Internet Engineering Task Force (IETF) S. Morris

Internet Engineering Task Force (IETF) S. Morris
Request for Comments: 7583 ISC
Category: Informational J. Ihren
ISSN: 2070-1721 Netnod
J. Dickinson
Sinodun
W. Mekking
Dyn
October 2015

DNSSEC Key Rollover Timing Considerations

Abstract

This document describes the issues surrounding the timing of events
in the rolling of a key in a DNSSEC-secured zone. It presents
timelines for the key rollover and explicitly identifies the
relationships between the various parameters affecting the process.

Status of This Memo

This document is not an Internet Standards Track specification; it is
published for informational purposes.

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). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see 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/rfc7583.

Morris, et al. Informational [Page 1]

RFC 7583 Key Timing October 2015

Copyright Notice

Copyright (c) 2015 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 ....................................................3
1.1. Key Rolling Considerations .................................3
1.2. Types of Keys ..............................................4
1.3. Terminology ................................................4
1.4. Limitation of Scope ........................................5
2. Rollover Methods ................................................5
2.1. ZSK Rollovers ..............................................5
2.2. KSK Rollovers ..............................................7
3. Key Rollover Timelines ..........................................8
3.1. Key States .................................................8
3.2. ZSK Rollover Timelines ....................................10
3.2.1. Pre-Publication Method .............................10
3.2.2. Double-Signature Method ............................12
3.3. KSK Rollover Timelines ....................................14
3.3.1. Double-KSK Method ..................................14
3.3.2. Double-DS Method ...................................17
3.3.3. Double-RRset Method ................................20
3.3.4. Interaction with Configured Trust Anchors ..........22
3.3.5. Introduction of First Keys .........................24
4. Standby Keys ...................................................24
5. Algorithm Considerations .......................................25
6. Summary ........................................................26
7. Security Considerations ........................................26
8. Normative References ...........................................26
Appendix A. List of Symbols ......................................28
Acknowledgements ..................................................31
Authors' Addresses ................................................31

Morris, et al. Informational [Page 2]

RFC 7583 Key Timing October 2015

1. Introduction

1.1. Key Rolling Considerations

When a zone is secured with DNSSEC, the zone manager must be prepared
to replace ("roll") the keys used in the signing process. The
rolling of keys may be caused by compromise of one or more of the
existing keys, or it may be due to a management policy that demands
periodic key replacement for security or operational reasons. In
order to implement a key rollover, the keys need to be introduced
into and removed from the zone at the appropriate times.
Considerations that must be taken into account are:

o DNSKEY records and associated information (such as the DS records
or RRSIG records created with the key) are not only held at the
authoritative nameserver, they are also cached by resolvers. The
data on these systems can be interlinked, e.g., a validating
resolver may try to validate a signature retrieved from a cache
with a key obtained separately.

o Zone "bootstrapping" events, where a zone is signed for the first
time, can be common in configurations where a large number of
zones are being served. Procedures should be able to cope with
the introduction of keys into the zone for the first time as well
as "steady-state", where the records are being replaced as part of
normal zone maintenance.

o To allow for an emergency re-signing of the zone as soon as
possible after a key compromise has been detected, standby keys
(additional keys over and above those used to sign the zone) need
to be present.

o A query for the DNSKEY RRset returns all DNSKEY records in the
zone. As there is limited space in the UDP packet (even with
EDNS0 support), key records no longer needed must be periodically
removed. (For the same reason, the number of standby keys in the
zone should be restricted to the minimum required to support the
key management policy.)

Management policy, e.g., how long a key is used for, also needs to be
considered. However, the point of key management logic is not to
ensure that a rollover is completed at a certain time but rather to
ensure that no changes are made to the state of keys published in the
zone until it is "safe" to do so ("safe" in this context meaning that
at no time during the rollover process does any part of the zone ever
go bogus). In other words, although key management logic enforces
policy, it may not enforce it strictly.

Morris, et al. Informational [Page 3]

RFC 7583 Key Timing October 2015

A high-level overview of key rollover can be found in [RFC6781]. In
contrast, this document focuses on the low-level timing detail of two
classes of operations described there, the rollover of Zone Signing
Keys (ZSKs), and the rollover of Key Signing Keys (KSKs).

Note that the process for the introduction of keys into a zone is
different from that of rolling a key; see Section 3.3.5 for more
information.

1.2. Types of Keys

Although DNSSEC validation treats all keys equally, [RFC4033]
recognizes the broad classification of ZSKs and KSKs. A ZSK is used
to authenticate information within the zone; a KSK is used to
authenticate the zone's DNSKEY RRset. The main implication for this
distinction concerns the consistency of information during a
rollover.

During operation, a validating resolver must use separate pieces of
information to perform an authentication. At the time of
authentication, each piece of information may be in its cache or may
need to be retrieved from an authoritative server. The rollover
process needs to happen in such a way that the information is
consistent at all times during the rollover. With a ZSK, the
information is the RRSIG (plus associated RRset) and the DNSKEY.
These are both obtained from the same zone. In the case of the KSK,
the information is the DNSKEY and DS RRset with the latter being
obtained from a different zone.

Although there are similarities in the algorithms to roll ZSKs and
KSKs, there are a number of differences. For this reason, the two
types of rollovers are described separately.

1.3. Terminology

The terminology used in this document is as defined in [RFC4033] and
[RFC5011].

A number of symbols are used to identify times, intervals, etc. All
are listed in Appendix A.

Morris, et al. Informational [Page 4]

RFC 7583 Key Timing October 2015

1.4. Limitation of Scope

This document represents current thinking at the time of publication.
However, the subject matter is evolving and it is not possible for
the document to be comprehensive. In particular, it does not cover:

o Rolling a key that is used as both a ZSK and KSK.

o Algorithm rollovers. Only the rolling of keys of the same
algorithm is described here: not transitions between algorithms.

o Changing TTLs.

Algorithm rollover is excluded from the document owing to the need
for there to be an RRSIG for at least one DNSKEY of each algorithm in
the DNSKEY RRset [RFC4035]. This introduces additional constraints
on rollovers that are not considered here. Such considerations do
not apply where other properties of the key, such as key length, are
changed during the rollover: the DNSSEC protocol does not impose any
restrictions in these cases.

Also excluded from consideration is the effect of changing the Time
to Live (TTL) of records in a zone. TTLs can be changed at any time,
but doing so around the time of a key rollover may have an impact on
event timings. In the timelines below, it is assumed that TTLs are
constant.

2. Rollover Methods

2.1. ZSK Rollovers

For ZSKs, the issue for the zone operator/signer is to ensure that
any caching validator that has access to a particular signature also
has access to the corresponding valid ZSK.

A ZSK can be rolled in one of three ways:

o Pre-Publication: described in [RFC6781], the new key is introduced
into the DNSKEY RRset, which is then re-signed. This state of
affairs remains in place for long enough to ensure that any cached
DNSKEY RRsets contain both keys. At that point, signatures
created with the old key can be replaced by those created with the
new key. During the re-signing process (which may or may not be
atomic depending on how the zone is managed), it doesn't matter
with which key an RRSIG record retrieved by a resolver was
created; cached copies of the DNSKEY RRset will contain both the
old and new keys.

Morris, et al. Informational [Page 5]

RFC 7583 Key Timing October 2015

Once the zone contains only signatures created with the new key,
there is an interval during which RRSIG records created with the
old key expire from caches. After this, there will be no
signatures anywhere that were created using the old key, and it
can be removed from the DNSKEY RRset.

o Double-Signature: also mentioned in [RFC6781], this involves
introducing the new key into the zone and using it to create
additional RRSIG records; the old key and existing RRSIG records
are retained. During the period in which the zone is being signed
(again, the signing process may not be atomic), validating
resolvers are always able to validate RRSIGs: any combination of
old and new DNSKEY RRset and RRSIGs allows at least one signature
to be validated.

Once the signing process is complete and enough time has elapsed
to make sure that all validators that have the DNSKEY and
signatures in cache have both the old and new information, the old
key and signatures can be removed from the zone. As before,
during this period any combination of DNSKEY RRset and RRSIGs will
allow validation of at least one signature.

o Double-RRSIG: strictly speaking, the use of the term "Double-
Signature" above is a misnomer as the method is not only double
signature, it is also double key as well. A true Double-Signature
method (here called the Double-RRSIG method) involves introducing
new signatures in the zone (while still retaining the old ones)
but not introducing the new key.

Once the signing process is complete and enough time has elapsed
to ensure that all caches that may contain an RR and associated
RRSIG have a copy of both signatures, the key is changed. After a
further interval during which the old DNSKEY RRset expires from
caches, the old signatures are removed from the zone.

Of the three methods, Double-Signature is conceptually the simplest:
introduce the new key and new signatures, then approximately one TTL
later remove the old key and old signatures. It is also the fastest,
but suffers from increasing the size of the zone and the size of
responses.

Pre-Publication is more complex: introduce the new key, approximately
one TTL later sign the records, and approximately one TTL after that
remove the old key. It does however keep the zone and response sizes
to a minimum.

Morris, et al. Informational [Page 6]

RFC 7583 Key Timing October 2015

Double-RRSIG is essentially the reverse of Pre-Publication: introduce
the new signatures, approximately one TTL later change the key, and
approximately one TTL after that remove the old signatures. However,
it has the disadvantage of the Pre-Publication method in terms of
time taken to perform the rollover, the disadvantage of the Double-
Signature rollover in terms of zone and response sizes, and none of
the advantages of either. For these reasons, it is unlikely to be
used in any real-world situations and so will not be considered
further in this document.

2.2. KSK Rollovers

In the KSK case, there should be no problem with a caching validator
not having access to a signature created with a valid KSK. The KSK
is only used for one signature (that over the DNSKEY RRset) and both
the key and the signature travel together. Instead, the issue is to
ensure that the KSK is trusted.

Trust in the KSK is due to either the existence of a signed and
validated DS record in the parent zone or an explicitly configured
trust anchor. If the former, the rollover algorithm will need to
involve the parent zone in the addition and removal of DS records, so
timings are not wholly under the control of the zone manager. If the
latter, [RFC5011] timings will be needed to roll the keys. (Even in
the case where authentication is via a DS record, the zone manager
may elect to include [RFC5011] timings in the key rolling process so
as to cope with the possibility that the key has also been explicitly
configured as a trust anchor.)

It is important to note that the need to interact with the parent
does not preclude the development of key rollover logic; in
accordance with the goal of the rollover logic, being able to
determine when a state change is "safe", the only effect of being
dependent on the parent is that there may be a period of waiting for
the parent to respond in addition to any delay the key rollover logic
requires. Although this introduces additional delays, even with a
parent that is less than ideally responsive, the only effect will be
a slowdown in the rollover state transitions. This may cause a
policy violation, but will not cause any operational problems.

Like the ZSK case, there are three methods for rolling a KSK:

o Double-KSK: the new KSK is added to the DNSKEY RRset, which is
then signed with both the old and new key. After waiting for the
old RRset to expire from caches, the DS record in the parent zone
is changed. After waiting a further interval for this change to
be reflected in caches, the old key is removed from the RRset.

Morris, et al. Informational [Page 7]

RFC 7583 Key Timing October 2015

o Double-DS: the new DS record is published. After waiting for this
change to propagate into caches, the KSK is changed. After a
further interval during which the old DNSKEY RRset expires from
caches, the old DS record is removed.

o Double-RRset: the new KSK is added to the DNSKEY RRset, which is
then signed with both the old and new key, and the new DS record
is added to the parent zone. After waiting a suitable interval
for the old DS and DNSKEY RRsets to expire from caches, the old
DNSKEY and DS records are removed.

In essence, Double-KSK means that the new KSK is introduced first and
used to sign the DNSKEY RRset. The DS record is changed, and finally
the old KSK is removed. It limits interactions with the parent to a
minimum but, for the duration of the rollover, the size of the DNSKEY
RRset is increased.

With Double-DS, the order of operations is the other way around:
introduce the new DS, change the DNSKEY, then remove the old DS. The
size of the DNSKEY RRset is kept to a minimum, but two interactions
are required with the parent.

Finally, Double-RRset is the fastest way to roll the KSK, but has the
drawbacks of both of the other methods: a larger DNSKEY RRset and two
interactions with the parent.

3. Key Rollover Timelines

3.1. Key States

DNSSEC validation requires both the DNSKEY and information created
from it (referred to as "associated data" in this section). In the
case of validation of an RR, the data associated with the key is the
corresponding RRSIG. Where there is a need to validate a chain of
trust, the associated data is the DS record.

During the rolling process, keys move through different states. The
defined states are:

Generated Although keys may be created immediately prior to first
use, some implementations may find it convenient to
create a pool of keys in one operation and draw from it
as required. (Note: such a pre-generated pool must be
secured against surreptitious use.) In the timelines
below, before the first event, the keys are considered to
be created but not yet used: they are said to be in the
"Generated" state.

Morris, et al. Informational [Page 8]

RFC 7583 Key Timing October 2015

Published A key enters the published state when either it or its
associated data first appears in the appropriate zone.

Ready The DNSKEY or its associated data have been published for
long enough to guarantee that copies of the key(s) it is
replacing (or associated data related to that key) have
expired from caches.

Active The data is starting to be used for validation. In the
case of a ZSK, it means that the key is now being used to
sign RRsets and that both it and the created RRSIGs
appear in the zone. In the case of a KSK, it means that
it is possible to use it to validate a DNSKEY RRset as
both the DNSKEY and DS records are present in their
respective zones. Note that when this state is entered,
it may not be possible for validating resolvers to use
the data for validation in all cases: the zone signing
may not have finished or the data might not have reached
the resolver because of propagation delays and/or caching
issues. If this is the case, the resolver will have to
rely on the predecessor data instead.

Retired The data has ceased to be used for validation. In the
case of a ZSK, it means that the key is no longer used to
sign RRsets. In the case of a KSK, it means that the
successor DNSKEY and DS records are in place. In both
cases, the key (and its associated data) can be removed
as soon as it is safe to do so, i.e., when all validating
resolvers are able to use the new key and associated data
to validate the zone. However, until this happens, the
current key and associated data must remain in their
respective zones.

Dead The key and its associated data are present in their
respective zones, but there is no longer information
anywhere that requires their presence for use in
validation. Hence, they can be removed at any time.

Removed Both the DNSKEY and its associated data have been removed
from their respective zones.

Revoked The DNSKEY is published for a period with the "revoke"
bit set as a way of notifying validating resolvers that
have configured it as a trust anchor, as used in
[RFC5011], that it is about to be removed from the zone.
This state is used when [RFC5011] considerations are in
effect (see Section 3.3.4).

Morris, et al. Informational [Page 9]

RFC 7583 Key Timing October 2015

3.2. ZSK Rollover Timelines

The following sections describe the rolling of a ZSK. They show the
events in the lifetime of a key (referred to as "key N") and cover
its replacement by its successor (key N+1).

3.2.1. Pre-Publication Method

In this method, the new key is introduced into the DNSKEY RRset.
After enough time to ensure that any cached DNSKEY RRsets contain
both keys, the zone is signed using the new key and the old
signatures are removed. Finally, when all signatures created with
the old key have expired from caches, the old key is removed.

The following diagram shows the timeline of a Pre-Publication
rollover. Time increases along the horizontal scale from left to
right and the vertical lines indicate events in the process.
Significant times and time intervals are marked.

|1| |2| |3| |4| |5| |6| |7| |8|
| | | | | | | |
Key N |<-Ipub->|<--->|<-------Lzsk------>|<-Iret->|<--->|
| | | | | | | |
Key N+1 | | | |<-Ipub->|<-->|<---Lzsk---- - -
| | | | | | | |
Key N Tpub Trdy Tact Tret Tdea Trem
Key N+1 Tpub Trdy Tact

---- Time ---->

Figure 1: Timeline for a Pre-Publication ZSK Rollover

Event 1: Key N's DNSKEY record is put into the zone, i.e., it is
added to the DNSKEY RRset, which is then re-signed with the currently
active KSKs. The time at which this occurs is the publication time
(Tpub), and the key is now said to be published. Note that the key
is not yet used to sign records.

Event 2: Before it can be used, the key must be published for long
enough to guarantee that any cached version of the zone's DNSKEY
RRset includes this key.

This interval is the publication interval (Ipub) and, for the second
or subsequent keys in the zone, is given by:

Ipub = Dprp + TTLkey

Morris, et al. Informational [Page 10]

RFC 7583 Key Timing October 2015

Here, Dprp is the propagation delay -- the time taken for a change
introduced at the master to replicate to all nameservers. TTLkey is
the TTL for the DNSKEY records in the zone. The sum is therefore the
maximum time taken for existing DNSKEY records to expire from caches,
regardless of the nameserver from which they were retrieved.

(The case of introducing the first ZSK into the zone is discussed in
Section 3.3.5.)

After a delay of Ipub, the key is said to be ready and could be used
to sign records. The time at which this event occurs is key N's
ready time (Trdy), which is given by:

Trdy(N) = Tpub(N) + Ipub

Event 3: At some later time, the key starts being used to sign
RRsets. This point is the activation time (Tact) and after this, key
N is said to be active.

Tact(N) >= Trdy(N)

Event 4: At some point thought must be given to its successor (key
N+1). As with the introduction of the currently active key into the
zone, the successor key will need to be published at least Ipub
before it is activated. The publication time of key N+1 depends on
the activation time of key N:

Tpub(N+1) <= Tact(N) + Lzsk - Ipub

Here, Lzsk is the length of time for which a ZSK will be used (the
ZSK lifetime). It should be noted that in the diagrams, the actual
key lifetime is represented; this may differ slightly from the
intended lifetime set by key management policy.

Event 5: While key N is still active, its successor becomes ready.
From this time onwards, key N+1 could be used to sign the zone.

Event 6: When key N has been in use for an interval equal to the ZSK
lifetime, it is retired (i.e., it will never again be used to
generate new signatures) and key N+1 activated and used to sign the
zone. This is the retire time of key N (Tret), and is given by:

Tret(N) = Tact(N) + Lzsk

It is also the activation time of the successor key N+1. Note that
operational considerations may cause key N to remain in use for a
longer (or shorter) time than the lifetime set by the key management
policy.

Morris, et al. Informational [Page 11]

RFC 7583 Key Timing October 2015

Event 7: The retired key needs to be retained in the zone whilst any
RRSIG records created using this key are still published in the zone
or held in caches. (It is possible that a validating resolver could
have an old RRSIG record in the cache, but the old DNSKEY RRset has
expired when it is asked to provide both to a client. In this case
the DNSKEY RRset would need to be looked up again.) This means that
once the key is no longer used to sign records, it should be retained
in the zone for at least the retire interval (Iret) given by:

Iret = Dsgn + Dprp + TTLsig

Dsgn is the delay needed to ensure that all existing RRsets have been
re-signed with the new key. Dprp is the propagation delay, required
to guarantee that the updated zone information has reached all slave
servers, and TTLsig is the maximum TTL of all the RRSIG records in
the zone created with the retiring key.

The time at which all RRSIG records created with this key have
expired from resolver caches is the dead time (Tdea), given by:

Tdea(N) = Tret(N) + Iret

... at which point the key is said to be dead.

Event 8: At any time after the key becomes dead, it can be removed
from the zone's DNSKEY RRset, which must then be re-signed with the
current KSK. This time is the removal time (Trem), given by:

Trem(N) >= Tdea(N)

... at which time the key is said to be removed.

3.2.2. Double-Signature Method

In this rollover, a new key is introduced and used to sign the zone;
the old key and signatures are retained. Once all cached DNSKEY and/
or RRSIG information contains copies of the new DNSKEY and RRSIGs
created with it, the old DNSKEY and RRSIGs can be removed from the
zone.

The timeline for a Double-Signature rollover is shown below. The
diagram follows the convention described in Section 3.2.1.

Morris, et al. Informational [Page 12]

RFC 7583 Key Timing October 2015

|1| |2| |3| |4|
| | | |
Key N |<-------Lzsk----------->|<--->|
| | | |
| |<--Iret-->| |
| | | |
Key N+1 | |<----Lzsk------- - -
| | | |
Key N Tact Tdea Trem
Key N+1 Tact

---- Time ---->

Figure 2: Timeline for a Double-Signature ZSK Rollover

Event 1: Key N is added to the DNSKEY RRset and is then used to sign
the zone; existing signatures in the zone are not removed. The key
is published and active: this is key N's activation time (Tact),
after which the key is said to be active.

Event 2: As the current key (key N) approaches the end of its actual
lifetime (Lzsk), the successor key (key N+1) is introduced into the
zone and starts being used to sign RRsets: neither the current key
nor the signatures created with it are removed. The successor key is
now also active.

Tact(N+1) = Tact(N) + Lzsk - Iret

Event 3: Before key N can be withdrawn from the zone, all RRsets that
need to be signed must have been signed by the successor key (key
N+1) and any old RRsets that do not include the new key or new RRSIGs
must have expired from caches. Note that the signatures are not
replaced: each RRset is signed by both the old and new key.

This takes Iret, the retire interval, given by the expression:

Iret = Dsgn + Dprp + max(TTLkey, TTLsig)

As before, Dsgn is the delay needed to ensure that all existing
RRsets have been signed with the new key and Dprp is the propagation
delay, required to guarantee that the updated zone information has
reached all slave servers. The final term (the maximum of TTLkey and
TTLsig) is the period to wait for key and signature data associated
with key N to expire from caches. (TTLkey is the TTL of the DNSKEY
RRset and TTLsig is the maximum TTL of all the RRSIG records in the
zone created with the ZSK. The two may be different: although the

Morris, et al. Informational [Page 13]

RFC 7583 Key Timing October 2015

TTL of an RRSIG is equal to the TTL of the RRs in the associated
RRset [RFC4034], the DNSKEY RRset only needs to be signed with the
KSK.)

At the end of this interval, key N is said to be dead. This occurs
at the dead time (Tdea) so:

Tdea(N) = Tact(N+1) + Iret

Event 4: At some later time, key N and the signatures generated with
it can be removed from the zone. This is the removal time (Trem),
given by:

Trem(N) >= Tdea(N)

3.3. KSK Rollover Timelines

The following sections describe the rolling of a KSK. They show the
events in the lifetime of a key (referred to as "key N") and cover it
replacement by its successor (key N+1). (The case of introducing the
first KSK into the zone is discussed in Section 3.3.5.)

3.3.1. Double-KSK Method

In this rollover, the new DNSKEY is added to the zone. After an
interval long enough to guarantee that any cached DNSKEY RRsets
contain the new DNSKEY, the DS record in the parent zone is changed.
After a further interval to allow the old DS record to expire from
caches, the old DNSKEY is removed from the zone.

The timeline for a Double-KSK rollover is shown below. The diagram
follows the convention described in Section 3.2.1.

Morris, et al. Informational [Page 14]

RFC 7583 Key Timing October 2015

|1| |2| |3| |4|
| | | |
Key N |<-IpubC->|<--->|<-Dreg->|<-----Lksk--- - -
| | | |
Key N+1 | | | |
| | | |
Key N Tpub Trdy Tsbm Tact
Key N+1

---- Time ---->

(continued ...)

|5| |6| |7| |8| |9| |10|
| | | | | |
Key N - - --------------Lksk------->|<-Iret->|<----->|
| | | | | |
Key N+1 |<-IpubC->|<--->|<-Dreg->|<--------Lksk----- - -
| | | | | |
Key N Tret Tdea Trem
Key N+1 Tpub Trdy Tsbm Tact

---- Time (cont.) ---->

Figure 3: Timeline for a Double-KSK Rollover

Event 1: Key N is introduced into the zone; it is added to the DNSKEY
RRset, which is then signed by all currently active KSKs. (So at
this point, the DNSKEY RRset is signed by both key N and its
predecessor KSK. If other KSKs were active, it is signed by these as
well.) This is the publication time of key N (Tpub); after this, the
key is said to be published.

Event 2: Before it can be used, the key must be published for long
enough to guarantee that any validating resolver that has a copy of
the DNSKEY RRset in its cache will have a copy of the RRset that
includes this key: in other words, that any prior cached information
about the DNSKEY RRset has expired.

The interval is the publication interval in the child zone (IpubC)
and is given by:

IpubC = DprpC + TTLkey

Morris, et al. Informational [Page 15]

RFC 7583 Key Timing October 2015

... where DprpC is the propagation delay for the child zone (the zone
containing the KSK being rolled) and TTLkey the TTL for the DNSKEY
RRset. The time at which this occurs is the key N's ready time,
Trdy, given by:

Trdy(N) = Tpub(N) + IpubC

Event 3: At some later time, the DS record corresponding to the new
KSK is submitted to the parent zone for publication. This time is
the submission time, Tsbm:

Tsbm(N) >= Trdy(N)

Event 4: The DS record is published in the parent zone. As this is
the point at which all information for authentication -- both DNSKEY
and DS record -- is available in the two zones, in analogy with other
rollover methods, this is called the activation time of key N (Tact):

Tact(N) = Tsbm(N) + Dreg

... where Dreg is the registration delay, the time taken after the DS
record has been submitted to the parent zone manager for it to be
placed in the zone. (Parent zones are often managed by different
entities, and this term accounts for the organizational overhead of
transferring a record. In practice, Dreg will not be a fixed time:
instead, the end of Dreg will be signaled by the appearance of the DS
record in the parent zone.)

Event 5: While key N is active, thought needs to be given to its
successor (key N+1). At some time before the scheduled end of the
KSK lifetime, the successor KSK is published in the zone. (As
before, this means that the DNSKEY RRset is signed by all KSKs.)
This time is the publication time of the successor key N+1, given by:

Tpub(N+1) <= Tact(N) + Lksk - Dreg - IpubC

... where Lksk is the actual lifetime of the KSK, and Dreg the
registration delay.

Event 6: After an interval IpubC, key N+1 becomes ready (in that all
caches that have a copy of the DNSKEY RRset have a copy of this key).
This time is the ready time of the successor key N+1 (Trdy).

Event 7: At the submission time of the successor key N+1, Tsbm(N+1),
the DS record corresponding to key N+1 is submitted to the parent
zone.

Morris, et al. Informational [Page 16]

RFC 7583 Key Timing October 2015

Event 8: The successor DS record is published in the parent zone and
the current DS record withdrawn. Key N is said to be retired and the
time at which this occurs is Tret(N), given by:

Tret(N) = Tsbm(N+1) + Dreg

Event 9: Key N must remain in the zone until any caches that contain
a copy of the DS RRset have a copy containing the new DS record.
This interval is the retire interval, given by:

Iret = DprpP + TTLds

... where DprpP is the propagation delay in the parent zone and TTLds
the TTL of a DS record in the parent zone.

As the key is no longer used for anything, it is said to be dead.
This point is the dead time (Tdea), given by:

Tdea(N) = Tret(N) + Iret

Event 10: At some later time, key N is removed from the zone's DNSKEY
RRset (at the remove time Trem); the key is now said to be removed.

Trem(N) >= Tdea(N)

3.3.2. Double-DS Method

In this rollover, the new DS record is published in the parent zone.
When any caches that contain the DS RRset contain a copy of the new
record, the KSK in the zone is changed. After a further interval for
the old DNSKEY RRset to expire from caches, the old DS record is
removed from the parent.

The timeline for a Double-DS rollover is shown below. The diagram
follows the convention described in Section 3.2.1.

Morris, et al. Informational [Page 17]

RFC 7583 Key Timing October 2015

|1| |2| |3| |4| |5|
| | | | |
Key N |<-Dreg->|<-IpubP->|<-->|<-------Lksk---- - -
| | | | |
Key N+1 | | | | |<--Dreg-- - -
| | | | |
Key N Tsbm Tpub Trdy Tact
Key N+1 Tsbm
---- Time ---->

(continued ...)

|6| |7| |8| |9| |10|
| | | | |
Key N - ----------Lksk--------->|<-Iret->|<---->|
| | | | |
Key N+1 - --Dreg-->|<-IpubP->|<-->|<---Lksk------- - -
| | | | |
Key N Tret Tdea Trem
Key N+1 Tpub Trdy Tact

---- Time ---->

Figure 4: Timeline for a Double-DS KSK Rollover

Event 1: The DS RR is submitted to the parent zone for publication.
This time is the submission time, Tsbm.

Event 2: After the registration delay, Dreg, the DS record is
published in the parent zone. This is the publication time (Tpub) of
key N, given by:

Tpub(N) = Tsbm(N) + Dreg

As before, in practice, Dreg will not be a fixed time. Instead, the
end of Dreg will be signaled by the appearance of the DS record in
the parent zone.

Event 3: At some later time, any cache that has a copy of the DS
RRset will have a copy of the DS record for key N. At this point,
key N, if introduced into the DNSKEY RRset, could be used to validate
the zone. For this reason, this time is known as the ready time,
Trdy, and is given by:

Trdy(N) = Tpub(N) + IpubP

Morris, et al. Informational [Page 18]

RFC 7583 Key Timing October 2015

IpubP is the publication interval of the DS record (in the parent
zone) and is given by the expression:

IpubP = DprpP + TTLds

... where DprpP is the propagation delay for the parent zone and
TTLds the TTL assigned to DS records in that zone.

Event 4: At some later time, the key rollover takes place and the new
key (key N) is introduced into the DNSKEY RRset and used to sign it.
This time is key N's activation time (Tact) and at this point key N
is said to be active:

Tact(N) >= Trdy(N)

Event 5: At some point, thought must be given to key replacement.
The DS record for the successor key must be submitted to the parent
zone at a time such that when the current key is withdrawn, any cache
that contains the zone's DS records has data about the DS record of
the successor key. The time at which this occurs is the submission
time of the successor key N+1, given by:

Tsbm(N+1) <= Tact(N) + Lksk - IpubP - Dreg

... where Lksk is the actual lifetime of key N (which may differ
slightly from the lifetime set in the key management policy) and Dreg
is the registration delay.

Event 6. After an interval Dreg, the successor DS record is
published in the zone.

Event 7: The successor key (key N+1) enters the ready state, i.e.,
its DS record is now in caches that contain the parent DS RRset.

Event 8: When key N has been active for its lifetime (Lksk), it is
replaced in the DNSKEY RRset by key N+1; the RRset is then signed
with the new key. At this point, as both the old and new DS records
have been in the parent zone long enough to ensure that they are in
caches that contain the DS RRset, the zone can be authenticated
throughout the rollover. A validating resolver can authenticate
either the old or new KSK.

This time is the retire time (Tret) of key N, given by:

Tret(N) = Tact(N) + Lksk

This is also the activation time of the successor key N+1.

Morris, et al. Informational [Page 19]

RFC 7583 Key Timing October 2015

Event 9: At some later time, all copies of the old DNSKEY RRset have
expired from caches and the old DS record is no longer needed. In
analogy with other rollover methods, this is called the dead time,
Tdea, and is given by:

Tdea(N) = Tret(N) + Iret

... where Iret is the retire interval of the key, given by:

Iret = DprpC + TTLkey

As before, this term includes DprpC, the time taken to propagate the
RRset change through the master-slave hierarchy of the child zone and
TTLkey, the time taken for the DNSKEY RRset to expire from caches.

Event 10: At some later time, the DS record is removed from the
parent zone. In analogy with other rollover methods, this is the
removal time (Trem), given by:

Trem(N) >= Tdea(N)

3.3.3. Double-RRset Method

In the Double-RRset rollover, the new DNSKEY and DS records are
published simultaneously in the appropriate zones. Once enough time
has elapsed for the old DNSKEY and DS RRsets to expire from caches,
the old DNSKEY and DS records are removed from their respective
zones.

The timeline for this rollover is shown below. The diagram follows
the convention described in Section 3.2.1.

|1| |2| |3| |4| |5|
| | | | |
Key N |<-----------Lksk---------->|<---->|
| | | | |
| |<------Ipub----->| |
| | | | |
| |<-Dreg->|<-Iret->| |
| | | | |
Key N+1 | | |<----Lksk-------- - -
| | | | |
Key N Tact Tret Tdea Trem
Key N+1 Tpub Tact

---- Time ---->

Figure 5: Timeline for a Double-RRset KSK Rollover

Morris, et al. Informational [Page 20]

RFC 7583 Key Timing October 2015

Event 1: The DS and DNSKEY records have appeared in their respective
zones and the latter has been used to sign the DNSKEY RRset. The key
is published and active: this is key N's activation time (Tact).

Event 2: As the current key (key N) approaches the end of its actual
lifetime (Lksk), the successor key (key N+1) is introduced into the
zone and is used to sign the DNSKEY RRset. At the same time, the
successor DS record is submitted to the parent zone. This is the
publication time of the successor key (Tpub):

Tpub(N+1) <= Tact(N) + Lksk - Ipub

... where Ipub is defined below.

Event 3: After the registration delay (Dreg), the DS record appears
in the parent zone. The DNSKEY record is already in the child zone,
so with both the new key and its associated data now visible, this is
the key's activation time (Tact) and the key is now said to be
active.

Tact(N+1) = Tpub(N+1) + Dreg

Event 4: Before key N and its associated data can be withdrawn, all
RRsets in the caches of validating resolvers must contain the new DS
and/or DNSKEY. The time at which this occurs is the dead time of key
N (Tdea), given by:

Tdea(N) = Tpub(N+1) + Ipub

Ipub is the time it takes to guarantee that any prior cached
information about the DNSKEY and the DS RRsets have expired. For the
DNSKEY, this is the publication interval of the child (IpubC). For
the DS, the publication interval (IpubP) starts once the record
appears in the parent zone, which is Dreg after it has been
submitted. Hence:

Ipub = max(Dreg + IpubP, IpubC)

The parent zone's publication interval is given by:

IpubP = DprpP + TTLds

where DprpP is the parent zone's propagation delay and TTLds is the
TTL of the DS record in that zone.

Morris, et al. Informational [Page 21]

RFC 7583 Key Timing October 2015

The child zone's publication interval is given by a similar equation:

IpubC = DprpC + TTLkey

where DprpC is the propagation delay in the child zone and TTLkey the
TTL of a DNSKEY record.

In analogy with other rollovers, we can also define a retire interval
-- the interval between a key becoming active and the time at which
its predecessor is considered dead. In this case, Iret is given by:

Iret = Ipub - Dreg

In other words, the retire interval of the predecessor key is the
greater of the publication interval of the parent, or the publication
interval of the child minus the registration delay.

Event 5: At some later time, the key N's DS and DNSKEY records are
removed from their respective zones. In analogy with other rollover
methods, this is the removal time (Trem), given by:

Trem(N) >= Tdea(N)

3.3.4. Interaction with Configured Trust Anchors

Although the preceding sections have been concerned with rolling
KSKs, where the trust anchor is a DS record in the parent zone, zone
managers may want to take account of the possibility that some
validating resolvers may have configured trust anchors directly.

Rolling a configured trust anchor is dealt with in [RFC5011]. It
requires introducing the KSK to be used as the trust anchor into the
zone for a period of time before use and retaining it (with the
"revoke" bit set) for some time after use.

3.3.4.1. Addition of KSK

When the new key is introduced, the expression for the publication
interval of the DNSKEY (IpubC) in the Double-KSK and Double-RRset
methods is modified to:

IpubC >= DprpC + max(Itrp, TTLkey)

Morris, et al. Informational [Page 22]

RFC 7583 Key Timing October 2015

... where the right-hand side of the expression now includes the
"trust point" interval. This term is the interval required to
guarantee that a resolver configured for the automatic update of keys
according to [RFC5011] will accept the new key as a new trust point.
That interval is given by:

Itrp >= queryInterval + AddHoldDownTime + queryInterval

... where queryInterval is as defined in Section 2.3 of [RFC5011] and
AddHoldDownTime is the Add Hold-Down Time defined in Section 2.4.1 of
the same document.

The first term of the expression (queryInterval) represents the time
after which all validating resolvers can be guaranteed to have
obtained a copy of the DNSKEY RRset containing the new key. Once
retrieved, a validating resolver needs to wait for AddHoldDownTime.
Providing it does not see a validly signed DNSKEY RRset without the
new key in that period, it will treat it as a trust anchor the next
time it retrieves the RRset, a process that can take up to another
queryInterval (the third term).

However, the expression for queryInterval given in [RFC5011] contains
the DNSKEY's RRSIG expiration interval, a parameter that only the
validating resolver can really calculate. In practice, a modified
query interval that depends only on TTLkey can be used:

modifiedQueryInterval = MAX(1hr, MIN(15 days, TTLkey / 2))

(This is obtained by taking the expression for queryInterval in
[RFC5011] and assuming a worst case for RRsigExpirationInterval. It
is greater than or equal to queryInterval for all values of the
expiration time.) The expression above then becomes (after
collecting terms):

Itrp >= AddHoldDownTime + 2 * modifiedQueryInterval

In the Double-DS method, instead of swapping the KSK RRs in a single
step, there must now be a period of overlap. In other words, the new
KSK must be introduced into the zone at least:

DprpC + max(Itrp, TTLkey)

... before the switch is made.

Morris, et al. Informational [Page 23]

RFC 7583 Key Timing October 2015

3.3.4.2. Removal of KSK

The timeline for the removal of the key in all methods is modified by
introducing a new state, "revoked". When the key reaches its dead
time, instead of being declared "dead", it is revoked; the "revoke"
bit is set in the published DNSKEY RR, and the DNSKEY RRset re-signed
with the current and revoked keys. The key is maintained in this
state for the revoke interval, Irev, given by:

Irev >= DprpC + modifiedQueryInterval

As before, DprpC is the time taken for the revoked DNSKEY to
propagate to all slave zones, and modifiedQueryInterval is the time
after which it can be guaranteed that all validating resolvers that
adhere to RFC 5011 have retrieved a copy of the DNSKEY RRset
containing the revoked key.

After this time, the key is dead and can be removed from the zone.

3.3.5. Introduction of First Keys

There are no timing considerations associated with the introduction
of the first keys into a zone other that they must be introduced and
the zone validly signed before a chain of trust to the zone is
created.

In the case of a secure parent, it means ensuring that the DS record
is not published in the parent zone until there is no possibility
that a validating resolver can obtain the record yet is not able to
obtain the corresponding DNSKEY. In the case of an insecure parent,
i.e., the initial creation of a chain of trust or "security apex", it
is not possible to guarantee this. It is up to the operator of the
validating resolver to wait for the new KSK to appear at all servers
for the zone before configuring the trust anchor.

4. Standby Keys

Although keys will usually be rolled according to some regular
schedule, there may be occasions when an emergency rollover is
required, e.g., if the active key is suspected of being compromised.
The aim of the emergency rollover is to allow the zone to be
re-signed with a new key as soon as possible. As a key must be in
the ready state to sign the zone, having at least one additional key
(a standby key) in this state at all times will minimize delay.

In the case of a ZSK, a standby key only really makes sense with the
Pre-Publication method. A permanent standby DNSKEY RR should be
included in the zone or successor keys could be introduced as soon as

Morris, et al. Informational [Page 24]

RFC 7583 Key Timing October 2015

possible after a key becomes active. Either way results in one or
more additional ZSKs in the DNSKEY RRset that can immediately be used
to sign the zone if the current key is compromised.

(Although, in theory, the mechanism could be used with both the
Double-Signature and Double-RRSIG methods, it would require
pre-publication of the signatures. Essentially, the standby key
would be permanently active, as it would have to be periodically used
to renew signatures. Zones would also permanently require two sets
of signatures.)

It is also possible to have a standby KSK. The Double-KSK method
requires that the standby KSK be included in the DNSKEY RRset;
rolling the key then requires just the introduction of the DS record
in the parent. Note that the standby KSK should also be used to sign
the DNSKEY RRset. As the RRset and its signatures travel together,
merely adding the KSK without using it to sign the DNSKEY RRset does
not provide the desired time saving: for a KSK to be used in a
rollover, the DNSKEY RRset must be signed with it, and this would
introduce a delay while the old RRset (not signed with the new key)
expires from caches.

The idea of a standby KSK in the Double-RRset rollover method
effectively means having two active keys (as the standby KSK and
associated DS record would both be published at the same time in
their respective zones).

Finally, in the Double-DS method of rolling a KSK, it is not a
standby key that is present, it is a standby DS record in the parent
zone.

Whatever algorithm is used, the standby item of data can be included
in the zone on a permanent basis, or be a successor introduced as
early as possible.

5. Algorithm Considerations

The preceding sections have implicitly assumed that all keys and
signatures are created using a single algorithm. However,
Section 2.2 of [RFC4035] requires that there be an RRSIG for each
RRset using at least one DNSKEY of each algorithm in the zone apex
DNSKEY RRset.

Except in the case of an algorithm rollover -- where the algorithms
used to create the signatures are being changed -- there is no
relationship between the keys of different algorithms. This means
that they can be rolled independently of one another. In other

Morris, et al. Informational [Page 25]

RFC 7583 Key Timing October 2015

words, the key-rollover logic described above should be run
separately for each algorithm; the union of the results is included
in the zone, which is signed using the active key for each algorithm.

6. Summary

For ZSKs, the Pre-Publication method is generally considered to be
the preferred way of rolling keys. As shown in this document, the
time taken to roll is wholly dependent on parameters under the
control of the zone manager.

In contrast, the Double-RRset method is the most efficient for KSK
rollover due to the ability to have new DS records and DNSKEY RRsets
propagate in parallel. The time taken to roll KSKs may depend on
factors related to the parent zone if the parent is signed. For
zones that intend to comply with the recommendations of [RFC5011], in
many cases, the rollover time will be determined by the times defined
by RFC 5011. It should be emphasized that this delay is a policy
choice and not a function of timing values and that it also requires
changes to the rollover process due to the need to manage revocation
of trust anchors.

Finally, the treatment of emergency key rollover is significantly
simplified by the introduction of standby keys as standard practice
during all types of rollovers.

7. Security Considerations

This document does not introduce any new security issues beyond those
already discussed in [RFC4033], [RFC4034], [RFC4035], and [RFC5011].

8. Normative References

[RFC4033] Arends, R., Austein, R., Larson, M., Massey, D., and S.
Rose, "DNS Security Introduction and Requirements",
RFC 4033, DOI 10.17487/RFC4033, March 2005,
<http://www.rfc-editor.org/info/rfc4033>.

[RFC4034] Arends, R., Austein, R., Larson, M., Massey, D., and S.
Rose, "Resource Records for the DNS Security Extensions",
RFC 4034, DOI 10.17487/RFC4034, March 2005,
<http://www.rfc-editor.org/info/rfc4034>.

[RFC4035] Arends, R., Austein, R., Larson, M., Massey, D., and S.
Rose, "Protocol Modifications for the DNS Security
Extensions", RFC 4035, DOI 10.17487/RFC4035, March 2005,
<http://www.rfc-editor.org/info/rfc4035>.

Morris, et al. Informational [Page 26]

RFC 7583 Key Timing October 2015

[RFC5011] StJohns, M., "Automated Updates of DNS Security (DNSSEC)
Trust Anchors", STD 74, RFC 5011, DOI 10.17487/RFC5011,
September 2007, <http://www.rfc-editor.org/info/rfc5011>.

[RFC6781] Kolkman, O., Mekking, W., and R. Gieben, "DNSSEC
Operational Practices, Version 2", RFC 6781,
DOI 10.17487/RFC6781, December 2012,
<http://www.rfc-editor.org/info/rfc6781>.

Morris, et al. Informational [Page 27]

RFC 7583 Key Timing October 2015

Appendix A. List of Symbols

The document defines a number of symbols, all of which are listed
here. All are of the form:

<TYPE><id><ZONE>

where:

<TYPE> is an uppercase character indicating what type the symbol is.
Defined types are:

D delay: interval that is a feature of the process

I interval between two events

L lifetime: interval set by the zone manager

T a point in time

TTL TTL of a record

I, T, and TTL are self-explanatory. Like I, both D and L are time
periods, but whereas I values are intervals between two events, a "D"
interval (delay) is a feature of the process, probably outside
control of the zone manager, and an "L" interval (lifetime) is chosen
by the zone manager and is a feature of policy.

<id> is lowercase and defines what object or event the variable is
related to, e.g.,

act activation

pub publication

ret retire

<ZONE> is an optional uppercase letter that distinguishes between the
same variable applied to different zones and is one of:

C child

P parent

Within the rollover descriptions, times may have a number in
parentheses affixed to their end indicating the instance of the key
to which they apply, e.g., Tact(N) is the activation time of key N,
Tpub(N+1) the publication time of key N+1 etc.

Morris, et al. Informational [Page 28]

RFC 7583 Key Timing October 2015

The list of variables used in the text given below.

Dprp Propagation delay. The amount of time for a change made at
a master nameserver to propagate to all the slave
nameservers.

DprpC Propagation delay in the child zone.

DprpP Propagation delay in the parent zone.

Dreg Registration delay: the time taken for a DS record
submitted to a parent zone to appear in it. As a parent
zone is often managed by a different organization than that
managing the child zone, the delays associated with passing
data between organizations is captured by this term.

Dsgn Signing delay. After the introduction of a new ZSK, the
amount of time taken for all the RRs in the zone to be
signed with it.

Ipub Publication interval. The amount of time that must elapse
after the publication of a DNSKEY and/or its associated
data before it can be assumed that any resolvers that have
the relevant RRset cached have a copy of the new
information.

IpubC Publication interval in the child zone.

IpubP Publication interval in the parent zone.

Iret Retire interval. The amount of time that must elapse after
a DNSKEY or associated data enters the retire state for any
dependent information (e.g., RRSIG for a ZSK) to be purged
from validating resolver caches.

Irev Revoke interval. The amount of time that a KSK must remain
published with the "revoke" bit set to satisfy
considerations of [RFC5011].

Itrp Trust-point interval. The amount of time that a trust
anchor must be published for in order to guarantee that a
resolver configured for an automatic update of keys will
see the new key at least twice.

Morris, et al. Informational [Page 29]

RFC 7583 Key Timing October 2015

Lksk Lifetime of a KSK. This is the actual amount of time for
which this particular KSK is regarded as the active KSK.
Depending on when the key is rolled over, the actual
lifetime may be longer or shorter than the intended key
lifetime indicated by management policy.

Lzsk Lifetime of a ZSK. This is the actual amount of time for
which the ZSK is used to sign the zone. Depending on when
the key is rolled over, the actual lifetime may be longer
or shorter than the intended key lifetime indicated by
management policy.

Tact Activation time. The time at which the key is regarded as
the principal key for the zone.

Tdea Dead time. The time at which any information held in
validating resolver caches is guaranteed to contain
information related to the successor key. At this point,
the current key and its associated information are not
longed required for validation purposes.

Tpub Publication time. The time that the key or associated data
appears in the zone for the first time.

Trem Removal time. The time at which the key and its associated
information starts being removed from their respective
zones.

Tret Retire time. The time at which successor information
starts being used.

Trdy Ready time. The time at which it can be guaranteed that
validating resolvers that have information about the key
and/or associated data cached have a copy of the new
information.

Tsbm Submission time. The time at which the DS record of a KSK
is submitted to the parent zone.

TTLds Time to live of a DS record.

TTLkey Time to live of a DNSKEY record. (By implication, this is
also the time to live of the signatures on the DNSKEY
RRset.)

TTLsig The maximum time to live of all the RRSIG records in the
zone that were created with the ZSK.

Morris, et al. Informational [Page 30]

RFC 7583 Key Timing October 2015

Acknowledgements

The authors gratefully acknowledge help and contributions from Roy
Arends, Tim Wicinski, and Wouter Wijngaards.

Authors' Addresses

Stephen Morris
Internet Systems Consortium
950 Charter Street
Redwood City, CA 94063
United States

Email: [email protected]
URI: http://www.isc.org

Johan Ihren
Netnod
Franzengatan 5
Stockholm SE-112 51
Sweden

Email: [email protected]
URI: http://www.netnod.se

John Dickinson
Sinodun Internet Technologies Ltd
Magdalen Centre
Oxford Science Park
Robert Robertson Avenue
Oxford, Oxfordshire OX4 4GA
United Kingdom

Email: [email protected]
URI: http://www.sinodun.com

W. (Matthijs) Mekking
Dyn, Inc.
150 Dow St
Manchester NH 03101
United States

Email: [email protected]
URI: https://www.dyn.com

Morris, et al. Informational [Page 31]