always shows zero. Conditions: This symptom is observed on a Cisco RPM-XF that is configured for SAR-based CBWFQ. Workaround: Determine the drop rate from the number of exceeded packets in the output of the "show policy-map interface [interface-name] output" command.
CSCee63435	Symptoms: RPMXF VSI Slave does not reply with extended VSI Nak error codes 51-54. Conditions: Controller requests for extended VSI Nak51-54 codes from RPMXF VSI Slave in case of dangling connection and RPMXF VSI Slave still replies with old error codes. Workaround: None.
CSCef36941	Symptoms Apart from Real time traffic streams, the RPM-XF card compresses the UDP traffic also on the Multilink interface configured for IP Header compression and connected to Customer router. The compressed UDP traffic received by the CE creates some problems thus making the CE unstable. Conditions When there is UDP traffic destined to the CE apart from the Real time traffic. Workaround None.
CSCef92881	Symptoms: Output of traceroute incorrectly shows the next hop entry for an interface on RPM XF as 0.0.0.0 instead of showing the IP address of the interface. E.g. mwr1900-1-01#traceroute 199.40.49.2 Type escape sequence to abort. Tracing the route to 199.40.49.2 1 0.0.0.0 4 msec 4 msec 4 msec <<<<<Should be 199.40.45.2 2 199.40.49.2 4 msec * 4 msec Condition: This happens only for those interfaces on rpmxf where the IP address is configured using a virtual template. Workaround: None.
CSCeg10138	Symptoms: The 64-bit counters in the output of a "show policy-map" command may not provide correct information. Condition: This symptom is observed on a Cisco RPM-XF. Workaround: There is no workaround.
CSCeg16660	Symptom: Load balancing of traffic works inconsistently if the traffic flow reaching the particular PE was already load balanced at a previous hop by another PE. Conditions: The problem is seen only when Load balance needs to work in series, akin to the scenarios where you would see polarization effect of load balancing in play. Workaround: None.
CSCeg17058	Symptoms: There is not enough information to verify the exact cause of Memory related ECC errors for a PXF ASIC present on a Cisco RPM-XF Conditions: This is required when the PXF ASIC crashes due to external memory issues like ECC errors Workaround: None.
CSCeg17274	Symptoms: When you enter a timeout value shorter than 8 seconds on an compression-enabled interface, the value is not configured. Instead, a timeout value of 8 seconds is configured. Conditions: This symptom is observed on a Cisco RPM-XF that is configured for IP RAN when you use the "ppp iphc max-time [seconds]" command to enter the timeout value on a virtual template in a multilink configuration. Workaround: There is no workaround.
CSCeg18940	Symptom: Connection level parameter mismatch between RPM-XF and CWM Db Conditions: Create a FF-RPM connection using default parameters from CWM Workaround: None.
CSCeg20768	Symptoms: The OSPF cost calculation is not triggered when the DBF update is received. Condition: Changed the normalized bandwidth of the MPBUNDLE between XF and MPSM. Workaround: Perform shut/no Shut on the switch sub interface.
CSCeg24025	Symptoms: IPCP between MWR and RPM-XF does not come up after throttling Q2 on MPSM. Conditions: With IPCP/LCP/OSPF Up between MWR and RPM-XF, pump traffic. Throttle Q2 threshold to 0 on MPSM. Flap corresponding switch subinterface on XF. Workaround: Increase the Q2 threshold of the MPSM bundle to 10 or more.
CSCeg25053	Symptom: There is no notification message in the log buffer or on the console related to the Switch Connection Synchronization applicable to Auto-Resync or Manual Resync Conditions: This is observed on a Cisco RPM-XF running IOS version 12.3(11)T2 or lower Workaround: There is no workaround.
CSCeg27046	Symptom: Important system and CPU register values are not dumped/stored into Crashinfo file if router reload abnormally. Conditions: PCI SERR or Galileo interrupt is triggered and pciinfo/ crashinfo file is written to the bootflash. Workaround: None.
CSCeg28876	Symptoms: "no ip route-cache" is displayed under MLPPP interface in the configuration even though route-cache is enabled by default. In addition "no ip route-cache cef" is displayed sometimes Conditions: (a) For MLP interfaces on RPM-XF, "no ip route-cache" appears after the MLP interface is UP (b) "no ip route-cache cef" appears if IP Header compression is configured for the MLP interface (c) RPM-XF image versions 12.3(11)T2 and below Workaround: There is no workaround for this display issue but it is to be noted that the Route Caching is in effect.
CSCeg31236	Symptom: Spurious memory access is seen along with the traceback. Conditions: This traceback is seen with default configuration and when "sh pxf cpu tfib" command is executed such that there is no label to be verified. Workaround: None.
CSCeg34852	Symptom: VCCI drops on RPM-XF after resetting the MPSM card Conditions: When compressed udp traffic is received on rpmxf (more than 10000 pkts per sec) Workaround: Issuing "clear int virtual-access <x>" is seen to be clearing off the problem.
CSCeg36182	Symptom: RPM-XF card does not come up after reload and some tracebacks are observed. Condition: This condition is observed with Router is reloaded with Cisco internal image version 12.3(11.6)PI6. This issue may also be applicable to 12.3(11)Tx releases. Workaround: None.
CSCeg47178	Symptoms: Clear counters command takes long time to zero out the average offered/drop rate counters Conditions: While pumping IP traffic for a limited duration with 200bytes packet size at the rate of 1000 bps, clear counters took far took long (1:45 min) to zero out the average offered/drop rate counters Workaround: None.
CSCeg65037	Symptoms: swfpga cam overwrites which results in dangling connection. Conditions: When a new cam entry add request clashes with an existing entry, this results in cam entry overwrite which could result in dangling connection. Workaround: None.
CSCin81995	Symptoms: The Segmentation And Reassembly (SAR) engine on a Cisco RPM-XF shows buffer exhaustion, causing data drops. Conditions: This symptom is observed when ATM SAR-based class-based weighted fair queueing (CBWFQ) is enabled and when policy maps are attached to outgoing PPPoA interfaces. Workaround: There is no workaround.
CSCin84419	Symptoms: Multicast traffic is punted to the RP, the CPU utilization is high, and the output of the "show pxf cpu mroute vrf [vrf-name]" command shows that the "No_FS" flag is set for a (S,G) entry and does not clear. Conditions: This symptom is observed on a Cisco RPM-XF when you enter the "clear ip mroute [group] command and a data MDT group for the [group] argument. Workaround: Enter the "clear ip mroute vrf *" command.
CSCin84421	Symptom: Traffic outage on switching from data MDT to a default MDT. Conditions: This condition was observed when "no mdt data" command was executed. Workaround: Use "mdt data" with a high threshold value to switch to default instead of a "no mdt data".
CSCin84494	Symptom: CPU util is 99%. Conditions: Multicast traffic is being punted to the RP and process switched. Workaround: Multicast traffic is punted to RP to create state. For a finite but transient period the cpu util can spike up if the multicast traffic volume being punted to the RP is very high. The cpu util comes down to normal once the multicast state has been established.

Resolved Caveats—Release 12.3(7)T3

Table 4 lists the resolved caveats in Cisco IOS Release 12.3(7)T3 for MGX 5.0.10 as of August 18, 2004.

Table 4 Resolved Caveats in Cisco IOS Release 12.3(7)T3 for MGX 5.0.10 

Caveat Number	Description
CSCee54524	During SCR provisioning (changing the bandwidth) on the ePVCs, the RPM-PR (Route Processor Module) PE may reload.
CSCed41381	LSNT: Input drops on framed PVC i/f cause ldp session flap
CSCed74835	When copying big files from Cisco MGX8850 Route Processor Module (RPM-PR) to MGX8850 hard drive, it silently failed.
CSCea85395	BGP suppressed prefixes not reinstated after condition removed
CSCea62571	%UTIL-3-TREE: Data structure error with trace back logged on LSC
CSCed41823	LSNT: Tx SAR stuck after micro rel sar tx/rx issued
CSCed74882	OSPF: CPUHOG Traceback on reload with large no of sec. ip addresses
CSCed88043	The outgoing VCCI programmed in the FIB/TFIB in the PXF for a prefix is wrong.
CSCed92418	back to back clear int sw1 causes VSI down on PXM
CSCee06261	LSNT: RP crash on rpmxf_is_atm_mlp_configured while clear int sw1
CSCee12415	Multicast traffic is not getting switched correctly.
CSCed16744	Traffic does not resume after SAR is brought out of hang state.
CSCee23320	Router might reload upon deletion/re-apply of policy-map.

Resolved Caveats—Release 12.3(2)T6

Table 5 lists the resolved caveats in Cisco IOS Release 12.3(2)T6 as of June 4, 2004.

Table 5 Resolved Caveats in Cisco IOS Release 12.3(2)T6 

Caveat Number	Description
CSCdy81782	no shut on the ppp interface before VA goes down causes pxf to drop
CSCed46603	MIB walk on ifOutDiscards object OID returns an error message.
CSCed53155	After failure recovery, the SAR Segmenter is not programmed correctly.
CSCed62886	TagI counter always shows 0 in the output of "show pxf cpu cef mem" command.
CSCed68881	"sh controller" output is not part of sar info files.
CSCed71495	The exp bit on the topmost label is not changed when set mpls exp topmost is configured on the ingress interface of the P router.
CSCed75086	When issuing the command show pxf cpu rewrite verification x.x.x.x you receive an error message stating that the "Channel ID in the SAR header is non-zero (x) for MVC"
CSCed82673	An RPM-XF card may reload abnormally when issuing some of the display commands under "show rpm vsis core conn debug-command <WORD> <Argument#1>"
CSCed41381	Input cell drops may occur on an ingress frame PVC that is configured on a switch interface. This situation may cause LDP/TDP/OSPF flaps.
CSCed71750	Virtual-Access counters doesn't match the ATM subinterface counters.
CSCed91750	S,G entries are not being created in the core.
CSCee00031	The `average packet size' displayed under `show ip mroute count' doesn't exactly match the actual size of the multicast packet being sent.
CSCee00685	Wrong DSCP values set on IP packets.
CSCee02220	Multicast traffic flows use default MDT instead of data MDT for some VRFs.
CSCeb05118	A Cisco MGX Route Processor Module (RPM-XF) router that is configured as an Edge Label Switch Router (ELSR) may reload when deleting Multiprotocol Label Switching (MPLS) type subinterfaces.
CSCee40165	The "show policy interface multilink <int> output" command shows incorrect counts for the dscp value tabulation at the end of the command output when rtp header compression is enabled under the multilink interface.
CSCec21461	On the RPM-XF, the input packet count for Virtual-Access interfaces are higher than the actual number of packets received.
CSCed22425	On the Cisco RPM-XF router, there is no way to know which external Border Gateway Protocol (eBGP) path is chosen when there are multiple VRF interfaces to the VPN prefix.
CSCed41273	PXF gets reloaded abnormally several times after microcode reload.
CSCed42706	On the Cisco RPM-XF router, the PXF does not increment the correct drop code when dropping packets.
CSCed68717	Incoming traffic is not being forwarded.
CSCed70687	PXF buffer allocate failure happens on ELSR.
CSCed78131	Checksum errors are reported on cRTP traffic streams.
CSCed83738	Packets on cRTP-enabled PPPoA interfaces that match classes other than "class-default" will be dropped.
CSCed89382	On Multilink Protocol interfaces using Link Fragmentation and Interleaving, the Fragmented Pkts counter under the "show pxf cpu subblocks Multilink1" command, increments when it should not.
CSCed90333	Traffic is not forwarded through a newly added CBWFQ class.
CSCed94549	A compressed packet from XF is rejected by the RPM-PR as a CRC error.
CSCed96053	Does not show precedence ip accounting for rtp/udp compression packets.
CSCee00038	Protocols flap when the non-ATM (POS or GigE) interfaces are congested by high traffic.
CSCee02404	PXF buffer leak, loss of connectivity, BGP dn on the PE-CE VRF link with cRTP enabled.
CSCee03726	PXF buffers are leaked.
CSCee07654	Starting on Multicast traffic on the CE puts the PXF on the PE in a loop sometimes. LDP/BGP/OSPF all go down and there is no data continuity.
CSCee11775	When Parallel Express Forwarding (PXF) forwarding fails while a debugging operation is performed you may not be able to verify the string rewrite information of the PXF engine easily.
CSCee12335	PXF buffer leak is observed when the multilink interface is flapped. Traffic must be running across the card.
CSCee14274	With Data path check feature enabled, for some reason if the data path pings fail and even though traffic is flowing through switch1 "data path feature recovery" kicks in. Data path feature recovery is to reset the card.