OSPF Restart Signaling
A flag in the OSPF Hello packet that tells neighbours to hold on to an adjacency while a router is restarting.
Nguyen, L., Roy, A., and A. Zinin, "OSPF Restart Signaling", RFC 4812, DOI 10.17487/RFC4812, March 2007, <https://www.rfc-editor.org/info/rfc4812>.
@misc{rfc4812,
series = {Request for Comments},
number = 4812,
howpublished = {RFC 4812},
publisher = {RFC Editor},
doi = {10.17487/RFC4812},
url = {https://www.rfc-editor.org/info/rfc4812},
author = {L. Nguyen and A. Roy and A. Zinin},
title = {{OSPF Restart Signaling}},
pagetotal = 7,
year = 2007,
month = mar,
} RFC text
Contents
Network Working Group L. Nguyen
Request for Comments: 4812 A. Roy
Category: Informational Cisco Systems
A. Zinin
Alcatel-Lucent
March 2007
OSPF Restart Signaling
Status of This Memo
This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.
Copyright Notice
Copyright (C) The IETF Trust (2007).
Abstract
OSPF is a link-state intra-domain routing protocol used in IP
networks. Routers find new and detect unreachable neighbors via the
Hello subprotocol. Hello OSPF packets are also used to ensure two-
way connectivity within time. When a router restarts its OSPF
software, it may not know its neighbors. If such a router sends a
Hello packet on an interface, its neighbors are going to reset the
adjacency, which may not be desirable in certain conditions.
This memo describes a vendor-specific mechanism that allows OSPF
routers to inform their neighbors about the restart process. Note
that this mechanism requires support from neighboring routers. The
mechanism described in this document was proposed before Graceful
OSPF Restart, as described in RFC 3623, came into existence. It is
implemented/supported by at least one major vendor and is currently
deployed in the field. The purpose of this document is to capture
the details of this mechanism for public use. This mechanism is not
an IETF standard.
Table of Contents
1. Introduction ....................................................2
2. Proposed Solution ...............................................2
2.1. Sending Hello Packets with the RS-bit Set ..................3
2.2. Receiving Hello Packets with the RS-Bit Set ................3
2.3. Ensuring Topology Stability ................................4
3. Backward Compatibility ..........................................4
4. Security Considerations .........................................4
5. IANA Considerations .............................................4
6. References ......................................................5
6.1. Normative References .......................................5
6.2. Informative References .....................................5
Appendix A. Acknowledgements ......................................6
1. Introduction
While performing a graceful restart of OSPF software [RFC3623],
routers need to prevent their neighbors from resetting their
adjacencies. However, after a reload, routers may not be aware of
the neighbors they had adjacencies with in their previous
incarnations. If such a router sends a Hello packet on an interface
and this packet does not list some neighbors, those neighbors will
reset the adjacency with the restarting router.
This document describes a technique that allows restarting routers to
inform their neighbors that they may not know about some neighbors
yet and the absence of some router IDs in the Hello packets should be
ignored.
2. Proposed Solution
With this Restart Signaling Solution, a new bit, called RS (restart
signal), is introduced into the Extended Options (EO) TLV in the
Link-Local Signaling (LLS) block (see [RFC4813]). The value of this
bit is 0x00000002; see Figure 1 below.
+---+---+---+---+---+---+---+- -+---+---+---+---+---+---+---+---+
| * | * | * | * | * | * | * |...| * | * | * | * | * | * | RS| LR|
+---+---+---+---+---+---+---+- -+---+---+---+---+---+---+---+---+
Figure 1. Bits in Extended Options TLV
For a definition of the LR-bit, see [RFC4811].
2.1. Sending Hello Packets with the RS-bit Set
OSPF routers should set the RS-bit in the EO-TLV attached to a Hello
packet when it is not known that all neighbors are listed in this
packet, but the restarting router wants them to preserve their
adjacencies. The RS-bit must not be set in Hello packets longer than
RouterDeadInterval seconds.
2.2. Receiving Hello Packets with the RS-Bit Set
When an OSPF router receives a Hello packet containing the LLS block
with the EO-TLV that has the RS-bit set, the router should skip the
two-way connectivity check with the announcing neighbor (i.e., the
router should not generate a 1-WayReceived event for the neighbor if
it does not find its own router ID in the list of neighbors as
described in Section 10.5 of [RFC2328]), provided that the neighbor
Finite State Machine (FSM) for this neighbor is in the Full state.
The router should also send a unicast Hello back to the sender in
reply to a Hello packet with RS-bit set. This is to speed up
learning of previously known neighbors. When sending such a reply
packet, care must be taken to ensure that the RS-bit is clear in it.
Two additional fields are introduced in the neighbor data structure:
RestartState flag and ResyncTimeout timer. RestartState flag
indicates that a Hello packet with the RS-bit set has been received
and the local router expects its neighbor to go through the Link
State Database (LSDB) resynchronization procedure using [RFC4811].
ResyncTimeout is a single-shot timer limiting the delay between the
first seen Hello packet with the RS-bit set and initialization of the
LSDB resynchronization procedure. The length of ResyncTimeout timer
is RouterDeadInterval seconds.
When a Hello packet with the RS-bit set is received and RestartState
flag is not set for the neighbor, the router sets RestartState flag
and starts ResyncTimeout timer. If ResyncTimeout expires,
RestartState flag is cleared and a 1-WayReceived event is generated
for the neighbor. If, while ResyncTimeout timer is running, the
neighbor starts LSDB resynchronization procedure using [RFC4811],
ResyncTimeout timer is canceled. The router also clears RestartState
flag on completion of the LSDB resynchronization process.
Two or more routers on the same segment cannot have Hello packets
with the RS-bit set at the same time, as can be the case when two or
more routers restart at about the same time. In such a scenario, the
routers should clear the RestartState flag, cancel the ResyncTimeout
timer, and generate a 1-WayReceived event.
2.3. Ensuring Topology Stability
Under certain circumstances, it might be desirable to stop announcing
the restarting router as fully adjacent if this may lead to possible
routing loops. In order to provide this functionality, a
configurable option is provided on the neighboring routers that
instructs the OSPF process to follow the logics described below.
When an OSPF router schedules a routing table calculation due to a
change in the contents of its LSDB, it should also reset all
adjacencies with restarting routers (those with RestartState set to
TRUE) by clearing the RestartState neighbor flags, canceling
ResyncTimeout timers (if running), and generating the 1-WayReceived
events for the neighbor FSMs.
3. Backward Compatibility
The described technique requires cooperation from neighboring
routers. However, if neighbors do not support this technique, they
will just reset the adjacency.
4. Security Considerations
The described technique does not introduce any new security issues
into the OSPF protocol.
5. IANA Considerations
Please refer to the "IANA Considerations" section of [RFC4813] for
more information on the Extended Options bit definitions.
6. References
6.1. Normative References
[RFC2328] Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998.
[RFC3623] Moy, J., Pillay-Esnault, P., and A. Lindem, "Graceful OSPF
Restart", RFC 3623, November 2003.
6.2. Informative References
[RFC4813] Friedman, B., Nguyen, L., Roy, A., Yeung, D., and A.
Zinin, "OSPF Link-Local Signaling", RFC 4813, March 2007.
[RFC4811] Nguyen, L., Roy, A., and A. Zinin, "OSPF Out-of-Band Link
State Database (LSDB) Resynchronization", RFC 4811, March
2007.
Appendix A. Acknowledgments
The authors would like to thank John Moy, Russ White, Don Slice, and
Alvaro Retana for their valuable comments.
Authors' Addresses
Liem Nguyen
Cisco Systems
225 West Tasman Drive
San Jose, CA 95134
USA
EMail: lhnguyen@cisco.com
Abhay Roy
Cisco Systems
225 West Tasman Drive
San Jose, CA 95134
USA
EMail: akr@cisco.com
Alex Zinin
Alcatel-Lucent
Mountain View, CA
USA
EMail: alex.zinin@alcatel-lucent.com
Full Copyright Statement
Copyright (C) The IETF Trust (2007).
This document is subject to the rights, licenses and restrictions
contained in BCP 78, and except as set forth therein, the authors
retain all their rights.
This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND
THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS
OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF
THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Intellectual Property
The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at
ietf-ipr@ietf.org.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
ISO Plain Language
On this page
§1The Problem With a Router That Forgets Its Neighbours
OSPF is a link-state intra-domain routing protocol used in IP networks. Routers find new neighbours and detect unreachable ones with the Hello subprotocol, and Hello packets are also used to confirm two-way connectivity within a period of time.
While performing a graceful restart of its OSPF software, a router needs to stop its neighbours from resetting their adjacencies. After a reload, though, a router may not know which neighbours it had adjacencies with in its previous incarnations. If such a router sends a Hello packet on an interface and that packet does not list some neighbours, those neighbours will reset the adjacency with the restarting router.
The technique in this document lets a restarting router tell its neighbours that it may not know about some of them yet, and that the absence of some router IDs in the Hello packets should be ignored.
The abstract is explicit about the status of this mechanism. It describes a vendor-specific mechanism, notes that it requires support from the neighbouring routers, records that it was proposed before Graceful OSPF Restart came into existence, and says it is implemented and supported by at least one major vendor and is deployed in the field. Its purpose is to capture the details of the mechanism for public use, and it is not an IETF standard.
OSPF neighbours check each other's Hello packets to confirm the link is working both ways. A router that has just rebooted cannot list the neighbours it had before, so its neighbours think it has gone and drop the link.
- Hello packetsOSPF routers find new neighbours and detect unreachable ones with the Hello subprotocol, and Hello packets confirm two-way connectivity within a period of time.
- The restart problemAfter a reload a router may not know the neighbours it had adjacencies with in its previous incarnations.
- The consequenceA Hello packet that does not list some neighbours makes those neighbours reset the adjacency with the restarting router.
- The signalA bit in the Hello packet tells the neighbours that the router is restarting and that missing router IDs should be ignored.
§2The RS Bit in the Extended Options TLV
With the restart signalling solution, a new bit called RS, for restart signal, is introduced into the Extended Options TLV in the Link-Local Signaling block defined in RFC 4813. The value of this bit is 0x00000002. It sits next to the LR bit, which is defined in RFC 4811, and the remaining positions of the TLV are shown as unused in the figure the RFC gives.
A router should set the RS bit in the Extended Options TLV attached to a Hello packet when it is not known that all neighbours are listed in that packet, but the restarting router wants them to preserve their adjacencies. The RS bit must not be set in Hello packets longer than RouterDeadInterval seconds, so that the signal cannot be used to hold an adjacency open indefinitely.
The bit is carried inside the existing Link-Local Signaling block, which is the mechanism RFC 4813 defines for carrying information about the local link in OSPF packets. The IANA considerations of this document point to the IANA considerations section of RFC 4813 for the definitions of the Extended Options bits.
The signal is a single bit carried in a block that OSPF already has, next to another bit of the same kind. It tells the neighbour that the list of neighbours in this Hello packet may be incomplete.
- 0x00000002RS, restart signal — The bit introduced by this document in the Extended Options TLV of the Link-Local Signaling block.
- LRLink-local resynchronization — The bit next to RS, whose definition is in RFC 4811.
- RouterDeadIntervalLimit on the RS bit — The RS bit must not be set in Hello packets longer than this many seconds, and the ResyncTimeout timer lasts this long.
§2.2How a Neighbour Handles a Restart Signal
When a router receives a Hello packet containing the Link-Local Signaling block with the Extended Options TLV that has the RS bit set, it should skip the two-way connectivity check with the announcing neighbour. That means it should not generate a 1-WayReceived event for the neighbour if it does not find its own router ID in the list of neighbours, provided that the neighbour's finite state machine for that neighbour is in the Full state.
The router should also send a unicast Hello back to the sender in reply to a Hello packet with the RS bit set, which speeds up the learning of previously known neighbours. When sending such a reply, care must be taken to ensure that the RS bit is clear in it, so the reply does not itself look like another restart.
Two additional fields are introduced in the neighbour data structure. The RestartState flag indicates that a Hello packet with the RS bit set has been received and that the local router expects its neighbour to go through the link state database resynchronisation procedure defined in RFC 4811. The ResyncTimeout timer is a single-shot timer limiting the delay between the first Hello packet with the RS bit set and the start of that resynchronisation, and its length is RouterDeadInterval seconds.
When a Hello packet with the RS bit set is received and the RestartState flag is not already set for that neighbour, the router sets the flag and starts the ResyncTimeout timer. If the timer expires, the flag is cleared and a 1-WayReceived event is generated for the neighbour, which resets the adjacency in the ordinary way. If the neighbour starts the link state database resynchronisation procedure while the timer is running, the timer is cancelled, and the flag is also cleared when the resynchronisation completes.
Two or more routers on the same segment cannot have Hello packets with the RS bit set at the same time, as can be the case when two or more routers restart at about the same time. In that situation the routers should clear the RestartState flag, cancel the ResyncTimeout timer, and generate a 1-WayReceived event.
The neighbour ignores the missing names in the Hello packet, but only for a limited time. If the restarting router does not catch up before the timer expires, the neighbour drops the adjacency as it normally would.
- Adjacent start
- Restart expected
- Adjacency down
| From | To | Event |
|---|---|---|
| Adjacent | Restart expected | Hello with the RS bit set |
| Restart expected | Adjacent | Resynchronization completes |
| Restart expected | Adjacency down | ResyncTimeout expires |
| Restart expected | Adjacency down | Another router on the segment sets the RS bit |
§2.3Ensuring Topology Stability
Under certain circumstances it may be desirable to stop announcing the restarting router as fully adjacent, if continuing to do so could lead to routing loops. To provide that, a configurable option is available on the neighbouring routers that instructs the OSPF process to follow a particular set of rules.
When a router schedules a routing table calculation because of a change in the contents of its link state database, it should also reset all adjacencies with restarting routers, that is, with the routers whose RestartState flag is set. It does this by clearing the RestartState neighbour flags, cancelling any running ResyncTimeout timers, and generating the 1-WayReceived events for the neighbour state machines.
If the network changes while a router is restarting, its neighbours can be told to stop treating it as fully connected. That avoids a routing loop, at the cost of the restart being abandoned early.
§4Compatibility and Security
The technique requires cooperation from the neighbouring routers. If the neighbours do not support it, they will simply reset the adjacency, which is the behaviour without the mechanism.
The security considerations section is short and says that the described technique does not introduce any new security issues into the OSPF protocol.
The mechanism only helps where the neighbouring routers also understand it. Where they do not, the adjacency is reset as before, and the RFC adds no new security issues of its own.