4 DNS Operations M. Larson
5 Internet-Draft P. Barber
6 Expires: August 14, 2006 VeriSign
10 Observed DNS Resolution Misbehavior
11 draft-ietf-dnsop-bad-dns-res-05
15 By submitting this Internet-Draft, each author represents that any
16 applicable patent or other IPR claims of which he or she is aware
17 have been or will be disclosed, and any of which he or she becomes
18 aware will be disclosed, in accordance with Section 6 of BCP 79.
20 Internet-Drafts are working documents of the Internet Engineering
21 Task Force (IETF), its areas, and its working groups. Note that
22 other groups may also distribute working documents as Internet-
25 Internet-Drafts are draft documents valid for a maximum of six months
26 and may be updated, replaced, or obsoleted by other documents at any
27 time. It is inappropriate to use Internet-Drafts as reference
28 material or to cite them other than as "work in progress."
30 The list of current Internet-Drafts can be accessed at
31 http://www.ietf.org/ietf/1id-abstracts.txt.
33 The list of Internet-Draft Shadow Directories can be accessed at
34 http://www.ietf.org/shadow.html.
36 This Internet-Draft will expire on August 14, 2006.
40 Copyright (C) The Internet Society (2006).
44 This memo describes DNS iterative resolver behavior that results in a
45 significant query volume sent to the root and top-level domain (TLD)
46 name servers. We offer implementation advice to iterative resolver
47 developers to alleviate these unnecessary queries. The
48 recommendations made in this document are a direct byproduct of
49 observation and analysis of abnormal query traffic patterns seen at
50 two of the thirteen root name servers and all thirteen com/net TLD
55 Larson & Barber Expires August 14, 2006 [Page 1]
57 Internet-Draft Observed DNS Resolution Misbehavior February 2006
60 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
61 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
62 document are to be interpreted as described in RFC 2119 [1].
67 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
68 1.1. A note about terminology in this memo . . . . . . . . . . 3
69 2. Observed iterative resolver misbehavior . . . . . . . . . . . 5
70 2.1. Aggressive requerying for delegation information . . . . . 5
71 2.1.1. Recommendation . . . . . . . . . . . . . . . . . . . . 6
72 2.2. Repeated queries to lame servers . . . . . . . . . . . . . 7
73 2.2.1. Recommendation . . . . . . . . . . . . . . . . . . . . 7
74 2.3. Inability to follow multiple levels of indirection . . . . 8
75 2.3.1. Recommendation . . . . . . . . . . . . . . . . . . . . 9
76 2.4. Aggressive retransmission when fetching glue . . . . . . . 9
77 2.4.1. Recommendation . . . . . . . . . . . . . . . . . . . . 10
78 2.5. Aggressive retransmission behind firewalls . . . . . . . . 10
79 2.5.1. Recommendation . . . . . . . . . . . . . . . . . . . . 11
80 2.6. Misconfigured NS records . . . . . . . . . . . . . . . . . 11
81 2.6.1. Recommendation . . . . . . . . . . . . . . . . . . . . 12
82 2.7. Name server records with zero TTL . . . . . . . . . . . . 12
83 2.7.1. Recommendation . . . . . . . . . . . . . . . . . . . . 13
84 2.8. Unnecessary dynamic update messages . . . . . . . . . . . 13
85 2.8.1. Recommendation . . . . . . . . . . . . . . . . . . . . 14
86 2.9. Queries for domain names resembling IPv4 addresses . . . . 14
87 2.9.1. Recommendation . . . . . . . . . . . . . . . . . . . . 14
88 2.10. Misdirected recursive queries . . . . . . . . . . . . . . 15
89 2.10.1. Recommendation . . . . . . . . . . . . . . . . . . . . 15
90 2.11. Suboptimal name server selection algorithm . . . . . . . . 15
91 2.11.1. Recommendation . . . . . . . . . . . . . . . . . . . . 16
92 3. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 17
93 4. IANA considerations . . . . . . . . . . . . . . . . . . . . . 18
94 5. Security considerations . . . . . . . . . . . . . . . . . . . 19
95 6. Internationalization considerations . . . . . . . . . . . . . 20
96 7. Informative References . . . . . . . . . . . . . . . . . . . . 20
97 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 21
98 Intellectual Property and Copyright Statements . . . . . . . . . . 22
111 Larson & Barber Expires August 14, 2006 [Page 2]
113 Internet-Draft Observed DNS Resolution Misbehavior February 2006
118 Observation of query traffic received by two root name servers and
119 the thirteen com/net TLD name servers has revealed that a large
120 proportion of the total traffic often consists of "requeries". A
121 requery is the same question (<QNAME, QTYPE, QCLASS>) asked
122 repeatedly at an unexpectedly high rate. We have observed requeries
123 from both a single IP address and multiple IP addresses (i.e., the
124 same query received simultaneously from multiple IP addresses).
126 By analyzing requery events we have found that the cause of the
127 duplicate traffic is almost always a deficient iterative resolver,
128 stub resolver or application implementation combined with an
129 operational anomaly. The implementation deficiencies we have
130 identified to date include well-intentioned recovery attempts gone
131 awry, insufficient caching of failures, early abort when multiple
132 levels of indirection must be followed, and aggressive retry by stub
133 resolvers or applications. Anomalies that we have seen trigger
134 requery events include lame delegations, unusual glue records, and
135 anything that makes all authoritative name servers for a zone
136 unreachable (DoS attacks, crashes, maintenance, routing failures,
139 In the following sections, we provide a detailed explanation of the
140 observed behavior and recommend changes that will reduce the requery
141 rate. None of the changes recommended affects the core DNS protocol
142 specification; instead, this document consists of guidelines to
143 implementors of iterative resolvers.
145 1.1. A note about terminology in this memo
147 To recast an old saying about standards, the nice thing about DNS
148 terms is that there are so many of them to choose from. Writing or
149 talking about DNS can be difficult and cause confusion resulting from
150 a lack of agreed-upon terms for its various components. Further
151 complicating matters are implementations that combine multiple roles
152 into one piece of software, which makes naming the result
153 problematic. An example is the entity that accepts recursive
154 queries, issues iterative queries as necessary to resolve the initial
155 recursive query, caches responses it receives, and which is also able
156 to answer questions about certain zones authoritatively. This entity
157 is an iterative resolver combined with an authoritative name server
158 and is often called a "recursive name server" or a "caching name
161 This memo is concerned principally with the behavior of iterative
162 resolvers, which are typically found as part of a recursive name
163 server. This memo uses the more precise term "iterative resolver",
167 Larson & Barber Expires August 14, 2006 [Page 3]
169 Internet-Draft Observed DNS Resolution Misbehavior February 2006
172 because the focus is usually on that component. In instances where
173 the name server role of this entity requires mentioning, this memo
174 uses the term "recursive name server". As an example of the
175 difference, the name server component of a recursive name server
176 receives DNS queries and the iterative resolver component sends
179 The advent of IPv6 requires mentioning AAAA records as well as A
180 records when discussing glue. To avoid continuous repetition and
181 qualification, this memo uses the general term "address record" to
182 encompass both A and AAAA records when a particular situation is
183 relevant to both types.
223 Larson & Barber Expires August 14, 2006 [Page 4]
225 Internet-Draft Observed DNS Resolution Misbehavior February 2006
228 2. Observed iterative resolver misbehavior
230 2.1. Aggressive requerying for delegation information
232 There can be times when every name server in a zone's NS RRset is
233 unreachable (e.g., during a network outage), unavailable (e.g., the
234 name server process is not running on the server host) or
235 misconfigured (e.g., the name server is not authoritative for the
236 given zone, also known as "lame"). Consider an iterative resolver
237 that attempts to resolve a query for a domain name in such a zone and
238 discovers that none of the zone's name servers can provide an answer.
239 We have observed a recursive name server implementation whose
240 iterative resolver then verifies the zone's NS RRset in its cache by
241 querying for the zone's delegation information: it sends a query for
242 the zone's NS RRset to one of the parent zone's name servers. (Note
243 that queries with QTYPE=NS are not required by the standard
244 resolution algorithm described in section 4.3.2 of RFC 1034 [2].
245 These NS queries represent this implementation's addition to that
248 For example, suppose that "example.com" has the following NS RRset:
250 example.com. IN NS ns1.example.com.
251 example.com. IN NS ns2.example.com.
253 Upon receipt of a query for "www.example.com" and assuming that
254 neither "ns1.example.com" nor "ns2.example.com" can provide an
255 answer, this iterative resolver implementation immediately queries a
256 "com" zone name server for the "example.com" NS RRset to verify it
257 has the proper delegation information. This implementation performs
258 this query to a zone's parent zone for each recursive query it
259 receives that fails because of a completely unresponsive set of name
260 servers for the target zone. Consider the effect when a popular zone
261 experiences a catastrophic failure of all its name servers: now every
262 recursive query for domain names in that zone sent to this recursive
263 name server implementation results in a query to the failed zone's
264 parent name servers. On one occasion when several dozen popular
265 zones became unreachable, the query load on the com/net name servers
268 We believe this verification query is not reasonable. Consider the
269 circumstances: When an iterative resolver is resolving a query for a
270 domain name in a zone it has not previously searched, it uses the
271 list of name servers in the referral from the target zone's parent.
272 If on its first attempt to search the target zone, none of the name
273 servers in the referral is reachable, a verification query to the
274 parent would be pointless: this query to the parent would come so
275 quickly on the heels of the referral that it would be almost certain
279 Larson & Barber Expires August 14, 2006 [Page 5]
281 Internet-Draft Observed DNS Resolution Misbehavior February 2006
284 to contain the same list of name servers. The chance of discovering
285 any new information is slim.
287 The other possibility is that the iterative resolver successfully
288 contacts one of the target zone's name servers and then caches the NS
289 RRset from the authority section of a response, the proper behavior
290 according to section 5.4.1 of RFC 2181 [3], because the NS RRset from
291 the target zone is more trustworthy than delegation information from
292 the parent zone. If, while processing a subsequent recursive query,
293 the iterative resolver discovers that none of the name servers
294 specified in the cached NS RRset is available or authoritative,
295 querying the parent would be wrong. An NS RRset from the parent zone
296 would now be less trustworthy than data already in the cache.
298 For this query of the parent zone to be useful, the target zone's
299 entire set of name servers would have to change AND the former set of
300 name servers would have to be deconfigured or decommissioned AND the
301 delegation information in the parent zone would have to be updated
302 with the new set of name servers, all within the TTL of the target
303 zone's NS RRset. We believe this scenario is uncommon:
304 administrative best practices dictate that changes to a zone's set of
305 name servers happen gradually when at all possible, with servers
306 removed from the NS RRset left authoritative for the zone as long as
307 possible. The scenarios that we can envision that would benefit from
308 the parent requery behavior do not outweigh its damaging effects.
310 This section should not be understood to claim that all queries to a
311 zone's parent are bad. In some cases, such queries are not only
312 reasonable but required. Consider the situation when required
313 information, such as the address of a name server (i.e., the address
314 record corresponding to the RDATA of an NS record), has timed out of
315 an iterative resolver's cache before the corresponding NS record. If
316 the name of the name server is below the apex of the zone, then the
317 name server's address record is only available as glue in the parent
318 zone. For example, consider this NS record:
320 example.com. IN NS ns.example.com.
322 If a cache has this NS record but not the address record for
323 "ns.example.com", it is unable to contact the "example.com" zone
324 directly and must query the "com" zone to obtain the address record.
325 Note, however, that such a query would not have QTYPE=NS according to
326 the standard resolution algorithm.
328 2.1.1. Recommendation
330 An iterative resolver MUST NOT send a query for the NS RRset of a
331 non-responsive zone to any of the name servers for that zone's parent
335 Larson & Barber Expires August 14, 2006 [Page 6]
337 Internet-Draft Observed DNS Resolution Misbehavior February 2006
340 zone. For the purposes of this injunction, a non-responsive zone is
341 defined as a zone for which every name server listed in the zone's NS
344 1. is not authoritative for the zone (i.e., lame), or,
346 2. returns a server failure response (RCODE=2), or,
348 3. is dead or unreachable according to section 7.2 of RFC 2308 [4].
350 2.2. Repeated queries to lame servers
352 Section 2.1 describes a catastrophic failure: when every name server
353 for a zone is unable to provide an answer for one reason or another.
354 A more common occurrence is when a subset of a zone's name servers
355 are unavailable or misconfigured. Different failure modes have
356 different expected durations. Some symptoms indicate problems that
357 are potentially transient; for example, various types of ICMP
358 unreachable messages because a name server process is not running or
359 a host or network is unreachable, or a complete lack of a response to
360 a query. Such responses could be the result of a host rebooting or
361 temporary outages; these events don't necessarily require any human
362 intervention and can be reasonably expected to be temporary.
364 Other symptoms clearly indicate a condition requiring human
365 intervention, such as lame server: if a name server is misconfigured
366 and not authoritative for a zone delegated to it, it is reasonable to
367 assume that this condition has potential to last longer than
368 unreachability or unresponsiveness. Consequently, repeated queries
369 to known lame servers are not useful. In this case of a condition
370 with potential to persist for a long time, a better practice would be
371 to maintain a list of known lame servers and avoid querying them
372 repeatedly in a short interval.
374 It should also be noted, however, that some authoritative name server
375 implementations appear to be lame only for queries of certain types
376 as described in RFC 4074 [5]. In this case, it makes sense to retry
377 the "lame" servers for other types of queries, particularly when all
378 known authoritative name servers appear to be "lame".
380 2.2.1. Recommendation
382 Iterative resolvers SHOULD cache name servers that they discover are
383 not authoritative for zones delegated to them (i.e. lame servers).
384 If this caching is performed, lame servers MUST be cached against the
385 specific query tuple <zone name, class, server IP address>. Zone
386 name can be derived from the owner name of the NS record that was
387 referenced to query the name server that was discovered to be lame.
391 Larson & Barber Expires August 14, 2006 [Page 7]
393 Internet-Draft Observed DNS Resolution Misbehavior February 2006
396 Implementations that perform lame server caching MUST refrain from
397 sending queries to known lame servers based on a time interval from
398 when the server is discovered to be lame. A minimum interval of
399 thirty minutes is RECOMMENDED.
401 An exception to this recommendation occurs if all name servers for a
402 zone are marked lame. In that case, the iterative resolver SHOULD
403 temporarily ignore the servers' lameness status and query one or more
404 servers. This behavior is a workaround for the type-specific
405 lameness issue described in the previous section.
407 Implementors should take care not to make lame server avoidance logic
408 overly broad: note that a name server could be lame for a parent zone
409 but not a child zone, e.g., lame for "example.com" but properly
410 authoritative for "sub.example.com". Therefore a name server should
411 not be automatically considered lame for subzones. In the case
412 above, even if a name server is known to be lame for "example.com",
413 it should be queried for QNAMEs at or below "sub.example.com" if an
414 NS record indicates it should be authoritative for that zone.
416 2.3. Inability to follow multiple levels of indirection
418 Some iterative resolver implementations are unable to follow
419 sufficient levels of indirection. For example, consider the
420 following delegations:
422 foo.example. IN NS ns1.example.com.
423 foo.example. IN NS ns2.example.com.
425 example.com. IN NS ns1.test.example.net.
426 example.com. IN NS ns2.test.example.net.
428 test.example.net. IN NS ns1.test.example.net.
429 test.example.net. IN NS ns2.test.example.net.
431 An iterative resolver resolving the name "www.foo.example" must
432 follow two levels of indirection, first obtaining address records for
433 "ns1.test.example.net" or "ns2.test.example.net" in order to obtain
434 address records for "ns1.example.com" or "ns2.example.com" in order
435 to query those name servers for the address records of
436 "www.foo.example". While this situation may appear contrived, we
437 have seen multiple similar occurrences and expect more as new generic
438 top-level domains (gTLDs) become active. We anticipate many zones in
439 new gTLDs will use name servers in existing gTLDs, increasing the
440 number of delegations using out-of-zone name servers.
447 Larson & Barber Expires August 14, 2006 [Page 8]
449 Internet-Draft Observed DNS Resolution Misbehavior February 2006
452 2.3.1. Recommendation
454 Clearly constructing a delegation that relies on multiple levels of
455 indirection is not a good administrative practice. However, the
456 practice is widespread enough to require that iterative resolvers be
457 able to cope with it. Iterative resolvers SHOULD be able to handle
458 arbitrary levels of indirection resulting from out-of-zone name
459 servers. Iterative resolvers SHOULD implement a level-of-effort
460 counter to avoid loops or otherwise performing too much work in
461 resolving pathological cases.
463 A best practice that avoids this entire issue of indirection is to
464 name one or more of a zone's name servers in the zone itself. For
465 example, if the zone is named "example.com", consider naming some of
466 the name servers "ns{1,2,...}.example.com" (or similar).
468 2.4. Aggressive retransmission when fetching glue
470 When an authoritative name server responds with a referral, it
471 includes NS records in the authority section of the response.
472 According to the algorithm in section 4.3.2 of RFC 1034 [2], the name
473 server should also "put whatever addresses are available into the
474 additional section, using glue RRs if the addresses are not available
475 from authoritative data or the cache." Some name server
476 implementations take this address inclusion a step further with a
477 feature called "glue fetching". A name server that implements glue
478 fetching attempts to include address records for every NS record in
479 the authority section. If necessary, the name server issues multiple
480 queries of its own to obtain any missing address records.
482 Problems with glue fetching can arise in the context of
483 "authoritative-only" name servers, which only serve authoritative
484 data and ignore requests for recursion. Such an entity will not
485 normally generate any queries of its own. Instead it answers non-
486 recursive queries from iterative resolvers looking for information in
487 zones it serves. With glue fetching enabled, however, an
488 authoritative server invokes an iterative resolver to look up an
489 unknown address record to complete the additional section of a
492 We have observed situations where the iterative resolver of a glue-
493 fetching name server can send queries that reach other name servers,
494 but is apparently prevented from receiving the responses. For
495 example, perhaps the name server is authoritative-only and therefore
496 its administrators expect it to receive only queries and not
497 responses. Perhaps unaware of glue fetching and presuming that the
498 name server's iterative resolver will generate no queries, its
499 administrators place the name server behind a network device that
503 Larson & Barber Expires August 14, 2006 [Page 9]
505 Internet-Draft Observed DNS Resolution Misbehavior February 2006
508 prevents it from receiving responses. If this is the case, all glue-
509 fetching queries will go answered.
511 We have observed name server implementations whose iterative
512 resolvers retry excessively when glue-fetching queries are
513 unanswered. A single com/net name server has received hundreds of
514 queries per second from a single such source. Judging from the
515 specific queries received and based on additional analysis, we
516 believe these queries result from overly aggressive glue fetching.
518 2.4.1. Recommendation
520 Implementers whose name servers support glue fetching SHOULD take
521 care to avoid sending queries at excessive rates. Implementations
522 SHOULD support throttling logic to detect when queries are sent but
523 no responses are received.
525 2.5. Aggressive retransmission behind firewalls
527 A common occurrence and one of the largest sources of repeated
528 queries at the com/net and root name servers appears to result from
529 resolvers behind misconfigured firewalls. In this situation, an
530 iterative resolver is apparently allowed to send queries through a
531 firewall to other name servers, but not receive the responses. The
532 result is more queries than necessary because of retransmission, all
533 of which are useless because the responses are never received. Just
534 as with the glue-fetching scenario described in Section 2.4, the
535 queries are sometimes sent at excessive rates. To make matters
536 worse, sometimes the responses, sent in reply to legitimate queries,
537 trigger an alarm on the originator's intrusion detection system. We
538 are frequently contacted by administrators responding to such alarms
539 who believe our name servers are attacking their systems.
541 Not only do some resolvers in this situation retransmit queries at an
542 excessive rate, but they continue to do so for days or even weeks.
543 This scenario could result from an organization with multiple
544 recursive name servers, only a subset of whose iterative resolvers'
545 traffic is improperly filtered in this manner. Stub resolvers in the
546 organization could be configured to query multiple recursive name
547 servers. Consider the case where a stub resolver queries a filtered
548 recursive name server first. The iterative resolver of this
549 recursive name server sends one or more queries whose replies are
550 filtered, so it can't respond to the stub resolver, which times out.
551 Then the stub resolver retransmits to a recursive name server that is
552 able to provide an answer. Since resolution ultimately succeeds the
553 underlying problem might not be recognized or corrected. A popular
554 stub resolver implementation has a very aggressive retransmission
555 schedule, including simultaneous queries to multiple recursive name
559 Larson & Barber Expires August 14, 2006 [Page 10]
561 Internet-Draft Observed DNS Resolution Misbehavior February 2006
564 servers, which could explain how such a situation could persist
565 without being detected.
567 2.5.1. Recommendation
569 The most obvious recommendation is that administrators SHOULD take
570 care not to place iterative resolvers behind a firewall that allows
571 queries to pass through but not the resulting replies.
573 Iterative resolvers SHOULD take care to avoid sending queries at
574 excessive rates. Implementations SHOULD support throttling logic to
575 detect when queries are sent but no responses are received.
577 2.6. Misconfigured NS records
579 Sometimes a zone administrator forgets to add the trailing dot on the
580 domain names in the RDATA of a zone's NS records. Consider this
581 fragment of the zone file for "example.com":
584 example.com. 3600 IN NS ns1.example.com ; Note missing
585 example.com. 3600 IN NS ns2.example.com ; trailing dots
587 The zone's authoritative servers will parse the NS RDATA as
588 "ns1.example.com.example.com" and "ns2.example.com.example.com" and
589 return NS records with this incorrect RDATA in responses, including
590 typically the authority section of every response containing records
591 from the "example.com" zone.
593 Now consider a typical sequence of queries. An iterative resolver
594 attempting to resolve address records for "www.example.com" with no
595 cached information for this zone will query a "com" authoritative
596 server. The "com" server responds with a referral to the
597 "example.com" zone, consisting of NS records with valid RDATA and
598 associated glue records. (This example assumes that the
599 "example.com" zone delegation information is correct in the "com"
600 zone.) The iterative resolver caches the NS RRset from the "com"
601 server and follows the referral by querying one of the "example.com"
602 authoritative servers. This server responds with the
603 "www.example.com" address record in the answer section and,
604 typically, the "example.com" NS records in the authority section and,
605 if space in the message remains, glue address records in the
606 additional section. According to Section 5.4 of RFC 2181 [3], NS
607 records in the authority section of an authoritative answer are more
608 trustworthy than NS records from the authority section of a non-
609 authoritative answer. Thus the "example.com" NS RRset just received
610 from the "example.com" authoritative server overrides the
611 "example.com" NS RRset received moments ago from the "com"
615 Larson & Barber Expires August 14, 2006 [Page 11]
617 Internet-Draft Observed DNS Resolution Misbehavior February 2006
620 authoritative server.
622 But the "example.com" zone contains the erroneous NS RRset as shown
623 in the example above. Subsequent queries for names in "example.com"
624 will cause the iterative resolver to attempt to use the incorrect NS
625 records and so it will try to resolve the nonexistent names
626 "ns1.example.com.example.com" and "ns2.example.com.example.com". In
627 this example, since all of the zone's name servers are named in the
628 zone itself (i.e., "ns1.example.com.example.com" and
629 "ns2.example.com.example.com" both end in "example.com") and all are
630 bogus, the iterative resolver cannot reach any "example.com" name
631 servers. Therefore attempts to resolve these names result in address
632 record queries to the "com" authoritative servers. Queries for such
633 obviously bogus glue address records occur frequently at the com/net
636 2.6.1. Recommendation
638 An authoritative server can detect this situation. A trailing dot
639 missing from an NS record's RDATA always results by definition in a
640 name server name that exists somewhere under the apex of the zone the
641 NS record appears in. Note that further levels of delegation are
642 possible, so a missing trailing dot could inadvertently create a name
643 server name that actually exists in a subzone.
645 An authoritative name server SHOULD issue a warning when one of a
646 zone's NS records references a name server below the zone's apex when
647 a corresponding address record does not exist in the zone AND there
648 are no delegated subzones where the address record could exist.
650 2.7. Name server records with zero TTL
652 Sometimes a popular com/net subdomain's zone is configured with a TTL
653 of zero on the zone's NS records, which prohibits these records from
654 being cached and will result in a higher query volume to the zone's
655 authoritative servers. The zone's administrator should understand
656 the consequences of such a configuration and provision resources
657 accordingly. A zero TTL on the zone's NS RRset, however, carries
658 additional consequences beyond the zone itself: if an iterative
659 resolver cannot cache a zone's NS records because of a zero TTL, it
660 will be forced to query that zone's parent's name servers each time
661 it resolves a name in the zone. The com/net authoritative servers do
662 see an increased query load when a popular com/net subdomain's zone
663 is configured with a TTL of zero on the zone's NS records.
665 A zero TTL on an RRset expected to change frequently is extreme but
666 permissible. A zone's NS RRset is a special case, however, because
667 changes to it must be coordinated with the zone's parent. In most
671 Larson & Barber Expires August 14, 2006 [Page 12]
673 Internet-Draft Observed DNS Resolution Misbehavior February 2006
676 zone parent/child relationships we are aware of, there is typically
677 some delay involved in effecting changes. Further, changes to the
678 set of a zone's authoritative name servers (and therefore to the
679 zone's NS RRset) are typically relatively rare: providing reliable
680 authoritative service requires a reasonably stable set of servers.
681 Therefore an extremely low or zero TTL on a zone's NS RRset rarely
682 makes sense, except in anticipation of an upcoming change. In this
683 case, when the zone's administrator has planned a change and does not
684 want iterative resolvers throughout the Internet to cache the NS
685 RRset for a long period of time, a low TTL is reasonable.
687 2.7.1. Recommendation
689 Because of the additional load placed on a zone's parent's
690 authoritative servers resulting from a zero TTL on a zone's NS RRset,
691 under such circumstances authoritative name servers SHOULD issue a
692 warning when loading a zone.
694 2.8. Unnecessary dynamic update messages
696 The UPDATE message specified in RFC 2136 [6] allows an authorized
697 agent to update a zone's data on an authoritative name server using a
698 DNS message sent over the network. Consider the case of an agent
699 desiring to add a particular resource record. Because of zone cuts,
700 the agent does not necessarily know the proper zone to which the
701 record should be added. The dynamic update process requires that the
702 agent determine the appropriate zone so the UPDATE message can be
703 sent to one of the zone's authoritative servers (typically the
704 primary master as specified in the zone's SOA MNAME field).
706 The appropriate zone to update is the closest enclosing zone, which
707 cannot be determined only by inspecting the domain name of the record
708 to be updated, since zone cuts can occur anywhere. One way to
709 determine the closest enclosing zone entails walking up the name
710 space tree by sending repeated UPDATE messages until success. For
711 example, consider an agent attempting to add an address record with
712 the name "foo.bar.example.com". The agent could first attempt to
713 update the "foo.bar.example.com" zone. If the attempt failed, the
714 update could be directed to the "bar.example.com" zone, then the
715 "example.com" zone, then the "com" zone, and finally the root zone.
717 A popular dynamic agent follows this algorithm. The result is many
718 UPDATE messages received by the root name servers, the com/net
719 authoritative servers, and presumably other TLD authoritative
720 servers. A valid question is why the algorithm proceeds to send
721 updates all the way to TLD and root name servers. This behavior is
722 not entirely unreasonable: in enterprise DNS architectures with an
723 "internal root" design, there could conceivably be private, non-
727 Larson & Barber Expires August 14, 2006 [Page 13]
729 Internet-Draft Observed DNS Resolution Misbehavior February 2006
732 public TLD or root zones that would be the appropriate targets for a
735 A significant deficiency with this algorithm is that knowledge of a
736 given UPDATE message's failure is not helpful in directing future
737 UPDATE messages to the appropriate servers. A better algorithm would
738 be to find the closest enclosing zone by walking up the name space
739 with queries for SOA or NS rather than "probing" with UPDATE
740 messages. Once the appropriate zone is found, an UPDATE message can
741 be sent. In addition, the results of these queries can be cached to
742 aid in determining closest enclosing zones for future updates. Once
743 the closest enclosing zone is determined with this method, the update
744 will either succeed or fail and there is no need to send further
745 updates to higher-level zones. The important point is that walking
746 up the tree with queries yields cacheable information, whereas
747 walking up the tree by sending UPDATE messages does not.
749 2.8.1. Recommendation
751 Dynamic update agents SHOULD send SOA or NS queries to progressively
752 higher-level names to find the closest enclosing zone for a given
753 name to update. Only after the appropriate zone is found should the
754 client send an UPDATE message to one of the zone's authoritative
755 servers. Update clients SHOULD NOT "probe" using UPDATE messages by
756 walking up the tree to progressively higher-level zones.
758 2.9. Queries for domain names resembling IPv4 addresses
760 The root name servers receive a significant number of A record
761 queries where the QNAME looks like an IPv4 address. The source of
762 these queries is unknown. It could be attributed to situations where
763 a user believes an application will accept either a domain name or an
764 IP address in a given configuration option. The user enters an IP
765 address, but the application assumes any input is a domain name and
766 attempts to resolve it, resulting in an A record lookup. There could
767 also be applications that produce such queries in a misguided attempt
768 to reverse map IP addresses.
770 These queries result in Name Error (RCODE=3) responses. An iterative
771 resolver can negatively cache such responses, but each response
772 requires a separate cache entry, i.e., a negative cache entry for the
773 domain name "192.0.2.1" does not prevent a subsequent query for the
774 domain name "192.0.2.2".
776 2.9.1. Recommendation
778 It would be desirable for the root name servers not to have to answer
779 these queries: they unnecessarily consume CPU resources and network
783 Larson & Barber Expires August 14, 2006 [Page 14]
785 Internet-Draft Observed DNS Resolution Misbehavior February 2006
788 bandwidth. A possible solution is to delegate these numeric TLDs
789 from the root zone to a separate set of servers to absorb the
790 traffic. The "black hole servers" used by the AS 112 Project [8],
791 which are currently delegated the in-addr.arpa zones corresponding to
792 RFC 1918 [7] private use address space, would be a possible choice to
793 receive these delegations. Of course, the proper and usual root zone
794 change procedures would have to be followed to make such a change to
797 2.10. Misdirected recursive queries
799 The root name servers receive a significant number of recursive
800 queries (i.e., queries with the RD bit set in the header). Since
801 none of the root servers offers recursion, the servers' response in
802 such a situation ignores the request for recursion and the response
803 probably does not contain the data the querier anticipated. Some of
804 these queries result from users configuring stub resolvers to query a
805 root server. (This situation is not hypothetical: we have received
806 complaints from users when this configuration does not work as
807 hoped.) Of course, users should not direct stub resolvers to use
808 name servers that do not offer recursion, but we are not aware of any
809 stub resolver implementation that offers any feedback to the user
810 when so configured, aside from simply "not working".
812 2.10.1. Recommendation
814 When the IP address of a name server that supposedly offers recursion
815 is configured in a stub resolver using an interactive user interface,
816 the resolver could send a test query to verify that the server indeed
817 supports recursion (i.e., verify that the response has the RA bit set
818 in the header). The user could be immediately notified if the server
821 The stub resolver could also report an error, either through a user
822 interface or in a log file, if the queried server does not support
823 recursion. Error reporting SHOULD be throttled to avoid a
824 notification or log message for every response from a non-recursive
827 2.11. Suboptimal name server selection algorithm
829 An entire document could be devoted to the topic of problems with
830 different implementations of the recursive resolution algorithm. The
831 entire process of recursion is woefully under specified, requiring
832 each implementor to design an algorithm. Sometimes implementors make
833 poor design choices that could be avoided if a suggested algorithm
834 and best practices were documented, but that is a topic for another
839 Larson & Barber Expires August 14, 2006 [Page 15]
841 Internet-Draft Observed DNS Resolution Misbehavior February 2006
844 Some deficiencies cause significant operational impact and are
845 therefore worth mentioning here. One of these is name server
846 selection by an iterative resolver. When an iterative resolver wants
847 to contact one of a zone's authoritative name servers, how does it
848 choose from the NS records listed in the zone's NS RRset? If the
849 selection mechanism is suboptimal, queries are not spread evenly
850 among a zone's authoritative servers. The details of the selection
851 mechanism are up to the implementor, but we offer some suggestions.
853 2.11.1. Recommendation
855 This list is not conclusive, but reflects the changes that would
856 produce the most impact in terms of reducing disproportionate query
857 load among a zone's authoritative servers. I.e., these changes would
858 help spread the query load evenly.
860 o Do not make assumptions based on NS RRset order: all NS RRs SHOULD
861 be treated equally. (In the case of the "com" zone, for example,
862 most of the root servers return the NS record for "a.gtld-
863 servers.net" first in the authority section of referrals.
864 Apparently as a result, this server receives disproportionately
865 more traffic than the other 12 authoritative servers for "com".)
867 o Use all NS records in an RRset. (For example, we are aware of
868 implementations that hard-coded information for a subset of the
871 o Maintain state and favor the best-performing of a zone's
872 authoritative servers. A good definition of performance is
873 response time. Non-responsive servers can be penalized with an
874 extremely high response time.
876 o Do not lock onto the best-performing of a zone's name servers. An
877 iterative resolver SHOULD periodically check the performance of
878 all of a zone's name servers to adjust its determination of the
895 Larson & Barber Expires August 14, 2006 [Page 16]
897 Internet-Draft Observed DNS Resolution Misbehavior February 2006
902 The authors would like to thank the following people for their
903 comments that improved this document: Andras Salamon, Dave Meyer,
904 Doug Barton, Jaap Akkerhuis, Jinmei Tatuya, John Brady, Kevin Darcy,
905 Olafur Gudmundsson, Pekka Savola, Peter Koch and Rob Austein. We
906 apologize if we have omitted anyone; any oversight was unintentional.
951 Larson & Barber Expires August 14, 2006 [Page 17]
953 Internet-Draft Observed DNS Resolution Misbehavior February 2006
956 4. IANA considerations
958 There are no new IANA considerations introduced by this memo.
1007 Larson & Barber Expires August 14, 2006 [Page 18]
1009 Internet-Draft Observed DNS Resolution Misbehavior February 2006
1012 5. Security considerations
1014 The iterative resolver misbehavior discussed in this document exposes
1015 the root and TLD name servers to increased risk of both intentional
1016 and unintentional denial of service attacks.
1018 We believe that implementation of the recommendations offered in this
1019 document will reduce the amount of unnecessary traffic seen at root
1020 and TLD name servers, thus reducing the opportunity for an attacker
1021 to use such queries to his or her advantage.
1063 Larson & Barber Expires August 14, 2006 [Page 19]
1065 Internet-Draft Observed DNS Resolution Misbehavior February 2006
1068 6. Internationalization considerations
1070 There are no new internationalization considerations introduced by
1073 7. Informative References
1075 [1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
1076 Levels", BCP 14, RFC 2119, March 1997.
1078 [2] Mockapetris, P., "Domain names - concepts and facilities",
1079 STD 13, RFC 1034, November 1987.
1081 [3] Elz, R. and R. Bush, "Clarifications to the DNS Specification",
1082 RFC 2181, July 1997.
1084 [4] Andrews, M., "Negative Caching of DNS Queries (DNS NCACHE)",
1085 RFC 2308, March 1998.
1087 [5] Morishita, Y. and T. Jinmei, "Common Misbehavior Against DNS
1088 Queries for IPv6 Addresses", RFC 4074, May 2005.
1090 [6] Vixie, P., Thomson, S., Rekhter, Y., and J. Bound, "Dynamic
1091 Updates in the Domain Name System (DNS UPDATE)", RFC 2136,
1094 [7] Rekhter, Y., Moskowitz, R., Karrenberg, D., Groot, G., and E.
1095 Lear, "Address Allocation for Private Internets", BCP 5,
1096 RFC 1918, February 1996.
1098 [8] <http://www.as112.net>
1119 Larson & Barber Expires August 14, 2006 [Page 20]
1121 Internet-Draft Observed DNS Resolution Misbehavior February 2006
1128 21345 Ridgetop Circle
1129 Dulles, VA 20166-6503
1132 Email: mlarson@verisign.com
1137 21345 Ridgetop Circle
1138 Dulles, VA 20166-6503
1141 Email: pbarber@verisign.com
1175 Larson & Barber Expires August 14, 2006 [Page 21]
1177 Internet-Draft Observed DNS Resolution Misbehavior February 2006
1180 Intellectual Property Statement
1182 The IETF takes no position regarding the validity or scope of any
1183 Intellectual Property Rights or other rights that might be claimed to
1184 pertain to the implementation or use of the technology described in
1185 this document or the extent to which any license under such rights
1186 might or might not be available; nor does it represent that it has
1187 made any independent effort to identify any such rights. Information
1188 on the procedures with respect to rights in RFC documents can be
1189 found in BCP 78 and BCP 79.
1191 Copies of IPR disclosures made to the IETF Secretariat and any
1192 assurances of licenses to be made available, or the result of an
1193 attempt made to obtain a general license or permission for the use of
1194 such proprietary rights by implementers or users of this
1195 specification can be obtained from the IETF on-line IPR repository at
1196 http://www.ietf.org/ipr.
1198 The IETF invites any interested party to bring to its attention any
1199 copyrights, patents or patent applications, or other proprietary
1200 rights that may cover technology that may be required to implement
1201 this standard. Please address the information to the IETF at
1205 Disclaimer of Validity
1207 This document and the information contained herein are provided on an
1208 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
1209 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
1210 ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
1211 INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
1212 INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
1213 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
1218 Copyright (C) The Internet Society (2006). This document is subject
1219 to the rights, licenses and restrictions contained in BCP 78, and
1220 except as set forth therein, the authors retain all their rights.
1225 Funding for the RFC Editor function is currently provided by the
1231 Larson & Barber Expires August 14, 2006 [Page 22]