| Network Work group | N. Kumar |
| Internet-Draft | G. Swallow |
| Intended status: Standards Track | C. Pignataro |
| Expires: January 31, 2016 | Cisco Systems, Inc. |
| N. Akiya | |
| Big Switch Networks | |
| S. Kini | |
| Ericsson | |
| H. Gredler | |
| Juniper Networks | |
| M. Chen | |
| Huawei | |
| July 30, 2015 |
Label Switched Path (LSP) Ping/Trace for Segment Routing Networks Using MPLS Dataplane
draft-kumarkini-mpls-spring-lsp-ping-04
Segment Routing architecture leverages the source routing and tunneling paradigms and can be directly applied to MPLS data plane. A node steers a packet through a controlled set of instructions called segments, by prepending the packet with a Segment Routing header.
The segment assignment and forwarding semantic nature of Segment Routing raises additional consideration for connectivity verification and fault isolation in LSP with Segment Routing architecture. This document illustrates the problem and describe a mechanism to perform LSP Ping and Traceroute on Segment Routing network over MPLS data plane.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on January 31, 2016.
Copyright (c) 2015 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
[I-D.ietf-spring-segment-routing] introduces and explains Segment Routing architecture that leverages the source routing and tunneling paradigms. A node steers a packet through a controlled set of instructions called segments, by prepending the packet with Segment Routing header. A detailed definition about Segment Routing architecture is available in [I-D.ietf-spring-segment-routing] and different use-cases are discussed in [I-D.filsfils-spring-segment-routing-use-cases]
The Segment Routing architecture can be directly applied to MPLS data plane in a way that, the Segment identifier (Segment ID) will be of 20-bits size and Segment Routing header is the label stack.
Multi Protocol Label Switching (MPLS) has defined in [RFC4379] a simple and efficient mechanism to detect data plane failures in Label Switched Paths (LSP) by specifying information to be carried in an MPLS "echo request" and "echo reply" for the purposes of fault detection and isolation, and mechanisms for reliably sending the echo reply. The functionality is modeled after the ping/traceroute paradigm (ICMP echo request [RFC0792]) and is typically referred to as LSP ping and LSP traceroute.
Unlike LDP or RSVP which are the other well-known MPLS control plane protocols, segment assignment in Segment Routing architecture is not hop-by-hop basis.
This nature of Segment Routing raises additional consideration for fault detection and isolation in Segment Routing network. This document illustrates the problem and describe a mechanism to perform LSP Ping and Traceroute on Segment Routing network over MPLS data plane.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
This document uses the terminologies defined in [I-D.ietf-spring-segment-routing], [RFC4379], and so the readers are expected to be familiar with the same.
This document defines sub-TLVs for the Target FEC Stack TLV and explains how they can be used to tackle below challenges.
[RFC4379] defines the OAM machinery that helps with fault detection and isolation in MPLS dataplane path with the use of various Target FEC Stack Sub-TLV that are carried in MPLS Echo Request packets and used by the responder for FEC validation. While it is obvious that new Sub-TLVs need to be assigned, the unique nature of Segment Routing architecture raises a need for additional machinery for path validation. This section discuss the challenges as below:
L1
+--------+
| L2 |
R3-------R6
/ \
/ \
R1----R2 R7----R8
\ /
\ /
R4-------R5
Figure 1: Segment Routing network
The Node segment IDs for R1, R2, R3, R4, R5, R6, R7 and R8 are 5001, 5002, 5003, 5004, 5005, 5006, 5007, 5008 respectively.
9136 --> Adjacency Segment ID from R3 to R6 over link L1.
9236 --> Adjacency Segment ID from R3 to R6 over link L2.
9124 --> Adjacency segment ID from R2 to R4.
9123 --> Adjacency Segment ID from R2 to R3.
The forwarding semantic of Adjacency Segment ID is to pop the segment ID and send the packet to a specific neighbor over a specific link. A malfunctioning node may forward packets using Adjacency Segment ID to incorrect neighbor or over incorrect link. Exposed segment ID (after incorrectly forwarded Adjacency Segment ID) might still allow such packet to reach the intended destination, although the intended strict traversal has been broken.
Assume in above topology, R1 sends traffic with segment stack as {9124, 5008} so that the path taken will be R1-R2-R4-R5-R7-R8. If the Adjacency Segment ID 9124 is misprogrammed in R2 to send the packet to R1 or R3, it will still be delivered to R8 but is not via the expected path.
MPLS traceroute may help with detecting such deviation in above mentioned scenario. However, in a different example, it may not be helpful. For example if R3, due to misprogramming, forwards packet with Adjacency Segment ID 9236 via link L1 while it is expected to be forwarded over Link L2.
A Segment ID can represent a service based instruction. An Segment Routing header can have label stack entries where the label represents a service to be applied along the path. Since these labels are part of the label stack, they can influence the path taken by a packet and consequently have implications on MPLS OAM. In section 6.5 of this document, it is described how the procedures of [RFC4379] can be applied to in the absence of service-labels in Section 6.5. Additional considerations for service labels are included in Section 7 and requires further discussion.
The format of the following Segment ID sub-TLVs follows the philosophy of Target FEC Stack TLV carrying FECs corresponding to each label in the label stack. When operated with the procedures defined in [RFC4379], this allows LSP ping/traceroute operations to function when Target FEC Stack TLV contains more FECs than received label stack at responder nodes.
Three new sub-TLVs are defined for TLVs type 1, 16 and 21.
sub-Type Value Field -------- --------------- TBD1 IPv4 Prefix Node Segment ID TBD2 IPv6 Prefix Node Segment ID TBD3 Adjacency Segment ID
Service Segments and FRR will be considered in future version.
The format is as below:
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | IPv4 Prefix | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |Prefix Length | Protocol | Reserved | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
IPv4 Prefix
Prefix Length
Protocol
The format is as below:
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | IPv6 Prefix | | | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |Prefix Length | Protocol | Reserved | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
IPv6 Prefix
Prefix Length
Protocol
The format is as below:
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Adj. Type | Protocol | Reserved | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Local Interface ID (4 or 16 octets) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Remote Interface ID (4 or 16 octets) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ ~ | Advertising Node Identifier (4 or 6 octets) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ ~ | Receiving Node Identifier (4 or 6 octets) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Adj. Type
Protocol
Local Interface ID
Remote Interface ID
Advertising Node Identifier
Receiving Node Identifier
In an echo reply, the Downstream Mapping TLV [RFC4379] is used to report for each interface over which a FEC could be forwarded. For a FEC, there are multiple protocols that may be used to distribute label mapping. The "Protocol" field of the Downstream Mapping TLV is used to return the protocol that is used to distribute a specific a label. The following protocols are defined in section 3.2 of [RFC4379]:
Protocol # Signaling Protocol
---------- ------------------
0 Unknown
1 Static
2 BGP
3 LDP
4 RSVP-TE
With segment routing, OSPF or ISIS can be used for label distribution, this document adds two new protocols as follows:
Protocol # Signaling Protocol
---------- ------------------
5 OSPF
6 ISIS
This section describes aspects of LSP Ping and traceroute operations that require further considerations beyond [RFC4379].
When LSP echo request packets are generated by an initiator, FECs carried in Target FEC Stack TLV may need to have deviating contents. This document outlines expected Target FEC Stack TLV construction mechanics by initiator for known scenarios.
Section 3.3.1.3 of [RFC6424] defines a new sub-TLV that a router must include when the FEC stack changes.
The network node which advertised the Node Segment ID is responsible for generating FEC Stack Change sub-TLV of &pop& operation for Node Segment ID, regardless of if PHP is enabled or not.
The network node that is immediate downstream of the node which advertised the Adjacency Segment ID is responsible for generating FEC Stack Change sub-TLV of &pop& operation for Adjacency Segment ID.
The forwarding semantic of Node Segment ID with PHP flag is equivalent to usage of implicit Null in MPLS protocols. Adjacency Segment ID is also similar in a sense that it can be thought as next hop destined locally allocated segment that has PHP enabled. Procedures described in Section 4.4 of [RFC4379] relies on Stack-D and Stack-R explicitly having Implicit Null value. It may simplify implementations to reuse Implicit Null for Node Segment ID PHP and Adjacency Segment ID cases. However, it is technically incorrect for Implicit Null value to externally appear. Therefore, implicit Null MUST NOT be placed in Stack-D and Interface and Label Stack TLV for Node Segment ID PHP and Adjacency Segment ID cases.
LSP Traceroute operation can properly traverse every hop of Segment Routing network in Uniform Model described in [RFC3443]. If one or more LSRs employ Short Pipe Model described in [RFC3443], then LSP Traceroute may not be able to properly traverse every hop of Segment Routing network due to absence of TTL copy operation when outer label is popped. In such scenario, following TTL manipulation technique MAY be used.
When tracing a LSP according to the procedures in [RFC4379] the TTL is incremented by one in order to trace the path sequentially along the LSP. However when a source routed LSP has to be traced there are as many TTLs as there are labels in the stack. The LSR that initiates the traceroute SHOULD start by setting the TTL to 1 for the tunnel in the LSP's label stack it wants to start the tracing from, the TTL of all outer labels in the stack to the max value, and the TTL of all the inner labels in the stack to zero. Thus a typical start to the traceroute would have a TTL of 1 for the outermost label and all the inner labels would have TTL 0. If the FEC Stack TLV is included it should contain only those for the inner stacked tunnels. The lack of an echo response or the Return Code/Subcode should be used to diagnose the tunnel as described in [RFC4379]. When the tracing of a tunnel in the stack is complete, then the next tunnel in the stack should be traced. The end of a tunnel can be detected from the "Return Code" when it indicates that the responding LSR is an egress for the stack at depth 1. Thus the traceroute procedures in [RFC4379] can be recursively applied to traceroute a source routed LSP.
Source stacking can be optionally used to apply services on the packet at a LSR along the path, where a label in the stack is used to trigger service application. A data plane failure detection and isolation mechanism should provide its functionality without applying these services. This is mandatory for services that are stateful, though for stateless services [RFC4379] could be used as-is. It MAY also provide a mechanism to detect and isolate faults within the service function itself.
To prevent services from being applied to an "echo request" packet, the TTL of service labels MUST be 0. However TTL processing rules of a service label must be the same as any MPLS label. Due to this a TTL of 0 in the service label would prevent the packet from being forwarded beyond the LSR that provides the service. To avoid this problem, the originator of the "echo request" MUST NOT include the service label in the label stack of an echo request above the tunnel label of the tunnel that is being currently traced. In other words the ingress must remove all service-labels above the label of the tunnel being currently traced, but retain service labels below it when sending the echo request. Note that load balancing may affect the path when the service labels are removed, resulting in a newer path being traversed. However this new path is potentially different only up to the LSR that provides the service. Since this portion of the path was traced when the tunnels above this tunnel in the stack were traced and followed the exact path as the source routed LSP, this should not be a major concern. Sometimes the newer path may have a problem that was not in the original path resulting in a false positive. In such a case the original path can be traversed by changing the label stack to reach the intermediate LSR with labels that route along each hop explicitly.
IANA is requested to assign 3 new Sub-TLVs from "Sub-TLVs for TLV Types 1, 16 and 21" sub-registry.
Sub-Type Sub-TLV Name Reference
---------- ----------------- ------------
TBD1 IPv4 Prefix Node Segment ID Section 4.1 (this document)
TBD2 IPv6 Prefix Node Segment ID Section 4.2 (this document)
TBD3 Adjacency Segment ID Section 4.3 (this document)
This document defines additional Sub-TLVs and follows the mechanism defined in [RFC4379]. So all the security consideration defined in [RFC4379] will be applicable for this document and in addition it does not impose any security challenges to be considered.
The authors would like to thank Stefano Previdi, Les Ginsberg, Balaji Rajagopalan and Harish Sitaraman for their review and comments.
The authors wold like to thank Loa Andersson for his comments and recommendation to merge drafts.
Tarek Saad
Cisco Systems
Email: tsaad@cisco.com
Siva Sivabalan
Cisco Systems
Email: msiva@cisco.com
Balaji Rajagopalan
Juniper Networks
Email: balajir@juniper.net
| [RFC6291] | Andersson, L., van Helvoort, H., Bonica, R., Romascanu, D. and S. Mansfield, "Guidelines for the Use of the "OAM" Acronym in the IETF", BCP 161, RFC 6291, DOI 10.17487/RFC6291, June 2011. |
| [RFC6425] | Saxena, S., Swallow, G., Ali, Z., Farrel, A., Yasukawa, S. and T. Nadeau, "Detecting Data-Plane Failures in Point-to-Multipoint MPLS - Extensions to LSP Ping", RFC 6425, DOI 10.17487/RFC6425, November 2011. |