Internet Engineering Task Force (IETF) L. Berger Request for Comments: 5710 LabN Category: Standards Track D. Papadimitriou ISSN: 2070-1721 Alcatel Lucent JP. Vasseur Cisco January 2010
PathErr Message Triggered MPLS and GMPLS LSP Reroutes
This document describes how Resource ReserVation Protocol (RSVP) PathErr messages may be used to trigger rerouting of Multi-Protocol Label Switching (MPLS) and Generalized MPLS (GMPLS) point-to-point Traffic Engineering (TE) Label Switched Paths (LSPs) without first removing LSP state or resources. Such LSP rerouting may be desirable in a number of cases, including, for example, soft-preemption and graceful shutdown. This document describes the usage of existing Standards Track mechanisms to support LSP rerouting. In this case, it relies on mechanisms already defined as part of RSVP-TE and simply describes a sequence of actions to be executed. While existing protocol definitions can be used to support reroute applications, this document also defines a new reroute-specific error code to allow for the future definition of reroute-application-specific error values.
Status of This Memo
This is an Internet Standards Track document.
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 5741.
Copyright (c) 2010 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.
The Resource ReserVation Protocol (RSVP), see [RFC2205], has been extended to support the control of Traffic Engineering (TE) Label Switched Paths (LSPs) for both Multi-Protocol Label Switching (MPLS) and Generalized MPLS (GMPLS) in, respectively, [RFC3209] and [RFC3473]. In all cases, a PathErr message is used to report errors to nodes upstream of the error-detecting node. As defined in [RFC2205] and left unmodified by [RFC3209], PathErr messages "do not change path state in the nodes through which they pass". Notwithstanding this definition, PathErr messages are most commonly used to report errors during LSP establishment, i.e., the RSVP-TE processing that occurs prior to the ingress receiving a Resv message. (See [RFC5711] for a broader discussion on PathErr message handling.) Support for such usage was enhanced via the introduction of the Path_State_Removed flag in [RFC3473], which enables a processing node to free related LSP state and resources. The usage of PathErr messages during LSP establishment was further covered in [RFC4920], which describes in detail how a node may indicate that it or one of its associated resources should be avoided, i.e., routed around, during LSP establishment.
PathErr messages can also be used to support a number of other cases that can occur after an LSP is established. This document focuses on the cases where PathErr messages can be used for a node to indicate that it desires an upstream node to reroute an LSP around the indicating node or resources associated with the indicating node. Some examples of such cases are soft-preemption and graceful shutdown (see [RFC5712] and [GRACEFUL]).
This document uses the terminology "reroute request" to refer to the indication by a node that an upstream reroute should take place. This document describes how a node can initiate a reroute request without disrupting LSP data traffic or, when so desired, with the disruption of data traffic and removal of LSP-associated state and resources. The applicability of this document is limited to point- to-point LSPs. Support for point-to-multipoint LSPs are for further study.
The mechanisms used to indicate reroute requests are derived from the mechanisms described in [RFC4920] and the error codes defined in [RFC4736]. This document describes (1) how a non-disruptive reroute request may be issued and, (2) based on an optional "timeout" period, how rerouting may be forced by removing LSP state and associated resources and signaling such removal. While this document describes how existing protocol definitions can be used to support rerouting, it also defines a new reroute-specific error code to allow for the future definition of reroute-application-specific error values.
This section describes how a downstream node can indicate that it desires a node upstream (along the LSP path) to initiate the rerouting of an LSP, and how the upstream nodes can respond to such a request. Initiating nodes, transit nodes, and ingress nodes are described separately.
When a transit or egress node desires to request the rerouting of an established LSP, it first determines if it can act on the reroute request locally. Such a check MUST be performed on the condition that the Explicit Route Object (ERO), see [RFC3209], received in the LSP's incoming Path message does not preclude LSP rerouting. Examples of requests that may trigger reroutes are avoiding an outgoing interface, a component, label resource, or a next hop not explicitly listed in the ERO. In all cases, the actual repair action SHOULD be performed after verification that the local policy allows local repair for that LSP/state. That is, any traffic-rerouting action (associated to this state) must be initiated and completed only as allowed by local node policy.
When the node cannot act locally, it MUST issue a PathErr message indicating its inability to perform local rerouting. The PathErr message MUST contain an ERROR_SPEC object of the format defined in [RFC2205] or [RFC3473]. Such a message MUST include one of the following combinations of error codes and error values:
1. "Notify/Local node maintenance required" to support backwards compatibility and to reroute around the local node.
2. "Notify/Local link maintenance required" to support backwards compatibility and to reroute around a local interface.
3. "Reroute/<any Reroute error value>" for future compatibility and when backwards compatibility is not a concern.
Berger, et al. Standards Track [Page 4]
RFC 5710 MPLS and GMPLS LSP Reroutes January 2010
The rest of the ERROR_SPEC object is constructed based on the local rerouting decision and the resource that is to be avoided by an upstream node. It is important to note that the address and TLVs carried by the ERROR_SPEC object identify the resource to be avoided and not the error code and value.
When the reroute decision redirects traffic around the local node, the local node MUST be indicated in the ERROR_SPEC object. Otherwise, i.e., when the reroute decision does not redirect traffic around the local node, the impacted interface MUST be indicated in the ERROR_SPEC object and the IF_ID [RFC3473] ERROR_SPEC object formats SHOULD be used to indicate the impacted interface.
The IF_ID [RFC3473] ERROR_SPEC object format MUST be used to indicate a reroute request that is more specific than an interface. The TLVs defined in [RFC3471], as updated by [RFC3477], [RFC4201], and [RFC4920] MAY be used to provide specific, additional reroute request information, e.g., reroute around a specific label. The principles related to ERROR_SPEC object construction, defined in Section 6.3.1 of [RFC4920], SHOULD be followed.
Reroute request timeouts are used to remove an LSP when there is no response to a reroute request. A reroute request timeout is used when an LSP is to be removed at the expiration of the reroute request timeout period. When such LSP removal is desired, and after initiating a reroute request, the initiating node MUST initiate a timeout during which it expects to receive a response to the reroute request. Valid responses are a PathTear message or a trigger Path message with an ERO, avoiding the resource that was indicated in the reroute request. If either type of message is received, the timeout period MUST be canceled and no further action is needed. Note, normal refresh processing is not modified by the introduction of reroute request timeouts. Such processing may result in Path state being removed during the timeout period, in which case the timeout period MUST also be canceled.
If the reroute request timeout is reached, the initiating node MUST remove the LSP and its associated state and resources. Removal of LSP state is indicated downstream via a corresponding PathTear message. Removal is indicated upstream via a PathErr message with the error code of "Service preempted". The Path_State_Removed flag MUST be set if supported. When the Path_State_Removed flag is not supported, a corresponding ResvTear MUST also be sent.
When a transit node's policy permits it to support reroute request processing and local repair, the node MUST examine incoming PathErr messages to see it the node can perform a requested reroute. A reroute request is indicated in a received PathErr message, which carries one of the error code and value combinations listed above in Section 2.1. Note that a conformant implementation MUST check for any of the three combinations listed in Section 2.1.
A transit node MAY act on a reroute request locally when the ERO received in the LSP's incoming Path message does not preclude the reroute. As before, examples include loosely routed LSP next hops. When the reroute request can be processed locally, standard, local repair processing MUST be followed. The node SHOULD limit the number of local repair attempts. Again, the expected norm is for local repair, and thereby this case, to be precluded due to policy.
When the transit node supports [RFC4920] and is a boundary node, and Boundary rerouting is allowed, it SHOULD use a route request as a trigger to reroute the LSP. (Per [RFC4920], the Flags field of the LSP_ATTRIBUTES object of the initial Path message indicates "Boundary rerouting".) In the case the node triggers rerouting, it first MUST identify an alternate path within the domain. When such a path is available, the node MUST terminate the PathErr message and issue a Path message reflecting the identified alternate path. Processing then continues per [RFC4920]. When an alternate path is not available, the node cannot act on the reroute request.
When a transit node cannot act on a reroute request locally, per standard processing, it MUST propagate the received PathErr message to the previous hop.
When reroute processing is supported, an ingress node MUST check received PathErr messages to identify them as indicating reroute requests. A reroute request is indicated in a received PathErr message, which carries one of the error code and value combinations listed above in Section 2.1. Note that a conformant implementation MUST check for any of the three combinations listed in Section 2.1.
Upon receiving a reroute request, the ingress MUST attempt to identify an alternate path, avoiding the node, interface, resource, etc. identified within the ERROR_SPEC object. When an alternate path cannot be identified, the reroute request MUST be discarded. When an
Berger, et al. Standards Track [Page 6]
RFC 5710 MPLS and GMPLS LSP Reroutes January 2010
alternate path is identified, a corresponding make-before-break LSP SHOULD be initiated and standard make-before-break procedures MUST be followed.
This section provides example reroute requests. This section is informative rather than prescriptive. Reroute requests are always sent via PathErr messages. As described above, a PathErr message may contain either an [RFC2205] format ERROR_SPEC object, or an IF_ID [RFC3473] format ERROR_SPEC object; it is the address and TLVs carried by the ERROR_SPEC object, and not the error value, that indicates the resource that is to be avoided by the reroute.
The node address is set to the local node's TE router address. Error code is set to either "Notify/Local link maintenance required" or "Reroute/<any Reroute error value>". IP address is set to the TE address of the interface to be avoided.
The node address is set to the local TE address used in the advertisement of the bundle associated with the component. Error code is set to either "Notify/Local link maintenance required" or "Reroute/<any Reroute error value>". Router ID is set to the local router ID, and Interface ID is the identifier assigned to the component link by the local node.
The node address is set to the local node's TE router address. Error code is set to either "Notify/Local link maintenance required" or "Reroute/<any Reroute error value>". IP address is set to the TE address of the interface that supports the label to be avoided. DOWNSTREAM_LABEL indicates the label to be avoided.
Sections 9 of [RFC4920] and [RFC4736] should be used as the starting point for reviewing the security considerations related to the formats and mechanisms discussed in this document. This document introduces a new error code, but this code is functionally equivalent to existing semantics, in particular, the error code/error value combinations of "Notify/Local node maintenance required" and "Notify/Local link maintenance required". As such, this document introduces no new security considerations beyond what already applies to these existing formats and mechanisms. Future documents may define new error values; any considerations specific to those values should be discussed in the document defining them. 6. References