RFC 5852 PROPOSED STANDARD GMPLS · RSVP-TE · TRANSPORT
RSVP-TE Handover: RSVP-TE Signaling Extension for LSP Handover from the Management Plane to the Control Plane in a GMPLS-Enabled Transport Network
How an RSVP-TE signaling flag moves an existing connection between management-system control and GMPLS control without touching traffic.
RFC EDITOR
Caviglia, D., Ceccarelli, D., Bramanti, D., Li, D., and S. Bardalai, "RSVP-TE Signaling Extension for LSP Handover from the Management Plane to the Control Plane in a GMPLS-Enabled Transport Network", RFC 5852, DOI 10.17487/RFC5852, April 2010, <https://www.rfc-editor.org/info/rfc5852>.
BIBTEX
@misc{rfc5852,
series = {Request for Comments},
number = 5852,
howpublished = {RFC 5852},
publisher = {RFC Editor},
doi = {10.17487/RFC5852},
url = {https://www.rfc-editor.org/info/rfc5852},
author = {D. Caviglia and D. Ceccarelli and D. Bramanti and D. Li and S. Bardalai},
title = {{RSVP-TE Signaling Extension for LSP Handover from the Management Plane to the Control Plane in a GMPLS-Enabled Transport Network}},
pagetotal = 23,
year = 2010,
month = apr,
} RFC text
Contents
Internet Engineering Task Force (IETF) D. Caviglia
Request for Comments: 5852 D. Ceccarelli
Category: Standards Track D. Bramanti
ISSN: 2070-1721 Ericsson
D. Li
Huawei Technologies
S. Bardalai
Fujitsu Network
April 2010
RSVP-TE Signaling Extension for LSP Handover from the Management Plane
to the Control Plane in a GMPLS-Enabled Transport Network
Abstract
In a transport network scenario, Data Plane connections controlled by
either a Generalized Multiprotocol Label Switching (GMPLS) Control
Plane (Soft Permanent Connections - SPC) or a Management System
(Permanent Connections - PC) may independently coexist. The ability
of transforming an existing PC into an SPC and vice versa -- without
actually affecting Data Plane traffic being carried over it -- is a
requirement. The requirements for the conversion between permanent
connections and switched connections in a GMPLS Network are defined
in RFC 5493.
This memo describes an extension to GMPLS Resource Reservation
Protocol - Traffic Engineering (RSVP-TE) signaling that enables the
transfer of connection ownership between the Management and the
Control Planes. Such a transfer is referred to as a Handover. This
document defines all Handover-related procedures. This includes the
handling of failure conditions and subsequent reversion to original
state. A basic premise of the extension is that the Handover
procedures must never impact an already established Data Plane
connection.
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.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc5852.
Copyright Notice
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.
Table of Contents
1. Introduction ....................................................4
1.1. Dedication .................................................4
2. Terminology .....................................................4
3. Motivation ......................................................4
4. Procedures ......................................................5
4.1. MP-to-CP Handover: LSP Ownership Transfer from
Management Plane to Control Plane ..........................6
4.2. MP-to-CP Handover Procedure Failure Handling ...............7
4.2.1. MP-to-CP Handover Failure - Path Failure ............8
4.2.1.1. MP-to-CP Handover Failure - Path
Message and Data Plane Failure .............8
4.2.1.2. MP-to-CP Handover Failure - Path
Message and Communication Failure ..........8
4.2.2. MP-to-CP Handover Failure - Resv Error ..............9
4.2.2.1. MP-to-CP Handover Failure - Resv
Error and Data Plane Failure ...............9
4.2.2.2. MP-to-CP Handover Failure - Resv
Error and Communication Failure ...........10
4.2.2.3. MP-to-CP Handover Failure - Node
Graceful Restart ..........................12
4.3. CP-to-MP Handover: LSP Ownership Transfer from
Control Plane to Management Plane .........................15
4.4. CP-to-MP Handover Procedure Failure .......................16
5. Minimum Information for MP-to-CP Handover ......................17
6. RSVP Message Formats ...........................................19
7. Objects Modification ...........................................19
7.1. Administrative Status Object ..............................19
7.2. Error Spec Object .........................................19
8. Compatibility ..................................................20
9. Security Considerations ........................................20
10. IANA Considerations ...........................................20
11. Acknowledgments ...............................................21
12. Contributors ..................................................21
13. References ....................................................21
13.1. Normative References .....................................21
13.2. Informative References ...................................22
1. Introduction
In a typical traditional transport network scenario, Data Plane (DP)
connections between two endpoints are controlled by means of a
Network Management System (NMS) operating within the Management Plane
(MP). NMS/MP is the owner of such transport connections, being
responsible for their setup, teardown, and maintenance.
The adoption of a Generalized MPLS (GMPLS) [RFC3945] Control Plane
(CP) in a network that is already in service -- controlled by the NMS
at the MP level -- introduces the need for a procedure able to
coordinate a controlled Handover of a Data Plane connection from the
MP to the CP.
In addition, the control Handover in the opposite direction, from CP
to MP should be possible as well. The procedures described in this
memo can be applied to a Label Switched Path (LSP) in any DP
switching technology and any network architecture.
This memo describes an extension to GMPLS Resource reSerVation
Protocol - Traffic Engineering (RSVP-TE) [RFC3471] [RFC3473]
signaling that enables the Handover of connection ownership between
the Management and the Control Planes. All Handover-related
procedures are defined below. This includes the handling of failure
conditions and subsequent reversion to original state. A basic
premise of the extension is that the Handover procedures must never
impact the exchange of user data on LSPs that are already established
in the DP.
1.1. Dedication
We would like to dedicate this work to our friend and colleague Dino
Bramanti, who passed away at the early age of 38. Dino has been
involved in this work since its beginning.
2. Terminology
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].
3. Motivation
The main motivation behind this work is the definition of a simple
and very low-impact procedure that satisfies the requirements defined
in [RFC5493]. Such a procedure is aimed at giving the transport
network operators the chance to hand over the ownership of existing
LSPs provisioned by NMS from the MP to the CP without disrupting user
traffic flowing on them. Handover from the MP to the CP (i.e., when
existing DP connection ownership and control is passed from the MP to
the CP) has been defined as a mandatory requirement, while the
opposite operation, CP-to-MP Handover, has been considered as a nice-
to-have feature that can be seen as an emergency procedure to disable
the CP and take manual control of the LSP. For more details on
requirements and motivations, please refer to [RFC5493].
4. Procedures
The modification defined in this document refers only to the
ADMIN_STATUS Object, that is, the message flow is left unmodified for
both LSP setup and deletion. Moreover, a new Error Value is defined
to identify the failure of a Handover procedure.
The following paragraphs give detailed description of the "MP-to-CP
Handover" and "CP-to-MP Handover" procedures, based on the use of a
newly defined bit called "H bit".
Just as when setting up an LSP using the CP [RFC3473], the Path
message may contain full information about the explicit route
including the links and labels traversed by the LSP. This
information is encoded in the Explicit Route Object (ERO), and must
be supplied by the MP using details recorded when the LSP was
provisioned, or collected by the MP by inspecting the nodes along the
path.
Alternatively, and also just as when setting up an LSP using the CP
[RFC3473], the ERO may include less information such that the details
of the next hop have to be determined by each node along the LSP as
it processes the Path message. This approach may be desirable when
the full information is not available to the MP or cannot be passed
to the head-end node when initiating the Handover from the MP to the
CP.
This section (Section 4) describes the general procedures and
protocol extensions for MP-to-CP Handover, and it uses the case of a
fully detailed ERO to describe the mechanism. Section 5 describes
how each node behaves in the case of a limited amount of information
in the ERO.
Note that when Handover is being performed for a bidirectional LSP
and the ERO contains full information including labels, the ERO
SHOULD include both upstream and downstream labels. Per Section
5.1.1 of [RFC3473], the labels are indicated on an output basis; this
means that the labels are used by the upstream node to create the
LABEL_SET Object and, for bidirectional LSPs, the UPSTREAM_LABEL
Object used in the outgoing Path message.
4.1. MP-to-CP Handover: LSP Ownership Transfer from Management Plane to
Control Plane
The MP-to-CP Handover procedure MUST create an RSVP-TE session along
the path of the LSP to be moved from the MP to the CP, associating it
with the existing cross-connected resources owned by the MP (e.g.,
lambdas, time slots, or reserved bandwidth) and at the same time
transferring their ownership to the CP.
The operator instructs the ingress node to hand over control of the
LSP from the MP to the CP. In this Handover mode, it supplies the
exact path of the LSP including any resource reservation and label
information.
The ingress MUST check that no corresponding Path state exists and
that corresponding Data Plane state does exist. If there is an
error, this MUST be reported to the operator and further protocol
action MUST NOT be taken.
The ingress signals the LSP using a Path message with the H bit and R
bit set in the ADMIN_STATUS Object. In this mode of Handover, the
Path message also carries an ERO that includes Label subobjects
indicating the labels used by the LSP at each hop. The ingress MUST
start the Expiration timer (see Section 4.2.1.2 for expiration of
this timer). Such a timer SHOULD be configurable per LSP and have a
default value of 30 seconds.
Each Label Switching Router (LSR) that receives a Path message with
the H bit set checks to see whether there is any matching Path state.
o If matching Path state is found with the H bit set, this is a Path
refresh and should be treated accordingly [RFC3473].
o If matching Path state is found with the H bit clear, this is an
error and MUST be treated according to the error case description
in Section 4.2.1.1.
o If no Path state is found, the LSR goes on to check whether there
is any matching Data Plane state.
o If no matching Data Plane state is found (including only partially
matching Data Plane state), this is an error or a race condition.
It MUST be handled according to the description in Section
4.2.1.1.
o If matching Data Plane state is found, the LSR MUST save the Path
state (including the set H bit), and it MUST forward the Path
message to the egress. The LSR MUST retain any MP state
associated with the LSP at this point.
An egress LSR MUST act as any other LSR, except that there is no
downstream node to which to forward the Path message. If all checks
are passed, the egress MUST respond with a Resv with the H bit set.
A transit LSR MUST process each Resv according to the normal rules of
[RFC3473].
When an ingress LSR receives a Resv message carrying the H bit set,
it checks the Expiration timer.
o If the timer is not running, the Resv is treated as a refresh and
no special action is taken [RFC3473].
o If the timer is running, the ingress MUST cancel the timer and
SHOULD notify the operator that the first stage of Handover is
complete. The ingress MUST send a Path message that is no
different from the previous message except that the H bit MUST be
clear.
The Path message with the H bit clear will travel the length of the
LSP and will result in the return of a Resv with the H bit clear
according to normal processing [RFC3473]. As a result, the H bit
will be cleared in the stored Path state at each transit LSR and at
the egress LSR. Each LSR SHOULD release any associated MP state
associated with the LSP when it receives the Path message with H bit
clear, but MAY retain the information according to local policy for
use in future MP processing.
When the ingress receives a Resv with the H bit clear, the Handover
is completed. The ingress SHOULD notify the operator that the
Handover is correctly completed.
4.2. MP-to-CP Handover Procedure Failure Handling
In the case of MP-to-CP Handover, two different failure scenarios can
happen: Path Failure and Resv Failure. Moreover, each failure can be
due to two different causes: DP Failure or Communication Failure. In
any case, the LSP ownership MUST be immediately rolled back to the
one previous to the Handover procedure. A section for each
combination will be analyzed in the following.
4.2.1. MP-to-CP Handover Failure - Path Failure
4.2.1.1. MP-to-CP Handover Failure - Path Message and Data Plane
Failure
In the following paragraph, we will analyze the case where the
Handover procedure fails during the Path message processing.
| Path | | |
|--------------->| Path | |
| |---------------X| |
| | PathErr | |
| PathErr |<---------------| |
|<---------------| | |
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 1: MP2CP - Path Msg and DP Failure
If an error occurs, the node detecting the error MUST respond to the
received Path message with a PathErr message, and MUST abort the
Handover procedure. The PathErr message SHOULD have the
Path_State_Removed flag set [RFC3473], but implementations MAY retain
their local state and wait for Path state timeout as per normal RSVP
processing.
Nodes receiving a PathErr message MUST follow standard PathErr
message processing and the associated DP resources MUST NOT be
impacted. If the local CP state indicates that a Handover is in
progress (based on the H bit in the Path message), the LSR MUST
revert the LSP ownership to the MP.
4.2.1.2. MP-to-CP Handover Failure - Path Message and Communication
Failure
Other possible scenarios are shown in the following figures and are
based on the inability to reach a node along the path of the LSP.
The below scenario postulates the use of a reliable message delivery
based on the mechanism defined in [RFC2961].
| Path | | |
|--------------->| Path | |
| |---------------X| |
| |---------------X| |
| | ... | |
| |---------------X| |
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 2: MP2CP - Path Msg and Communication Failure
(Reliable Delivery)
The Path message sent from LSR A towards LSR B is lost or does not
reach the destination for any reason. As a reliable delivery
mechanism is implemented, LSR A retransmits the Path message for a
configurable number of times, and if no ack is received, the Handover
procedure will be aborted (via the Expiration timer).
In the next scenario RSVP-TE messages are sent without reliable
message delivery, that is, no [RFC2961] MessageID procedure is used.
| Path | | |
|--------------->| Path | |
| |----------X | |
| | | |
TIMER EXPIRES | | |
| Path Tear | Path Tear | Path Tear |
|--------------->|--------------->|--------------->|
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 3: MP2CP - Path Msg and Communication Failure
(No Reliable Delivery)
If the Resv message is not received before the expiration of the
Expiration timer, the Handover procedure is aborted as described in
Section 4.2.1.1. Please note that any node that has forwarded a Path
(LSR A), i.e., has installed local path state, will send a PathTear
when that state is removed (according to [RFC2205]).
4.2.2. MP-to-CP Handover Failure - Resv Error
4.2.2.1. MP-to-CP Handover Failure - Resv Error and Data Plane Failure
In the case of a failure occurrence during the Resv message
processing (in case there has been any change in the Data Plane
during the signaling), the node MUST send a PathErr message [RFC2205]
in the upstream direction. The PathErr message is constructed and
processed as defined above in Section 4.2.1.1. The failure detection
node MUST also send a PathTear message downstream. The PathTear
message is constructed and processed as defined above in
Section 4.2.1.1.
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| |X---------------| |
| PathErr | PathTear | PathTear |
|<---------------|--------------->|--------------->|
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 4: MP2CP - Resv Error and DP Failure
In the case shown in Figure 4, the failure occurs in LSR A. A
PathTear message is sent towards B and a PathErr message (with
ErrorCode set to "Handover Procedure Failure") is sent in the
upstream direction. The PathErr and PathTear messages remove the
Path state established by the Path messages along the nodes of the
LSP and abort the Handover procedure.
Please note that the failure occurred after the Handover procedure
was successfully completed in LSR B, but Handover state will still be
maintained locally as, per Section 4.1, a Path message with the H bit
clear will have not yet been sent or received. A node that receives
a PathTear when it has Path state with the H bit set MUST remove Path
state, but MUST NOT change Data Plane state. It MUST return LSP
ownership to the MP.
4.2.2.2. MP-to-CP Handover Failure - Resv Error and Communication
Failure
When a Resv message cannot reach one or more of the upstream nodes,
the procedure is quite similar to the one previously seen about the
Path message. Even in this case, it is possible to distinguish two
different scenarios.
In the first scenario we consider the utilization of a reliable
message delivery based on the mechanism defined in [RFC2961]. After
a correct forwarding of the Path message along the nodes of the LSP,
the Egress LSR sends a Resv message in the opposite direction. It
might happen that the Resv message does not reach the ingress Label
Edge Router (LER) or an LSR, say LSR A. LSR B MUST send a Resv
message again for a configurable number of times and then, if the
delivery does not succeed, the adoption procedure will be aborted
(via the Expiration timer).
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| | X---------| |
| | X---------| |
| | ... | |
| | X---------| |
| | | |
Ingress
LSR A LSR B Egress LER
Figure 5: MP2CP - Resv Error and Communication Failure
(Reliable Delivery)
Considering that the Resv message did not manage to reach LSR A, it
is highly probable that the PathErr would fail too. Due to this
fact, the Expiration timer is used on the ingress LER after sending
the path and on LSR A after forwarding it. When the timer expires,
if no Resv or PathErr message is received, the Handover procedure is
aborted as described in Section 4.2.1.1 and the LSP ownership is
returned to the Management Plane.
Figure 6, on the other hand, shows the scenario in which no reliable
delivery mechanism is implemented.
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| | X---------| |
TIMER EXPIRES | | |
| Path Tear | Path Tear | Path Tear |
|--------------->|--------------->|--------------->|
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 6: MP2CP - Resv Error and Communication Failure
(No Reliable Delivery)
If no Resv message is received before the Expiration timer expires,
the ingress LER follows the same procedure defined in Section 4.1.
4.2.2.3. MP-to-CP Handover Failure - Node Graceful Restart
If node restart and graceful restart are enabled, then one of the
following scenarios will happen.
Case I - Finite Restart Time
In this case, the Restart Time (see [RFC3473]) is finite, i.e., not a
value of 0xffffffff. In the sequence diagram below, assume LSR A
restarts. If the ingress LER does not receive the Resv message in
time, it MUST abort the Handover process by generating a PathTear
message downstream. Also, if LSR A does not complete the restart
process within the restart time interval, then LSR B MUST start
tearing down all LSPs between LSR A and LSR B and this includes the
LSP that is being used to carry out the Handover of MP resources to
the CP. LSP B MUST generate a PathTear message downstream and a
PathErr message upstream. Both LSR B and the egress LER MUST NOT
release the DP resources because, in both nodes, the H bit is set in
the local Path state.
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| X X---------| |
| PathTear | |
|-------X Restart Timer |
| Expires |
| PathErr | PathTear |
| X--------|--------------->|
| | |
| X | |
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 7: MP2CP - Node Graceful Restart - Case I
Case II - Infinite Restart Time
In this case, the Restart Time (see [RFC3473]) indicates that the
restart of the sender's Control Plane may occur over an indeterminate
interval, i.e., is 0xffffffff. The sequence is quite similar to the
previous one. In this sequence, the restart timer will not expire in
LSR B since it is run infinitely. Instead, after LSR A restarts, LSR
B MUST start the recovery timer. The recovery timer will expire
since there will be no Path message with the RECOVERY LABEL received
from LSR A given the ingress node had already removed the local Path
state after it aborts the Handover process. Thus, LSR B MUST tear
down the specific LSP that is being used to convert the MP resources
to CP by generating a PathTear message downstream and PathErr message
upstream. Similarly to the previous case, both LSR B and the egress
LER MUST NOT release the DP resources because the H bit is set in the
local Path state.
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| X X---------| |
| PathTear | |
|-------X | |
| | |
| X | |
| | | |
| | Recovery Timer |
| | Expires |
| PathErr | PathErr | PathTear |
|<---------------|<---------------|--------------->|
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 8: MP2CP - Node Graceful Restart - Case II
Case III
In this case, the ingress LER does not abort the Handover process.
When LSR A restarts, the ingress LER detects the restart and MUST
re-generate the Path message with the H bit set in order to restart
the Handover.
When LSR B receives the Path message, it sees the H-bit set on the
message and also sees that it has the H-bit set in its own state and
that it has sent the Resv. But it is also aware that LSR A has
restarted and could have sent a Path message with a RECOVERY LABEL
object. LSR B may attempt to resume the Handover process or may
abort the Handover. This choice is made according to local policy.
If resuming the Handover, LSR B MUST treat the received Path message
as a retransmission, and MUST retransmit its Resv. If aborting
Handover, LSR B MUST return a PathErr and MUST send a PathTear
downstream. In both cases, LSR B MUST NOT modify the DP state.
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| X X---------| |
| | |
| X | |
| | | |
| Path | Path | |
|--------------->|--------------->| |
| PathErr | PathErr | PathTear |
|<---------------|<---------------|--------------->|
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 9: MP2CP - Node Graceful Restart - Case III
4.3. CP-to-MP Handover: LSP Ownership Transfer from Control Plane to
Management Plane
Let's now consider the case of LSP ownership transfer from Control
Plane to Management Plane. Also in this section, we will analyze the
Handover procedure success and failure.
The scenario is still a DP connection between two nodes acting as
ingress and egress for a LSP, but in this case, the CP has the
ownership and control of the LSP. The CP-to-MP Handover procedure
MUST delete the existing RSVP-TE session information and MUST NOT
affect the cross-connected resources, but just move their ownership
to the MP.
In other words, after LSP ownership transfer from CP to MP, the LSP
is no longer under the control of RSVP-TE, which is no more able to
"see" the LSP itself. The CP-to-MP Handover procedure MUST be a
standard LSP deletion procedure as described in Section 7.2.1 of
[RFC3473]. The procedure is initiated at the ingress node of the LSP
by an MP entity. The ingress node and MP exchange the relevant
information for this task and then propagate it over CP by means of
RSVP-TE tear down signaling as described below.
The ingress node MUST send a Path message in the downstream direction
with Handover and Reflect bits set in the ADMIN_STATUS Object. No
action is taken over the DP and transit LSRs must forward such
message towards the egress node. All of the nodes MUST keep track of
the procedure storing the H bit in their local Path and Resv states.
Then, every node waits for the H bit to be received within the
related Resv message. After the Resv message is received by the
ingress LER, it MUST send a PathTear message in order to clear the
whole LSP information recorded on the RSVP-TE data structures of the
nodes. Downstream nodes processing a PathTear message that follows a
Path message with the H bit set, MUST NOT remove any associated Data
Plane state. In other words, a normal LSP tear down signaling is
exchanged between nodes traversed by the LSP, but the H bit set in
the Path message indicates that no DP action has to correspond to CP
signaling.
4.4. CP-to-MP Handover Procedure Failure
Failures during CP-to-MP Handover procedure MUST NOT result in the
removal of any associated Data Plane state. To that end, when a Resv
message containing an ADMIN_STATUS Object with the H bit not received
during the period of time described in Section 7.2.2. of [RFC3473]
different processing is required. While the H bit is set in the Path
state, a node MUST NOT send a PathTear when a failure is detected.
Instead, the failure is reported upstream using a PathErr. The only
node that can send a PathTear is the ingress node, and it can only do
this as a step in the procedures specified in this document. This
applies to both MP-to-CP and CP-to-MP Handover. The ingress node MAY
choose to report the failure in the CP-to-MP Handover procedure via
the MP.
The CP-to-MP Handover procedure can also fail due to two causes:
PathTear lost or node down. In the former case, if the LSP is not
under MP control after the Expiration timer elapses, a manual
intervention from the network operator is requested, while in the
latter case, different scenarios may happen:
- CASE I - Path message and node down
| Path | Path X |
|--------------->|--------------X |
| | |
| | X |
| | | |
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 10: Case I - Path Message and Node Down
Per [RFC3473], Section 7.2.2, the ingress node should wait for a
configurable amount of time (30 seconds by default) and then send a
PathTear message. In this case, the normal deletion procedure MUST
NOT be followed. When the Expiration timer elapses, a manual
intervention from network operator is requested and normal, i.e.,
pre-CP-to-MP Handover, LSP processing continues.
- CASE II - Resv message and node down
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| | | Resv |
| | Resv |<---------------|
| X X---------| |
| | |
| X | |
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 11: Case II - Resv Message and Node Down
The procedure to be followed is the same depicted in CASE I. The
network operator can ask for the automatic CP-to-MP procedure again
after the failed node comes back up. Per [RFC3473], section 7.2, the
nodes will forward the new Path and Resv messages correctly.
- CASE III - PathTear message and node down
| Path | Path | Path |
|--------------->|--------------->|--------------->|
| Resv | Resv | Resv |
|<---------------|<---------------|<---------------|
| PathTear | | |
|--------------->| PathTear X |
| |------------X |
| | X |
| | | |
Ingress LER LSR A LSR B Egress LER
Figure 12: Case III - PathTear Message and Node Down
This scenario can be managed as a normal PathTear lost described
above in this section.
5. Minimum Information for MP-to-CP Handover
As described in Section 4, it is also possible for the ERO to contain
less than the full set of path information for the LSP being handed
over. This arises when only a minimal set of information is handed
to the CP by the MP at the LSP's head-end. Instead of collecting all
of the LSP information (including the labels) and formatting it into
an ERO, as described in Section 4, it is possible to start with a
minimum amount of information. The full ERO method and the
partial/no ERO method are not mutually exclusive; support of both
methods is required.
At the ingress node, the information needed to specify the LSP is the
outgoing interface ID, upstream label, and downstream label of this
interface and egress node ID. The remaining information about an
existing LSP can then be collected hop by hop, as the signaling is
going on, by looking up the cross-connection table in the DP at each
node along the LSP path.
Starting from the information available at the ingress LER about the
outgoing interface ID of that ingress node, the incoming interface ID
of the next hop can be found by looking up the link resource table/
database in the LER itself.
The Path message is hence built with the LABEL_SET Object ([RFC3473])
and the UPSTREAM_LABEL Object ([RFC3473]), where the upstream label
and downstream label of ingress outgoing interface of the LSP are
included in these two objects. In addition to the above mentioned
objects, the Path message MUST include the ADMIN_STATUS Object with
the H bit set, as already defined in previous chapters for the
detailed ERO-based way of proceeding. Such a Handover Path is sent
to the incoming interface of the next hop. When this Path message
reaches the second node along the LSP, the information about incoming
interface ID and the upstream and downstream labels of this interface
is extracted from it, and it is used to find next hop outgoing
interface ID and the upstream/downstream labels by looking up the DP
cross-connection table.
After having determined, in this way, the parameters describing the
LSPs next hop, the outgoing Path message to be sent is built
replacing the LABEL_SET Object and UPSTREAM_LABEL Object content with
the looked-up values of upstream and downstream labels.
By repeating this procedure for each transit node along the LSP, it
is possible to make the Handover Path message reach the egress node,
exactly following the LSP that is in place over DP. The ERO MAY, in
this case, be included in the Path message as an optional object, and
MAY be filled with the LSP-relevant information down to either the
port level with the interface ID or the label level with upstream and
downstream labels. The ERO can be used to check the consistency of
resource in the DP down to the port level or label level at each
intermediate node along the LSP.
Where the DP path continues beyond the egress, by indicating the
Egress label at the head-end of an LSP, the traffic can be directed
to the right destination. The GMPLS signaling procedure for egress
control is described in [RFC4003]
6. RSVP Message Formats
This memo does not introduce any modification in RSVP messages object
composition.
7. Objects Modification
The modifications required concern two RSVP objects: the ADMIN_STATUS
and ERROR_SPEC Objects.
7.1. Administrative Status Object
This memo introduces a new flag into the ADMIN_STATUS Object. The
ADMIN_STATUS Object is defined in [RFC3473]. This document uses the
H bit of the ADMIN_STATUS Object. The bit is bit number 25.
7.2. Error Spec Object
It is possible that a failure, such as the loss of the Data
Communication Network (DCN) connection or the restart of a node,
occurs during the LSP ownership handing over. In this case, the LSP
Handover procedure is interrupted, the ownership of the LSP must
remain with the ownership prior to the initiation of the Handover
procedure. It is important that the transaction failure not affect
the DP. The LSP is kept in place and no traffic hit occurs.
The failure is signaled by a PathErr message in the upstream
direction and PathTear messages in the downstream direction. The
PathErr messages include an ERROR_SPEC Object specifying the causes
of the failure.
This memo introduces a new Error Code (with different Error Values)
into the ERROR_SPEC Object, defined in [RFC2205].
The defined Error Code is "Handover Procedure Failure", and its value
is 35. For this Error Code, the following Error Value sub-codes are
defined:
1 = Cross-connection mismatch
2 = Other failure
8. Compatibility
The main requirement for the Handover procedure to work is that all
nodes along the path MUST support the extension defined in this
document. This requirement translates to an administrative
requirement as it is not enforced at the protocol level. As defined,
non-supporting nodes will simply propagate the H bit without setting
local state. This may result in an impact on data traffic during the
Handover procedure.
9. Security Considerations
The procedures described in this document rely completely on RSVP-TE
messages and mechanism. The use of the H bit being set in the
ADMIN_STATUS Object basically informs the receiving entity that no
operations are to be done over the DP as a consequence of such
special signaling flow. Using specially flagged signaling messages,
we want to limit the function of setup and teardown messages to the
CP, making them ineffective over related DP resource usage.
However, the Handover procedures allow the Control Plane to be used
to take an LSP out of the control of the Management Plane. This
could cause considerable disruption and could introduce a new
security concern. As a consequence, the use of GMPLS security
techniques is more important. For RSVP-TE security, please refer to
[RFC3473], for the GMPLS security framework, please refer to
[sec-fwk].
10. IANA Considerations
IANA manages the bit allocations for the ADMIN_STATUS Object
([RFC3473]). This document requires the allocation of the Handover
bit: the H bit. IANA has allocated a bit for this purpose.
Bit Number Hex Value Name Reference
---------- ----------- --------------------------------- ---------
25 0x00000040 Handover (H) [RFC5852]
IANA has also allocated a new Error Code:
35 Handover failure
This Error Code has the following globally defined Error
Value sub-codes:
1 = Cross-connection mismatch
2 = Other failure
11. Acknowledgments
We wish to thank Adrian Farrel, Lou Berger, Alan Elder, and Ben
Campbell for their assistance and precious advice to prepare this
document for publication. We also wish to thank Nicola Ciulli
(Nextworks) who contributed to the initial stage of this document.
12. Contributors
Shan Zhu
Fujitsu Network Communications Inc.
2801 Telecom Parkway,
Richardson, TX 75082
USA
EMail: Shan.Zhu@us.fujitsu.com
Tel: +1-972-479-2041
Igor Bryskin
ADVA Optical Networking Inc
7926 Jones Branch Drive, Suite 615
McLean, VA 22102
USA
EMail: ibryskin@advaoptical.com
Francesco Fondelli
Ericsson
Via Negrone 1A
Genova - 16145
Italy
EMail: francesco.fondelli@ericsson.com
Lou Berger
LabN Consulting, LLC
Phone: +1 301 468 9228
EMail: lberger@labn.net
13. References
13.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC2205] Braden, B., Zhang, L., Berson, S., Herzog, S., and S.
Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1
Functional Specification", RFC 2205, September 1997.
[RFC2961] Berger, L., Gan, D., Swallow, G., Pan, P., Tommasi, F.,
and S. Molendini, "RSVP Refresh Overhead Reduction
Extensions", RFC 2961, April 2001.
[RFC3471] Berger, L., "Generalized Multi-Protocol Label Switching
(GMPLS) Signaling Functional Description", RFC 3471,
January 2003.
[RFC3473] Berger, L., "Generalized Multi-Protocol Label Switching
(GMPLS) Signaling Resource ReserVation Protocol-Traffic
Engineering (RSVP-TE) Extensions", RFC 3473, January 2003.
[RFC3945] Mannie, E., "Generalized Multi-Protocol Label Switching
(GMPLS) Architecture", RFC 3945, October 2004.
[RFC4003] Berger, L., "GMPLS Signaling Procedure for Egress
Control", RFC 4003, February 2005.
13.2. Informative References
[RFC5493] Caviglia, D., Bramanti, D., Li, D., and D. McDysan,
"Requirements for the Conversion between Permanent
Connections and Switched Connections in a Generalized
Multiprotocol Label Switching (GMPLS) Network", RFC 5493,
April 2009.
[sec-fwk] Fang, L. and M. Behringer, "Security Framework for MPLS
and GMPLS Networks", Work in Progress, March 2010.
Authors' Addresses
Diego Caviglia
Ericsson
Via A. Negrone 1A
Genova - Sestri Ponente 16153
Italy
EMail: diego.caviglia@ericsson.com
Daniele Ceccarelli
Ericsson
Via A. Negrone 1A
Genova - Sestri Ponente 16153
Italy
EMail: daniele.ceccarelli@ericsson.com
Dino Bramanti
Ericsson
Dan Li
Huawei Technologies
F3-5-B R&D Center, Huawei Base
Shenzhen 518129
P.R. China
EMail: danli@huawei.com
Snigdho Bardalai
Fujitsu Network
2801 Telecom Parkway
Richardson, TX 75082
USA
EMail: sbardalai@gmail.com
AI ELI5
On this page
PROBLEMTransport networks can have connections owned by a management system (permanent) or by a GMPLS control plane (soft permanent). Operators need to switch ownership of a live connection without disturbing the traffic on it.
MECHANISMA new Handover (H) bit in the RSVP-TE ADMIN_STATUS object marks special Path and Resv exchanges that move ownership of a label switched path while leaving data plane state alone. A new error code reports failures and ownership rolls back.
WHY IT MATTERSA path can be handed from the management plane to the control plane, and back, with defined behavior for failures and restarts. All nodes on the path must support the extension.
§1Handing Over Connection Ownership
In a traditional transport network, a Network Management System (NMS) in the Management Plane (MP) owns the data plane connections between endpoints and handles their setup, teardown and maintenance. When a GMPLS Control Plane (CP) is added to a network already in service, there must be a way to hand a connection from the MP to the CP in a controlled way.
The abstract calls connections owned by the CP Soft Permanent Connections (SPC) and those owned by the management system Permanent Connections (PC). RFC 5493 sets the requirement for converting between them. This RFC extends GMPLS RSVP-TE signaling (RFC 3471, RFC 3473) to do so, a transfer it calls a Handover.
A basic premise is that Handover procedures must never impact user data on LSPs already established in the data plane. MP-to-CP Handover is a mandatory requirement. CP-to-MP Handover is described as a nice-to-have, an emergency procedure to disable the CP and take manual control of the LSP.
A connection can be run by a manager or by an automatic system. This memo says how to swap who is in charge without interrupting the people using it.
§2MP-to-CP Handover
The procedure changes only the ADMIN_STATUS object, with a new H bit. Message flow for LSP setup and deletion is left unmodified. The operator tells the ingress node to hand over an LSP and supplies its exact path, including resources and labels.
The ingress MUST check that no Path state exists and that data plane state does exist. It then sends a Path message with the H bit and the R bit set, carrying an ERO with Label subobjects, and starts an Expiration timer that SHOULD be configurable per LSP with a default of 30 seconds.
Each LSR that receives the Path message with the H bit set checks for matching state. A refresh is treated normally, a clash with H-clear state is an error, and missing data plane state is an error or race. If matching data plane state is found, the LSR MUST save the Path state with the H bit and forward the message, and it MUST retain any MP state at that point. The egress answers with a Resv carrying the H bit.
When the ingress receives that Resv while the timer runs, it cancels the timer, SHOULD tell the operator the first stage is done, and sends an unchanged Path message with the H bit clear. LSRs SHOULD release their MP state when they see it. When the ingress gets a Resv with the H bit clear, the Handover is complete.
The first person sends a note down the line saying, "We are about to change drivers." Everyone checks the bus is really there and answers. A second note says the change is done, and the old manager lets go.
IngressTransit LSRsEgress
- Path (H and R bits set, ERO with labels)Ingress starts the Expiration timer.
- Path (H bit set)
- Resv (H bit set)
- Resv (H bit set)Ingress cancels the timer.
- Path (H bit clear)
- Path (H bit clear)
- Resv (H bit clear)
- Resv (H bit clear)Handover is complete.
§3When MP-to-CP Handover Fails
Two failures can occur, a Path failure or a Resv failure, each caused by either a data plane failure or a communication failure. In every case LSP ownership MUST be rolled back immediately to what it was before the Handover.
If a node hits an error on the Path message, it MUST answer with a PathErr and abort. The PathErr SHOULD have the Path_State_Removed flag set, and the data plane resources MUST NOT be impacted. A node that sees a PathTear while holding Path state with the H bit set MUST remove the Path state, MUST NOT change data plane state, and MUST return ownership to the MP.
When messages are lost, the Expiration timer ends the attempt. With reliable delivery (RFC 2961) the sender retries first. Without it, the ingress sends a PathTear when the timer expires.
For node restarts with graceful restart, the RFC describes three cases. With a finite Restart Time, the neighbor tears down the LSP if the restarting node does not recover, but does not release data plane resources because the H bit is set. With an infinite Restart Time, a recovery timer decides. In the third case the ingress re-sends the Path message with the H bit set, and the neighbor either resumes or aborts according to local policy, never modifying the data plane.
If anything goes wrong during the swap, the old manager takes the wheel back at once. The bus never stops, and the setup is cleaned up.
§4CP-to-MP Handover
Going the other way, the control plane gives up its LSP and the management plane takes over. The procedure MUST delete the RSVP-TE session information and MUST NOT affect the cross-connected resources. It is a standard LSP deletion as in RFC 3473, started at the ingress by an MP entity.
The ingress MUST send a Path message with the Handover and Reflect bits set. Transit LSRs forward it and nodes store the H bit in their Path and Resv state. After the ingress receives the Resv, it MUST send a PathTear. Downstream nodes processing a PathTear that follows an H-bit Path message MUST NOT remove any data plane state.
Failures must not remove data plane state either. While the H bit is set, a node MUST NOT send a PathTear on failure and reports upstream with a PathErr instead. Only the ingress can send a PathTear, and only as a step in these procedures. If a PathTear is lost and the LSP is not under MP control after the Expiration timer elapses, the RFC asks for manual intervention by the operator.
To give the steering back, a note goes along the line saying "tear down the automatic paperwork but leave the bus alone." Only the front of the line is allowed to cancel the paperwork.
§5Handover With Minimal Information
The ERO need not carry the full path. The full ERO method and the partial or no-ERO method are not mutually exclusive, and support of both is required.
With minimal information, the ingress needs the outgoing interface ID, the upstream and downstream labels of that interface, and the egress node ID. The Path message carries a LABEL_SET object and an UPSTREAM_LABEL object with those labels, and MUST include the ADMIN_STATUS object with the H bit set. Each node on the path extracts the incoming interface and labels, looks up its data plane cross-connection table to find the next hop's outgoing interface and labels, and builds the next outgoing Path message from those values.
Repeating this at every transit node lets the Handover Path message follow the LSP that already exists in the data plane. The ERO MAY be included as an optional object and used to check resource consistency at port or label level.
If the first person only knows the first stretch of road, each stop along the way looks up the next stretch in its own map, until the note reaches the end.
§6The H Bit and the Error Code
The RFC adds no changes to the RSVP message composition. It changes two objects. The H bit is new in the ADMIN_STATUS object defined in RFC 3473, and it is bit number 25 (hex value 0x00000040). IANA has allocated it.
It also adds a new Error Code to the ERROR_SPEC object (from RFC 2205): Handover Procedure Failure, value 35. Failures are signaled by a PathErr upstream and PathTear messages downstream, and the PathErr carries the error. Two Error Value sub-codes are defined: 1 for a cross-connection mismatch and 2 for any other failure. On such a failure, ownership stays where it was before the Handover and no traffic hit occurs.
There is one new flag that means "this is only a paperwork change." There is also a new kind of error message that says the swap failed.
- 25ADMIN_STATUS H bit — Marks signaling as a Handover; hex value 0x00000040.
- 35/1Cross-connection mismatch — Error Code 35 with Error Value 1.
- 35/2Other failure — Error Code 35 with Error Value 2.
§7Compatibility
For Handover to work, all nodes along the path MUST support this extension. This is an administrative requirement rather than something enforced at the protocol level. Nodes that do not support it simply propagate the H bit without setting local state, and that may impact data traffic during the Handover.
Everyone along the line has to know the new rules. If someone does not, they just pass the note along without understanding it, and that could bump the passengers.
§8Security Considerations
The procedures rely entirely on RSVP-TE messages and mechanisms. The H bit tells the receiver that no data plane operations follow from this signaling, which limits setup and teardown messages to the control plane.
However, Handover lets the control plane take an LSP out of the management plane's control. The RFC says this could cause considerable disruption and could introduce a new security concern, so GMPLS security techniques become more important. It refers readers to RFC 3473 for RSVP-TE security.
Because the swap lets one system take charge from another, it is important to make sure only the right people can ask for it.
§9References
[RFC3473]RFC 3473Berger, L., "Generalized Multi-Protocol Label Switching (GMPLS) Signaling Resource ReserVation Protocol-Traffic Engineering (RSVP-TE) Extensions", RFC 3473, January 2003.February 2003[RFC2205]RFC 2205Braden, B., Zhang, L., Berson, S., Herzog, S., and S. Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1 Functional Specification", RFC 2205, September 1997.September 1997[RFC2961]RFC 2961Berger, L., Gan, D., Swallow, G., Pan, P., Tommasi, F., and S. Molendini, "RSVP Refresh Overhead Reduction Extensions", RFC 2961, April 2001.April 2001[RFC3471]RFC 3471Berger, L., "Generalized Multi-Protocol Label Switching (GMPLS) Signaling Functional Description", RFC 3471, January 2003.February 2003[RFC3945]RFC 3945Mannie, E., "Generalized Multi-Protocol Label Switching (GMPLS) Architecture", RFC 3945, October 2004.November 2004[RFC5493]RFC 5493Caviglia, D., Bramanti, D., Li, D., and D. McDysan, "Requirements for the Conversion between Permanent Connections and Switched Connections in a Generalized Multiprotocol Label Switching (GMPLS) Network", RFC 5493, April 2009.April 2009