DNS Operations Root Server System Advisory Committee Internet-Draft B.Manning Ed. Intended status: Informational Expires: 2012 June 18 27 December 2011 Root Name Server Operational Requirements draft-rssac-dnsop-rfc2870bis-00 Status of this Memo Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on June 18, 2012. IETF Legal Notices Copyright (c) 2011 IETF Trust and the persons identified as the document authors. All rights reserved. This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. 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. "This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights." "This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE." "The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79." "Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr." "The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org." Abstract This memo replaces RFC 2870. As the Internet has become critical to the world's social and economic infrastructure, attention has rightly focused on the correct, safe, reliable, and secure operation of the Internet infrastructure itself. The root domain and its name servers are a crucial part of that technical infrastructure. The primary focus of this document is to provide guidelines for secure, stable, and resilient name service for the root zone. Additionally it will look into some specifics for the operation of the name servers. Other operators of critical name servers--such as those for generic top-level domains (gTLDs), country code top-level domains (ccTLDs) and other major zones--may also find it useful. 1. Background The resolution of domain names on the Internet is critically dependent on the proper, safe and secure operation of the root domain name servers. The Internet Assigned Numbers Authority (IANA) publishes the "root hints" file (http://www.internic.net/zones/named.root), which lists the names and IP addresses of the thirteen root name servers that are the focus of this document. These servers are operated by twelve different organizations, which are listed at http://root-servers.org/. This document provides formal guidelines for the operation of these root name servers. 1.1 ICANN has appointed a Root Server System Advisory Committee (RSSAC) (http://www.icann.org/committees/dns-root/) to give technical and operational advice to the ICANN board. ICANN and RSSAC look to the IETF to provide engineering standards. 1.2 The root servers serve the root (".") and ROOT-SERVERS.NET zones. For legacy reasons, some of the root servers have also served other important zones, namely ARPA, IN-ADDR.ARPA. In the future, the root servers may not serve any additional zones or they may be asked to serve other zones. 1.3 The root servers are neither involved with nor dependent upon any WHOIS [5] data. 1.4 The domain name system has proven to be sufficiently robust that that the (presumably temporary) loss of most of the root server instances should not significantly affect operation of the Internet. 1.5 Experience has shown that the Internet is quite vulnerable to incorrect data in the root zone or top-level domain (TLD) zones. Providing authentication, validation, and security of these data and the communications channels they use are protected. Most of the root operators now provision more than one physical or logical host to provide root name service. In such configurations, incoming DNS queries to the IP address providing root name service are distributed to multiple hosts using a variety of techniques, including layer-three load balancing devices, equal-cost multipath routing and anycast routing. (Anycast is discussed further in Section 4.) For simplicity, this document uses the term "server" to refer to however many hosts a root operator provisions to provide root name service at a single address per address family. The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [1]. 2. The Servers Themselves The following are requirements for the technical details of the root servers themselves: 2.1 It would be short-sighted of this document to specify particular hardware, operating systems, or name serving software. Variations in these areas would actually add overall robustness. 2.2 Each server MUST run software which correctly implements the IETF standards for the DNS, currently RFC1035 [2], RFC2181 [3] and RFC4035[14]. While there are no accepted, formal test suites for standards compliance, the maintainers of software used on root servers are expected to take all reasonable actions to conform to the IETF's then-current documented expectations. 2.3 At any time, each server MUST be able to handle a load of requests for root data which is at least ten times the measured average number of requests on that server in then-current normal conditions. This capacity is usually expressed in queries per second. This requirement is intended to ensure continued operation of root services in the event of extreme scenarios, such as distributed denial-of-service (DDoS) attacks and other operational anomalies. 2.4 Each root server should have sufficient connectivity to the Internet to support the bandwidth needs of the above requirement. Connectivity to the Internet SHOULD be as diverse as possible. Root servers SHOULD have mechanisms in place to accept IP connectivity to the root server from any Internet provider delivering connectivity at their own cost. 2.5 Servers MUST provide authoritative responses only from the zones they serve. The servers MUST disable recursive lookup, forwarding or any other function that might allow them to provide cached answers. These restrictions help prevent undue load on the root servers and reduce the chance of their caching incorrect data. 2.6 Root servers MUST answer queries from any Internet host, i.e. root servers may not block root name resolution from any valid IP address, except in the case of queries causing operational problems, in which case the blocking SHOULD last only as long as the problem, and be as specific as reasonably possible. 2.7 Root servers SHOULD NOT answer AXFR, or other zone transfer, queries from clients other than other root servers. This restriction is intended to, among other things, prevent unnecessary load on the root servers from both well-meaning but misguided operators configuring their recursive name servers to secondary the root zone, loading from one of the root servers, and from malicious denial-of-service attacks. The root servers MAY distribute the root zone via FTP, HTTP or other access methods on other less-critical servers. 2.8 Servers MUST generate checksums when sending UDP datagrams and MUST verify checksums when receiving UDP datagrams containing a non-zero checksum. 3. Security Considerations The servers need both physical and protocol security as well as unambiguous authentication of their responses. 3.1. Physical Security Physical security MUST be ensured in a manner expected of data centers critical to a major enterprise. 3.1.1 Whether or not the overall site in which a root server is located has access control, the specific area in which the root server is located MUST have positive access control, i.e. the number of individuals permitted access to the area MUST be limited, controlled, and recorded. At a minimum, control measures SHOULD be either mechanical or electronic locks. Physical security MAY be enhanced by the use of intrusion detection and motion sensors, multiple serial access points, security personnel, etc. 3.1.2 Power continuity for at least 24 hours MUST be assured, whether through on-site batteries, on-site power generation or some combination thereof. This backup power source MUST supply the server itself, as well as the infrastructure necessary to connect the server to the Internet. There MUST be procedures that ensure the power fallback mechanisms and supplies are tested no less frequently than the specifications and recommendations of the manufacturer. 3.1.3 Fire detection and/or retardation MUST be provided. 3.1.4 Provision MUST be made for rapid return to operation after a system outage. Such provision SHOULD involve backup of system software and configuration but SHOULD also involve backup hardware which is pre-configured and ready to take over operation, which MAY require manual procedures. 3.2. Network Security Network security should be of the level provided for critical infrastructure of a major commercial enterprise. 3.2.1 The root servers themselves MUST NOT provide services other than DNS name service, secure shell, and network time, e.g. protocols such as HTTP, Telnet, rlogin, FTP, routing daemons, etc. The only login accounts permitted should be for the server administrator(s). Servers MUST have a secure mechanism for remote administrative access and maintenance. Failures happen; there will be times when something breaks badly enough that administrators will need to connect remotely. Remote logins MUST be protected by a secure means that is strongly authenticated and encrypted, and sites from which remote login is allowed MUST be protected and hardened. Remote logins SHOULD be restricted to a list of discrete IP addresses or address ranges. 3.2.2 Root name servers SHOULD NOT trust other hosts, except secondary servers trusting the primary server, for matters of network time, authentication, encryption keys, or other access or security information. If a root operator uses Kerberos authentication to manage access to the root server, then the associated Kerberos key server MUST be protected with the same prudence as the root server itself. This applies to all related services which are trusted in any manner. 3.2.3 The LAN segment on which a root server is located SHOULD NOT also have other Internet-reachable hosts. Secure monitoring hosts that passively monitor network traffic to and from the root server MAY be placed on the same LAN segment. LAN segments SHOULD be switched or routed so there is no possibility of masquerading. 3.2.4 The LAN segment(s) on which a root server is located SHOULD be separately firewalled or packet filtered to prohibit network access to any port other than those needed for name service. 3.2.5 The root servers SHOULD have their clocks synchronized via NTP [6], SNTP [7] or similar mechanisms, in as secure manner as possible. For this purpose, servers and their associated firewalls SHOULD allow the root servers to be NTP clients. Root servers MUST NOT act as NTP peers or servers. 3.2.6 All attempts at intrusion or other compromise SHOULD be logged, and all such logs from all root servers SHOULD be analysed by a cooperative security team communicating with all server operators to look for patterns, serious attempts, etc. Servers SHOULD log in UTC to facilitate log comparison. 3.2.7 Server logging SHOULD be to separate hosts which SHOULD be protected similarly to the root servers themselves. 3.2.8 The server SHOULD be protected from attacks based on source routing. The server MUST NOT rely on address- or name-based authentication. 3.2.9 The network on which the server is located SHOULD have in- addr.arpa service. 3.3. Protocol Authentication and Security Protocol authentication and security are required to ensure that data presented by the root servers are those created by those authorized to maintain the root zone data. 3.3.1 The root zone SHOULD be signed by IANA in accordance with DNSSEC ([8], [9] and [10]). 3.3.2 Once the root zone is DNSSEC-signed, the root servers MUST be DNSSEC-capable so that queries may be authenticated by clients with security and authentication concerns. 3.3.3 Transfer of the root zone between root servers MUST be authenticated and be as secure as reasonably possible. Out of band security validation of updates MUST be supported. Servers MUST use TSIG [4] to authenticate root zones received from other servers. 3.3.4 A "distribution" server, which only allows access by the authorized secondary root servers, MAY be used. 3.3.5 Root zone updates SHOULD only progress after a number of heuristic checks designed to detect erroneous updates have been passed. In case the update fails the tests, human intervention MUST be requested. 3.3.6 Root zone updates SHOULD normally be effective no later than one fourth of the time between root zone updates from notification of the root server operator. 3.3.7 A special procedure for emergency updates SHOULD be defined. Updates initiated by the emergency procedure SHOULD be made no later than 12 hours after notification. 3.3.8 In the event of a critical network failure, each root server MUST have a method to update the root zone data via a medium which is delivered through an alternative, non-network, path. In the event that the root server cannot serve current data, it MUST cease offering DNS service. See also Paragraph 4.3. 3.3.9 Each root MUST keep global statistics on the amount and types of queries received/answered on a daily basis. These statistics SHOULD be made available to RSSAC and RSSAC- sponsored researchers to help determine how to better deploy these machines more efficiently across the Internet. Each root MAY collect data snapshots to help determine data points such as DNS query storms, significant implementation bugs, etc. 3.3.10 Each root operator MUST monitor its server to detect operational problems, such as a host being down, a host not running a name server, a name server not returning authoritative answers for the root zone, etc. Servers MAY also be monitored from an external network. These monitoring processes SHOULD be automated. If problems are detected, the appropriate staff MUST be notified to troubleshoot and remedy the problem. 4. Anycast and Network Considerations The root zone has been available via shared unicast as described in RFC 3258 [11] from several of the authoritative root name servers since 2002. This technique is now commonly referred to as "anycast", despite its differences from the definition of that term in RFC 1546 [12]. Anycast has proven to be a successful method for increasing the capacity and geographic distribution of the root server system. For a root server to offer service successfully using anycast, several best practices should be followed. 4.1 A root server SHOULD be anycast using IPv4 address space that is a /24 from legacy allocations. Using any prefix longer than a /20 that is not from the swamp risks reachability problems because of filtering by ISPs who won't route such address space. IPv6 address space SHOULD be at least a /48. 4.2 Each of a root server's anycast instances MAY be sourced from a consistent origin autonomous system (AS). That is, the BGP routing announcement for all instances of a given root server's service address (i.e., the IPv4 address corresponding to the RDATA of that root name server's NS record) should have the same origin AS. 4.3 Each anycast instance of a root server MUST withdraw its BGP routing announcement upon service failure. Each anycast instance MUST monitor its ability to provide root name service and withdraw its route if it detects itself unable to provide service. Such monitoring SHOULD be automatic and not dependent on a human noticing a service failure. 4.4 Anycast instances SHOULD NOT use the BGP MULTI_EXIT_DISC (MED) attribute because of possible inter-domain routing (IDR) oscillation in networks using route reflectors or AS confederations. Suggested better alternatives are BGP origin code, altering AS path length (i.e., prepending), adjusting local preference and communities. Specific route oscillation scenarios and mitigations are described in detail in RFC 3345 [13]. 4.5 Some root operators provision different kinds of anycast instances for a given root server. Some instances are designated to be local to particular autonomous systems and thus advertise their routes with the BGP no-export community attribute. Other instances are designated "global" and reachable from anywhere on the Internet; these instances do not advertise with no-export. In the event of this configuration, the local and global instances SHOULD NOT advertise routes with the same prefix length. The global instances SHOULD advertise a shorter covering prefix. Failure to advertise a shorter covering prefix from the global instances can result in unreachability in certain scenarios. For example, consider AS A with a local root server anycast instance (i.e., advertising its route with the no-export attribute) announced as prefix P1. AS A prefers its local route to P1 over the other paths to P1 it may have received (corresponding to any global nodes). Now consider AS B that peers only with AS A. Since the local instance of prefix P1 is the best path and is marked no-export, AS A does not send this prefix to AS B. AS B thus has no route to prefix P1 and cannot reach any instances of this root server. Instead, consider if the root server operator advertised a shorter covering prefix P2 for its global instances. In the scenario above, AS A would send prefix P2 to AS B, making any of this root server's global instances reachable from AS B. For IPv4, if the root server prefix is a globally routable /24, and the operator does not have the necessary adjacent address space to aggregate and advertise a shorter prefix, the /24 itself SHOULD be advertised globally and a longer prefix (i.e., /25 or longer) designated local-only. Such a longer local-only prefix will not typically be passed across peering boundaries, which eliminates the need to tag this prefix as no-export. 4.6 There SHOULD NOT be more than three root servers (including anycast instances) in close physical or topological proximity. Multiple root servers near each other introduces the possibility of undesirable fate sharing, since the impact of any issue that could cause poor performance or an outage is magnified by the presence of multiple root servers in one place. For example, consider an AS hosting multiple root servers that suffers a catastrophic failure that drastically reduces performance but does not sever connectivity with the rest of the Internet. In such a situation, there could be sufficient connectivity to continue announcing routes but not to reliably deliver service traffic. If the routes to the root servers located in this AS are sufficiently attractive, some portion of the traffic destined for those root servers could be siphoned off to this AS and then dropped because of insufficient capacity. Obviously, the more root servers located in the AS, the more significant and problematic the resulting unreachability. This scenario is worse for a unicast root server because the server's entire traffic load is affected: there are no other anycast instances to handle that server's query load. 5. Communications Communications and coordination between root server operators and between the operators and IANA and ICANN are necessary. 5.1 Planned outages and other down times SHOULD be coordinated between root server operators to ensure that a significant number of the root servers are not all down at the same time. Announcement of planned outages also keeps other operators from wasting time wondering about any anomalies. 5.2 Root server operators SHOULD coordinate backup timing so that many servers are not off-line being backed up at the same time. Backups SHOULD be frequently transferred off site. 5.3 Root server operators SHOULD exchange log files, particularly as they relate to security, loading, and other significant events. This MAY be through a central log coordination point, or MAY be informal. 5.4 Statistics as they concern usage rates, loading, and resource utilization SHOULD be exchanged between operators, and SHOULD be reported to IANA for planning and reporting purposes. 5.5 Root name server administrative personnel MUST be available to provide service 24 hours a day, 7 days per week. On call personnel MAY be used to provide this service outside of normal working hours. 6. IANA considerations There are no new IANA considerations introduced by this memo. 7. Internationalization considerations There are no new internationalization considerations introduced by this memo. 8. References 8.1. Normative References [1] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [2] Mockapetris, P., "Domain names - implementation and specification", STD 13, RFC 1035, November 1987. [3] Elz, R. and R. Bush, "Clarifications to the DNS Specification", RFC 2181, July 1997. [4] Vixie, P., Gudmundsson, O., Eastlake, D., and B. Wellington, "Secret Key Transaction Authentication for DNS (TSIG)", RFC 2845, May 2000. 8.2. Informative References [5] Daigle, L., "WHOIS Protocol Specification", RFC 3912, September 2004. [6] Mills, D., "Network Time Protocol (Version 3) Specification, Implementation", RFC 1305, March 1992. [7] Mills, D., "Simple Network Time Protocol (SNTP) Version 4 for IPv4, IPv6 and OSI", RFC 4330, January 2006. [8] Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, "DNS Security Introduction and Requirements", RFC 4033, March 2005. [9] Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, "Resource Records for the DNS Security Extensions", RFC 4034, March 2005. [10] Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, "Protocol Modifications for the DNS Security Extensions", RFC 4035, March 2005. [11] Hardie, T., "Distributing Authoritative Name Servers via Shared Unicast Addresses", RFC 3258, April 2002. [12] Partridge, C., Mendez, T., and W. Milliken, "Host Anycasting Service", RFC 1546, November 1993. [13] McPherson, D., Gill, V., Walton, D., and A. Retana, "Border Gateway Protocol (BGP) Persistent Route Oscillation Condition", RFC 3345, August 2002. [14] R. Arends, R. Austein, M. Larson, D. Massey, S. Rose, "Protocol Modifications for the DNS Security Extensions", RFC 4035, March 2005 Authors' Addresses Root Server System Advisory Committee (RSSAC) Bill Manning, Editor PO Box 12317 Marina del Rey, CA 90295 USA