Network Working Group Manoj Leelanivas (Juniper Networks)
Internet Draft Yakov Rekhter(Juniper Networks)
Expiration Date: March 2003 Rahul Aggarwal (Redback Networks)
Graceful Restart Mechanism for LDP
draft-ietf-mpls-ldp-restart-05.txt
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026, except that the right to
produce derivative works is not granted.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that other
groups may also distribute working documents as Internet-Drafts.
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.''
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Abstract
This document describes a mechanism that helps to minimize the
negative effects on MPLS traffic caused by Label Switching Router's
(LSR's) control plane restart, and specifically by the restart of its
Label Distribution Protocol (LDP) component, on LSRs that are capable
of preserving the MPLS forwarding component across the restart.
The mechanism described in this document is applicable to all LSRs,
both those with the ability to preserve forwarding state during LDP
restart and those without (although the latter need to implement only
a subset of the mechanism described in this document). Supporting (a
subset of) the mechanism described here by the LSRs that can not
preserve their MPLS forwarding state across the restart would not
reduce the negative impact on MPLS traffic caused by their control
plane restart, but it would minimize the impact if their neighbor(s)
are capable of preserving the forwarding state across the restart of
draft-ietf-mpls-ldp-restart-05.txt [Page 1]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
their control plane and implement the mechanism described here.
The mechanism makes minimalistic assumptions on what has to be
preserved across restart - the mechanism assumes that only the actual
MPLS forwarding state has to be preserved; the mechanism does not
require any of the LDP-related state to be preserved across the
restart.
The procedures described in this document apply to downstream
unsolicited label distribution. Extending these procedures to
downstream on demand label distribution is for further study.
Specification of Requirements
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 RFC 2119 [RFC2119].
Summary for Sub-IP Area
Summary
This document describes a mechanism that helps to minimize the
negative effects on MPLS traffic caused by LSR's control plane
restart, and specifically by the restart of its LDP component, on
LSRs that are capable of preserving the MPLS forwarding component
across the restart.
Related documents
See the Reference Section
draft-ietf-mpls-ldp-restart-05.txt [Page 2]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
Where does it fit in the Picture of the Sub-IP Work
This work fits squarely in MPLS box.
Why is it Targeted at this WG
The LDP is a product of the MPLS WG. This document specifies
procedures to minimize the negative effects caused by the restart of
the control plane LDP module. Since the procedures described in this
document are directly related to LDP, it would be logical to target
this document at the MPLS WG.
Justification
The WG should consider this document, as it allows to minimize the
negative effects caused by the restart of the control plane LDP
module.
1. Motivation
For the sake of brevity in the context of this document by "the
control plane" we mean "the LDP component of the control plane".
For the sake of brevity in the context of this document by "MPLS
forwarding state" we mean either <incoming label -> (outgoing label,
next hop)> (non-ingress case), or <FEC->(outgoing label, next hop)>
(ingress case) mapping.
In the case where a Label Switching Router (LSR) could preserve its
MPLS forwarding state across restart of its control plane, and
specifically its LDP component [LDP], it is desirable not to perturb
the LSPs going through that LSR (and specifically, the LSPs
established by LDP). In this document, we describe a mechanism,
termed "LDP Graceful Restart", that allows to accomplish this goal.
The mechanism described in this document is applicable to all LSRs,
both those with the ability to preserve forwarding state during LDP
restart and those without (although the latter need to implement only
a subset of the mechanism described in this document). Supporting (a
subset of) the mechanism described here by the LSRs that can not
preserve their MPLS forwarding state across the restart would not
reduce the negative impact on MPLS traffic caused by their control
plane restart, but it would minimize the impact if their neighbor(s)
are capable of preserving the forwarding state across the restart of
their control plane and implement the mechanism described here.
draft-ietf-mpls-ldp-restart-05.txt [Page 3]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
The mechanism makes minimalistic assumptions on what has to be
preserved across restart - the mechanism assumes that only the actual
MPLS forwarding state has to be preserved. Clearly this is the
minimum amount of state that has to be preserved across the restart
in order not to perturb the LSPs traversing a restarting LSR. The
mechanism does not require any of the LDP-related state to be
preserved across the restart.
In the scenario where label binding on an LSR is created/maintained
not just by the LDP component of the control plane, but by other
protocol components as well (e.g., BGP, RSVP-TE), and the LSR
supports restart of the individual components of the control plane
that create/maintain label binding (e.g., restart of LDP, but no
restart of BGP), the LSR needs to preserve across the restart the
information about which protocol has assigned which labels.
The procedures described in this document apply to downstream
unsolicited label distribution. Extending these procedures to
downstream on demand label distribution is for further study.
2. LDP Extension
An LSR indicates that it is capable of supporting LDP Graceful
Restart, as defined in this document, by including the Fault Tolerant
(FT) Session TLV as an Optional Parameter in the LDP Initialization
message. The format of the FT Session TLV is defined in [FT-LDP].
The L (Learn from Network) flag MUST be set to 1, which indicates
that the procedures in this document are used. The rest of the FT
flags are set to 0 by a sender and ignored on receipt.
The value field of the FT Session TLV contains two components that
are used by the mechanisms defined in this document: FT Reconnect
Timeout, and Recovery Time.
The FT Reconnect Timeout is the time (in milliseconds) that the
sender of the TLV would like the receiver of that TLV to wait after
the receiver detects the failure of LDP communication with the
sender. While waiting, the receiver SHOULD retain the MPLS
forwarding state for the (already established) LSPs that traverse a
link between the sender and the receiver. The FT Reconnect Timeout
should be long enough to allow the restart of the control plane of
the sender of the TLV, and specifically its LDP component to bring it
to the state where the sender could exchange LDP messages with its
neighbors.
Setting the FT Reconnect Timeout to 0 indicates that the sender of
the TLV will not preserve its forwarding state across the restart,
draft-ietf-mpls-ldp-restart-05.txt [Page 4]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
yet the sender supports the procedures defined in Section "Restart of
LDP communication with a neighbor LSR" of this document, and
therefore could take advantage if its neighbor can preserve its
forwarding state across the restart.
For a restarting LSR the Recovery Time carries the time (in
milliseconds) the LSR is willing to retain its MPLS forwarding state
that it preserved across the restart. The time is from the moment the
LSR sends the Initialization message that carries the FT Session TLV
after restart. Setting this time to 0 indicates that the MPLS
forwarding state wasn't preserved across the restart (or even if it
was preserved, is no longer available).
The Recovery Time SHOULD be long enough to allow the neighboring
LSR's to re-sync all the LSP's in a graceful manner, without creating
congestion in the LDP control plane.
3. Operations
An LSR that supports functionality described in this document
advertises this to its LDP neighbors by carrying the FT Session TLV
in the LDP Initialization message.
This document assumes that in certain situations, as specified in
section "Egress LSR", in addition to the MPLS forwarding state, an
LSR can also preserve its IP forwarding state across the restart.
Procedures for preserving IP forwarding state across the restart are
defined in [OSPF-RESTART], [ISIS-RESTART], and [BGP-RESTART].
3.1. Procedures for the restarting LSR
After an LSR restarts its control plane, the LSR MUST check whether
it was able to preserve its MPLS forwarding state from prior to the
restart. If no, then the LSR sets the Recovery Time to 0 in the FT
Session TLV the LSR sends to its neighbors.
If the forwarding state has been preserved, then the LSR starts its
internal timer, called MPLS Forwarding State Holding timer (the value
of that timer SHOULD be configurable), and marks all the MPLS
forwarding state entries as "stale". At the expiration of the timer,
all the entries still marked as stale SHOULD be deleted. The value of
the Recovery Time advertised in the FT Session TLV is set to the
(current) value of the timer at the point when the Initialization
message carrying the FT Session TLV is sent.
We say that an LSR is in the process of restarting when the MPLS
draft-ietf-mpls-ldp-restart-05.txt [Page 5]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
Forwarding State Holding timer is not expired. Once the timer
expires, we say that the LSR completed its restart.
The following procedures apply when an LSR is in the process of
restarting.
3.1.1. Non-egress LSR
If the label carried in the newly received Mapping message is not an
Implicit NULL, the LSR searches its MPLS forwarding state for an
entry with the outgoing label equal to the label carried in the
message, and the next hop equal to one of the addresses (next hops)
received in the Address message from the peer. If such an entry is
found, the LSR no longer marks the entry as stale. In addition, if
the entry is of type <incoming label, (outgoing label, next hop)>
(rather than <FEC, (outgoing label, next hop)>), the LSR associates
the incoming label from that entry with the FEC received in the Label
Mapping message, and advertises (via LDP) <incoming label, FEC> to
its neighbors. If the found entry has no incoming label, or if no
entry is found, the LSR follows the normal LDP procedures. (Note
that this paragraph describes the scenario where the restarting LSR
is neither the egress, nor the penultimate hop that uses penultimate
hop popping for a particular LSP. Note also that this paragraph
covers the case where the restarting LSR is the ingress.)
If the label carried in the Mapping message is an Implicit NULL
label, the LSR searches its MPLS forwarding state for an entry that
indicates Label pop (means no outgoing label), and the next hop equal
to one of the addresses (next hops) received in the Address message
from the peer. If such an entry is found, the LSR no longer marks the
entry as stale, the LSR associates the incoming label from that entry
with the FEC received in the Label Mapping message from the neighbor,
and advertises (via LDP) <incoming label, FEC> to its neighbors. If
the found entry has no incoming label, or if no entry is found, the
LSR follows the normal LDP procedures. (Note that this paragraph
describes the scenario where the restarting LSR is a penultimate hop
for a particular LSP, and this LSP uses penultimate hop popping.)
The description in the above paragraph assumes that the restarting
LSR generates the same label for all the LSPs that terminate on the
same LSR (different from the restarting LSR), and for which the
restarting LSR is a penultimate hop. If this is not the case, and the
restarting LSR generates a unique label per each such LSP, then the
LSR needs to preserve across the restart not just <incoming label,
(outgoing label, next hop)> mapping, but also the FEC associated with
this mapping. In such case the LSR searches its MPLS forwarding state
for an entry that (a) indicates Label pop (means no outgoing label),
draft-ietf-mpls-ldp-restart-05.txt [Page 6]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
(b) the next hop equal to one of the addresses (next hops) received
in the Address message from the peer, and (c) has the same FEC as the
one received in the Label Mapping message. If such an entry is found,
the LSR no longer marks the entry as stale, the LSR associates the
incoming label from that entry with the FEC received in the Label
Mapping message from the neighbor, and advertises (via LDP) <incoming
label, FEC> to its neighbors. If the found entry has no incoming
label, or if no entry is found, the LSR follows the normal LDP
procedures.
3.1.2. Egress LSR
If an LSR determines that it is an egress for a particular FEC, the
LSR is configured to generate a non-NULL label for that FEC, and the
LSR is configured to generate the same (non-NULL) label for all the
FECs that share the same next hop and for which the LSR is an egress,
the LSR searches its MPLS forwarding state for an entry that
indicates Label pop (means no outgoing label), and the next hop equal
to the next hop for that FEC. (Determining the next hop for the FEC
depends on the type of the FEC. For example, when the FEC is an IP
address prefix, the next hop for that FEC is determined from the IP
forwarding table.) If such an entry is found, the LSR no longer marks
this entry as stale, the LSR associates the incoming label from that
entry with the FEC, and advertises (via LDP) <incoming label, FEC> to
its neighbors. If the found entry has no incoming label, or if no
entry is found, the LSR follows the normal LDP procedures.
If an LSR determines that it is an egress for a particular FEC, the
LSR is configured to generate a non-NULL label for that FEC, and the
LSR is configured to generate a unique label for each such FEC, then
the LSR needs to preserve across the restart not just <incoming
label, (outgoing label, next hop)> mapping, but also the FEC
associated with this mapping. In such case the LSR would search its
MPLS forwarding state for an entry that indicates Label pop (means no
outgoing label), and the next hop equal to the next hop for that FEC
associated with the entry (Determining the next hop for the FEC
depends on the type of the FEC. For example, when the FEC is an IP
address prefix, the next hop for that FEC is determined from the IP
forwarding table.) If such an entry is found, the LSR no longer marks
this entry as stale, the LSR associates the incoming label from that
entry with the FEC, and advertises (via LDP) <incoming label, FEC> to
its neighbors. If the found entry has no incoming label, or if no
entry is found, the LSR follows the normal LDP procedures.
If an LSR determines that it is an egress for a particular FEC, and
the LSR is configured to generate a NULL (either Explicit or
Implicit) label for that FEC, the LSR just advertises (via LDP) such
draft-ietf-mpls-ldp-restart-05.txt [Page 7]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
label (together with the FEC) to its neighbors.
3.2. Alternative procedures for the restarting LSR
In this section we describe an alternative to the procedures
described in Section "Procedures for the restarting LSR".
The procedures described in this section assumes that the restarting
LSR has (at least) as many unallocated as allocated labels. The
latter form the MPLS forwarding state that the LSR managed to
preserve across the restart.
After an LSR restarts its control plane, the LSR MUST check whether
it was able to preserve its MPLS forwarding state from prior to the
restart. If no, then the LSR sets the Recovery Time to 0 in the FT
Session TLV the LSR sends to its neighbors.
If the forwarding state has been preserved, then the LSR starts its
internal timer, called MPLS Forwarding State Holding timer (the value
of that timer SHOULD be configurable), and marks all the MPLS
forwarding state entries as "stale". At the expiration of the timer,
all the entries still marked as stale SHOULD be deleted. The value of
the Recovery Time advertised in the FT Session TLV is set to the
(current) value of the timer at the point when the Initialization
message carrying the FT Session TLV is sent.
We say that an LSR is in the process of restarting when the MPLS
Forwarding State Holding timer is not expired. Once the timer
expires, we say that the LSR completed its restart.
While an LSR is in the process of restarting, the LSR creates local
label binding by following the normal LDP procedures.
Note that while an LSR is in the process of restarting, the LSR may
have not one, but two local label bindings for a given FEC - one that
was retained from prior to restart, and another that was created
after the restart. Once the LSR completes its restart, the former
will be deleted. Both of these bindings though would have the same
outgoing label (and the same next hop).
3.3. Restart of LDP communication with a neighbor LSR
When an LSR detects that its LDP session with a neighbor went down,
and the LSR knows that the neighbor is capable of preserving its MPLS
forwarding state across the restart (as was indicated by the FT
Session TLV in the Initialization message received from the
draft-ietf-mpls-ldp-restart-05.txt [Page 8]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
neighbor), the LSR retains the label-FEC bindings received via that
session (rather than discarding the bindings), but marks them as
"stale".
After detecting that the LDP session with the neighbor went down, the
LSR tries to re-establish LDP communication with the neighbor
following the usual LDP procedures.
The amount of time the LSR keeps its stale label-FEC bindings is set
to the lesser of the FT Reconnect Timeout, as was advertised by the
neighbor, and a local timer, called the Neighbor Liveness Timer. If
within that time the LSR still doesn't establish an LDP session with
the neighbor, all the stale bindings SHOULD be deleted. The Neighbor
Liveness Timer is started when the LSR detects that its LDP session
with the neighbor went down. The value of the Neighbor Liveness timer
SHOULD be configurable.
If the LSR re-establishes an LDP session with the neighbor within the
lesser of the FT Reconnect Timeout and the Neighbor Liveness Timer,
and the LSR determines that the neighbor was not able to preserve its
MPLS forwarding state, the LSR SHOULD immediately delete all the
stale label-FEC bindings received from that neighbor. If the LSR
determines that the neighbor was able to preserve its MPLS forwarding
state (as was indicated by the non-zero Recovery Time advertised by
the neighbor), the LSR SHOULD further keep the stale label-FEC
bindings received from the neighbor for as long as the lesser of the
Recovery Time, advertised by the neighbor, and a local configurable
value, called Maximum Recovery Time.
The LSR SHOULD try to complete the exchange of its label mapping
information with the neighbor within 1/2 of the Recovery Time, as
specified in the FT Session TLV received from the neighbor.
The LSR handles the Label Mapping messages received from the neighbor
by following the normal LDP procedures, except that (a) it treats the
stale entries in its Label Information Base (LIB), as if these
entries have been received over the (newly established) session, (b)
if the label-FEC binding carried in the message is the same as the
one that is present in the LIB, but is marked as stale, the LIB entry
no longer is marked as stale, and (c) if for the FEC in the label-FEC
binding carried in the message there is already a label-FEC binding
in the LIB that is marked as stale, and the label in the LIB binding
is different from the label carried in the message, the LSR just
updates the LIB entry with the new label.
An LSR, once it creates a <label, FEC> binding, SHOULD keep the value
of the label in this binding for as long as the LSR has a route to
the FEC in the binding. If the route to the FEC disappears, and then
draft-ietf-mpls-ldp-restart-05.txt [Page 9]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
re-appears again later, then this may result in using a different
label value, as when the route re-appears, the LSR would create a new
<label, FEC> binding.
To minimize the potential mis-routing caused by the label change,
when creating a new <label, FEC> binding the LSR SHOULD pick up the
least recently used label. Once an LSR releases a label, the LSR
SHOULD NOT re-use this label for advertising a <label, FEC> binding
to a neighbor that supports graceful restart for at least the sum of
FT Reconnect Timeout plus Recovery Time, as advertised by the
neighbor to the LSR.
4. Security Consideration
The security considerations pertaining to the original LDP protocol
[RFC3036] remain relevant.
In addition, the mechanism described here renders LSRs that implement
it to additional denial-of-service attacks as follows:
An intruder may impersonate an LDP peer in order to force a
failure and reconnection of the TCP connection, but where the
intruder sets the Recovery Time to 0 on reconnection. This forces
all labels received from the peer to be released.
An intruder could intercept the traffic between LDP peers and
override the setting of the Recovery Time to be set to 0. This
forces all labels received from the peer to be released.
All of these attacks may be countered by use of an authentication
scheme between LDP peers, such as the MD5- based scheme outlined in
[LDP].
As with LDP, a security issue may exist if an LDP implementation
continues to use labels after expiration of the session that first
caused them to be used. This may arise if the upstream LSR detects
the session failure after the downstream LSR has released and re-used
the label. The problem is most obvious with the platform-wide label
space and could result in mis-routing of data to other than intended
destinations and it is conceivable that these behaviors may be
deliberately exploited to either obtain services without
authorization or to deny services to others.
In this document, the validity of the session may be extended by the
Reconnect Timeout, and the session may be re-established in this
period. After the expiry of the Reconnection Timeout the session
must be considered to have failed and the same security issue applies
draft-ietf-mpls-ldp-restart-05.txt [Page 10]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
as described above.
However, the downstream LSR may declare the session as failed before
the expiration of its Reconnection Timeout. This increases the
period during which the downstream LSR might reallocate the label
while the upstream LSR continues to transmit data using the old usage
of the label. To reduce this issue, this document requires that
labels are not re-used until for at least the sum of Reconnect
Timeout plus Recovery Time.
5. Intellectual Property Considerations
Juniper Networks, Inc. is seeking patent protection on some or all of
the technology described in this Internet-Draft. If technology in
this document is adopted as a standard, Juniper Networks agrees to
license, on reasonable and non-discriminatory terms, any patent
rights it obtains covering such technology to the extent necessary to
comply with the standard.
Redback Networks, Inc. is seeking patent protection on some of the
technology described in this Internet-Draft. If technology in this
document is adopted as a standard, Redback Networks agrees to
license, on reasonable and non-discriminatory terms, any patent
rights it obtains covering such technology to the extent necessary to
comply with the standard.
6. Acknowledgments
We would like to thank Loa Andersson, Chaitanya Kodeboyina, Ina
Minei, Nischal Sheth, Enke Chen, and Adrian Farrel for their
contributions to this document.
7. Normative References
[LDP] "Label Distribution Protocol", RFC3036
[FT-LDP] "Fault Tolerance for LDP and CR-LDP", work in progress
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
draft-ietf-mpls-ldp-restart-05.txt [Page 11]
Internet Draft draft-ietf-mpls-ldp-restart-05.txt September 2002
8. Non-normative References
[OSPF-RESTART] "Hitless OSPF Restart", draft-ietf-ospf-hitless-
restart-01.txt
[ISIS-RESTART] "Restart signaling for ISIS", draft-ietf-isis-
restart-01.txt
[BGP-RESTART] "Graceful Restart Mechanism for BGP", draft-ietf-idr-
restart-03.txt
9. Author Information
Manoj Leelanivas
Juniper Networks
1194 N.Mathilda Ave
Sunnyvale, CA 94089
e-mail: manoj@juniper.net
Yakov Rekhter
Juniper Networks
1194 N.Mathilda Ave
Sunnyvale, CA 94089
e-mail: yakov@juniper.net
Rahul Aggarwal
Redback Networks
350 Holger Way
San Jose, CA 95134
e-mail: rahul@redback.com
draft-ietf-mpls-ldp-restart-05.txt [Page 12]
