RFC 7048

Internet Engineering Task Force (IETF)                       E. Nordmark
Request for Comments: 7048                               Arista Networks
Updates: 4861                                               I. Gashinsky
Category: Standards Track                                         Yahoo!
ISSN: 2070-1721                                             January 2014

           Neighbor Unreachability Detection Is Too Impatient


   IPv6 Neighbor Discovery includes Neighbor Unreachability Detection.
   That function is very useful when a host has an alternative neighbor
   -- for instance, when there are multiple default routers -- since it
   allows the host to switch to the alternative neighbor in a short
   time.  By default, this time is 3 seconds after the node starts
   probing.  However, if there are no alternative neighbors, this
   timeout behavior is far too impatient.  This document specifies
   relaxed rules for Neighbor Discovery retransmissions that allow an
   implementation to choose different timeout behavior based on whether
   or not there are alternative neighbors.  This document updates RFC

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

Nordmark & Gashinsky         Standards Track                    [Page 1]

RFC 7048                  NUD Is Too Impatient              January 2014

Copyright Notice

   Copyright (c) 2014 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 ....................................................2
   2. Definition of Terms .............................................3
   3. Protocol Updates ................................................3
   4. Example Algorithm ...............................................6
   5. Acknowledgements ................................................7
   6. Security Considerations .........................................8
   7. References ......................................................8
      7.1. Normative References .......................................8
      7.2. Informative References .....................................8

1.  Introduction

   IPv6 Neighbor Discovery [RFC4861] includes Neighbor Unreachability
   Detection (NUD), which detects when a neighbor is no longer
   reachable.  The timeouts specified for NUD are very short (by
   default, three transmissions spaced one second apart).  These short
   timeouts can be appropriate when there are alternative neighbors to
   which the packets can be sent -- for example, if a host has multiple
   default routers in its Default Router List or if the host has a
   Neighbor Cache Entry (NCE) created by a Redirect message.  In those
   cases, when NUD fails, the host will try the alternative neighbor by
   redoing the next-hop selection.  That implies picking the next router
   in the Default Router List or discarding the NCE created by a
   Redirect message, respectively.

   The timeouts specified in [RFC4861] were chosen to be short in order
   to optimize scenarios where alternative neighbors are available.

   However, when there is no alternative neighbor, there are several
   benefits to making NUD probe for a longer time.  One benefit is to
   make NUD more robust against transient failures, such as spanning

Nordmark & Gashinsky         Standards Track                    [Page 2]

RFC 7048                  NUD Is Too Impatient              January 2014

   tree reconvergence and other layer 2 issues that can take many
   seconds to resolve.  Marking the NCE as unreachable, in that case,
   causes additional multicast on the network.  Assuming there are IP
   packets to send, the lack of an NCE will result in multicast Neighbor
   Solicitations being sent (to the solicited-node multicast address)
   every second instead of the unicast Neighbor Solicitations that NUD

   As a result, IPv6 Neighbor Discovery is operationally more brittle
   than the IPv4 Address Resolution Protocol (ARP).  For IPv4, there is
   no mandatory time limit on the retransmission behavior for ARP
   [RFC0826], which allows implementors to pick more robust schemes.

   The following constant values in [RFC4861] seem to have been made
   part of IPv6 conformance testing: MAX_MULTICAST_SOLICIT,
   MAX_UNICAST_SOLICIT, and RETRANS_TIMER.  While such strict
   conformance testing seems consistent with [RFC4861], it means that
   the standard needs to be updated to allow IPv6 Neighbor Discovery to
   be as robust as ARP.

   This document updates RFC 4861 to relax the retransmission rules.

   Additional motivations for making IPv6 Neighbor Discovery more robust
   in the face of degenerate conditions are covered in [RFC6583].

2.  Definition of Terms

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in [RFC2119].

3.  Protocol Updates

   Discarding the NCE after three packets spaced one second apart is
   only needed when an alternative neighbor is available, such as an
   additional default router or discarding an NCE created by a Redirect.

   If an implementation transmits more than MAX_UNICAST_SOLICIT/
   MAX_MULTICAST_SOLICIT packets, then it SHOULD use the exponential
   backoff of the retransmit timer.  This is to avoid any significant
   load due to a steady background level of retransmissions from
   implementations that retransmit a large number of Neighbor
   Solicitations (NS) before discarding the NCE.

   Even if there is no alternative neighbor, the protocol needs to be
   able to handle the case when the link-layer address of the neighbor/
   target has changed by switching to multicast Neighbor Solicitations
   at some point in time.

Nordmark & Gashinsky         Standards Track                    [Page 3]

RFC 7048                  NUD Is Too Impatient              January 2014

   In order to capture all the cases above, this document introduces a
   new UNREACHABLE state in the conceptual model described in [RFC4861].
   An NCE in the UNREACHABLE state retains the link-layer address, and
   IPv6 packets continue to be sent to that link-layer address.  But in
   the UNREACHABLE state, the NUD Neighbor Solicitations are multicast
   (to the solicited-node multicast address), using a timeout that
   follows an exponential backoff.

   In the places where [RFC4861] says to discard/delete the NCE after N
   probes (Sections 7.3 and 7.3.3, and Appendix C), this document
   instead specifies a transition to the UNREACHABLE state.

   If the Neighbor Cache Entry was created by a Redirect message, a node
   MAY delete the NCE instead of changing its state to UNREACHABLE.  In
   any case, the node SHOULD NOT use an NCE created by a Redirect to
   send packets if that NCE is in the UNREACHABLE state.  Packets should
   be sent following the next-hop selection algorithm in [RFC4861],
   Section 5.2, which disregards NCEs that are not reachable.

   Section 6.3.6 of [RFC4861] indicates that default routers that are
   "known to be reachable" are preferred.  For the purposes of that
   section, if the NCE for the router is in the UNREACHABLE state, it is
   not known to be reachable.  Thus, the particular text in
   Section 6.3.6 that says "in any state other than INCOMPLETE" needs to
   be extended to say "in any state other than INCOMPLETE or

   Apart from the use of multicast NS instead of unicast NS, and the
   exponential backoff of the timer, the UNREACHABLE state works the
   same as the current PROBE state.

   A node MAY garbage collect a Neighbor Cache Entry at any time as
   specified in [RFC4861].  This freedom to garbage collect does not
   change with the introduction of the UNREACHABLE state in the
   conceptual model.  An implementation MAY prefer garbage collecting
   UNREACHABLE NCEs over other NCEs.

   There is a non-obvious extension to the state-machine description in
   Appendix C of [RFC4861] in the case for "NA, Solicited=1, Override=0.
   Different link-layer address than cached".  There we need to add
   "UNREACHABLE" to the current list of "STALE, PROBE, Or DELAY".  That
   is, the NCE would be unchanged.  Note that there is no corresponding
   change necessary to the text in [RFC4861], Section 7.2.5, since it is
   phrased using "Otherwise" instead of explicitly listing the three

Nordmark & Gashinsky         Standards Track                    [Page 4]

RFC 7048                  NUD Is Too Impatient              January 2014

   The other state transitions described in Appendix C handle the
   introduction of the UNREACHABLE state without any change, since they
   are described using "not INCOMPLETE".

   There is also the more obvious change already described above.
   [RFC4861] has this:

   State           Event                   Action             New state

   PROBE           Retransmit timeout,     Discard entry         -
                   N or more

   That needs to be replaced by:

   State           Event                   Action             New state

   PROBE           Retransmit timeout,     Increase timeout  UNREACHABLE
                   N retransmissions.      Send multicast NS

   UNREACHABLE     Retransmit timeout      Increase timeout  UNREACHABLE
                                           Send multicast NS

   The exponential backoff SHOULD be clamped at some reasonable maximum
   retransmit timeout, such as 60 seconds (see MAX_RETRANS_TIMER below).
   If there is no IPv6 packet sent using the UNREACHABLE NCE, then it is
   RECOMMENDED to stop the retransmits of the multicast NS until either
   the NCE is garbage collected or there are IPv6 packets sent using the
   NCE.  The multicast NS and associated exponential backoff can be
   applied on the condition of continued use of the NCE to send IPv6
   packets to the recorded link-layer address.

   A node can unicast the first few Neighbor Solicitation messages even
   while in the UNREACHABLE state, but it MUST switch to multicast
   Neighbor Solicitations within 60 seconds of the initial
   retransmission to be able to handle a link-layer address change for
   the target.  The example below shows such behavior.

Nordmark & Gashinsky         Standards Track                    [Page 5]

RFC 7048                  NUD Is Too Impatient              January 2014

4.  Example Algorithm

   This section is NOT normative but specifies a simple implementation
   that conforms with this document.  The implementation is described
   using operator-configurable values that allow it to be configured to
   be compatible with the retransmission behavior in [RFC4861].  The
   operator can configure the values for MAX_UNICAST_SOLICIT,
   implementation to be as simple as:

   next_retrans = ($BACKOFF_MULTIPLE ^ $solicit_retrans_num) *
   $RetransTimer * $JitterFactor where solicit_retrans_num is zero for
   the first transmission, and JitterFactor is a random value between
   synchronization of transmissions from different hosts.

   After MARK_UNREACHABLE transmissions, the implementation would mark
   the NCE UNREACHABLE and as a result explore alternate next hops.
   After MAX_UNICAST_SOLICIT, the implementation would switch to
   multicast NUD probes.

   The behavior of this example algorithm is to have 5 attempts, with
   time spacing of 0 (initial request), 1 second later, 3 seconds after
   the first retransmission, then 9, then 27, and switch to UNREACHABLE
   after the first three transmissions.  Thus, relative to the time of
   the first transmissions, the retransmissions would occur at 1 second,
   4 seconds, 13 seconds, and finally 40 seconds.  At 4 seconds from the
   first transmission, the NCE would be marked UNREACHABLE.  That
   behavior corresponds to:


      RETRANS_TIMER=1 (default)




   After 3 retransmissions, the implementation would mark the NCE
   UNREACHABLE.  That results in trying an alternative neighbor, such as
   another default router, or ignoring an NCE created by a Redirect as
   specified in [RFC4861].  With the above values, that would occur
   after 4 seconds following the first transmission compared to the

Nordmark & Gashinsky         Standards Track                    [Page 6]

RFC 7048                  NUD Is Too Impatient              January 2014

   2 seconds using the fixed scheme in [RFC4861].  That additional
   delay is small compared to the default ReachableTime of
   30,000 milliseconds.

   After 5 transmissions, i.e., 40 seconds after the initial
   transmission, the example behavior is to switch to multicast NUD
   probes.  In the language of the state machine in [RFC4861], that
   corresponds to the action "Discard entry".  Thus, any attempts to
   send future packets would result in sending multicast NS packets.  An
   implementation MAY retain the backoff value as it switches to
   multicast NUD probes.  The potential downside of deferring switching
   to multicast is that it would take longer for NUD to handle a change
   in a link-layer address, i.e., the case when a host or a router
   changes its link-layer address while keeping the same IPv6 address.
   However, [RFC4861] says that a node MAY send unsolicited NS to handle
   that case, which is rather infrequent in operational networks.  In
   any case, the implementation needs to follow the "SHOULD" in
   Section 3 to switch to multicast solutions within 60 seconds after
   the initial transmission.

   you would get the same behavior as in [RFC4861].

   If the request was not answered at first -- due, for example, to a
   transitory condition -- an implementation following this algorithm
   would retry immediately and then back off for progressively longer
   periods.  This would allow for a reasonably fast resolution time when
   the transitory condition clears.

   Note that RetransTimer and ReachableTime are by default set from the
   protocol constants RETRANS_TIMER and REACHABLE_TIME but are
   overridden by values advertised in Router Advertisements as specified
   in [RFC4861].  That remains the case even with the protocol updates
   specified in this document.  The key values that the operator would

   It is useful to have a maximum value for
   ($BACKOFF_MULTIPLE^$solicit_attempt_num)*$RetransTimer so that the
   retransmissions are not too far apart.  The above value of 60 seconds
   for this MAX_RETRANS_TIMER is consistent with DHCPv6.

5.  Acknowledgements

   The comments from Thomas Narten, Philip Homburg, Joel Jaeggli, Hemant
   Singh, Tina Tsou, Suresh Krishnan, and Murray Kucherawy have helped
   improve this document.

Nordmark & Gashinsky         Standards Track                    [Page 7]

RFC 7048                  NUD Is Too Impatient              January 2014

6.  Security Considerations

   Relaxing the retransmission behavior for NUD is believed to have no
   impact on security.  In particular, it doesn't impact the application
   of Secure Neighbor Discovery [RFC3971].

7.  References

7.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3971]  Arkko, J., Kempf, J., Zill, B., and P. Nikander, "SEcure
              Neighbor Discovery (SEND)", RFC 3971, March 2005.

   [RFC4861]  Narten, T., Nordmark, E., Simpson, W., and H. Soliman,
              "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861,
              September 2007.

7.2.  Informative References

   [RFC0826]  Plummer, D., "Ethernet Address Resolution Protocol: Or
              converting network protocol addresses to 48.bit Ethernet
              address for transmission on Ethernet hardware", STD 37,
              RFC 826, November 1982.

   [RFC6583]  Gashinsky, I., Jaeggli, J., and W. Kumari, "Operational
              Neighbor Discovery Problems", RFC 6583, March 2012.

Authors' Addresses

   Erik Nordmark
   Arista Networks
   Santa Clara, CA

   EMail: nordmark@acm.org

   Igor Gashinsky
   45 W 18th St
   New York, NY

   EMail: igor@yahoo-inc.com

Nordmark & Gashinsky         Standards Track                    [Page 8]