Tracing Requirements for Generic Tunnels
Requirements for a traceroute tool that can see through tunnels
Bonica, R., Kompella, K., and D. Meyer, "Tracing Requirements for Generic Tunnels", RFC 3609, DOI 10.17487/RFC3609, October 2003, <https://www.rfc-editor.org/info/rfc3609>.
@misc{rfc3609,
series = {Request for Comments},
number = 3609,
howpublished = {RFC 3609},
publisher = {RFC Editor},
doi = {10.17487/RFC3609},
url = {https://www.rfc-editor.org/info/rfc3609},
author = {R. Bonica and K. Kompella and D. Meyer},
title = {{Tracing Requirements for Generic Tunnels}},
pagetotal = 9,
year = 2003,
month = oct,
} RFC text
Contents
Network Working Group R. Bonica
Request for Comments: 3609 MCI
Category: Informational K. Kompella
Juniper Networks
D. Meyer
Sprint
September 2003
Tracing Requirements for Generic Tunnels
Status of this Memo
This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2003). All Rights Reserved.
Abstract
This document specifies requirements for a generic route-tracing
application. It also specifies requirements for a protocol that will
support that application. Network operators will use the generic
route-tracing application to verify proper operation of the IP
forwarding plane. They will also use the application to discover
details regarding tunnels that support IP forwarding.
The generic route-tracing application, specified herein, supports a
superset of the functionality that "traceroute" currently offers.
Like traceroute, the generic route-tracing application can discover
the forwarding path between two interfaces that are contained by an
IP network. Unlike traceroute, this application can reveal details
regarding tunnels that support the IP forwarding path.
1. Introduction
IP networks utilize several tunneling technologies. Although these
tunneling technologies provide operators with many useful features,
they also present management challenges. Network operators require a
generic route-tracing application that they can use to verify the
correct operation of the IP forwarding plane. The generic
route-tracing application must be capable of detecting tunnels and
revealing tunnel details. The application also must be useful in
diagnosing tunnel faults.
Implementors also require a new protocol that will support the
generic-route tracing application. This document specifies
requirements for that protocol. It specifies requirements,
primarily, by detailing the desired capabilities of the generic
route-tracing application. A particular version of generic
route-tracing application may implement some subset of the desired
capabilities. It may also implement a superset of those
capabilities. However, protocol designers are not required to
consider the additional capabilities when designing the new protocol.
This document also specifies a few protocol requirements, stated as
such. These requirements are driven by desired characteristics of
the generic route-tracing application. Whenever a protocol
requirement is stated, it is mapped to the desired characteristic of
the route-tracing application.
2. Review of Existing Functionality
Currently, network operators use "traceroute" to trace through the
forwarding path of an IP network. Section 3.4 of [RFC-2151] provides
a thorough description of traceroute. Although traceroute is very
reliable and very widely deployed, it is deficient with regard to
tunnel tracing.
Depending upon tunnel type, traceroute may display an entire tunnel
as a single IP hop, or it may display the tunnel as a collection of
IP hops, without indicating that they are part of a tunnel.
For example, assume that engineers deploy an IP tunnel in an IP
network. Assume also that they configure the tunnel so that the
ingress router does not copy the TTL value from the inner IP header
to outer IP header. Instead, the ingress router always sets the
outer TTL value to its maximum permitted value. When engineers trace
through the network, traceroute will always display the tunnel as a
single IP hop, hiding all components except the egress interface.
Now assume that engineers deploy an MPLS LSP in an IP network.
Assume also that engineers configure the MPLS LSP so that the ingress
router propagates the TTL value from the IP header to the MPLS
header. When engineers trace through the network, traceroute will
display the LSP as a series of IP hops, without indicating that they
are part of a tunnel.
3. Application Requirements
Network operators require a new route-tracing application. The new
application must support all functionality that traceroute currently
offers. It also must provide enhanced tunnel tracing capabilities.
The following list provides specific requirements for the new
route-tracing application:
1) Support the notion of a security token as part of the tunnel
trace request. The security token identifies the tracer's
privileges in tracing tunnels. Network elements will use this
security token to determine whether or not to return the requested
information to the tracer. In particular, appropriate privileges
are required for items (2), (3), (6), (8), (10), (13), and (14).
Justification: Operators may need to discover network forwarding
details, while concealing those details from unauthorized parties.
2) Support in-line traces. An in-line trace reveals the path
between the host upon which the route-tracing application executes
and any interface in an IP network.
Justification: Operators need to discover how the network would
forward a datagram between any two IP interfaces.
3) Support third-party traces. A third-party trace reveals the
path between any two points in an IP network. The application
that initiates a third-party trace need not execute upon a host or
router that is part of the traced path. Unlike existing solutions
[RFC-2151] [RFC-2925], the application will not rely upon IP
options or require access to the SNMP agent in order to support
third-party traces.
Justification: Operators need to discover how the network would
forward a datagram between any two IP interfaces.
4) Support partial traces through broken paths or tunnels.
Justification: Operators need to identify the root cause of
forwarding plane failures.
5) When tracing through a tunnel, either as part of an in-line
trace or a third-party trace, display the tunnel either as a
single IP hop or in detail. The user's request determines how the
application displays tunnels, subject to the user having
permission to do this.
Justification: As they discover IP forwarding details, operators
may need to reveal or mask tunneling details.
6) When displaying a tunnel in detail, include the tunnel type
(e.g., GRE, MPLS), the tunnel name (if applicable), the tunnel
identifier (if applicable) and tunnel endpoint addresses. Also,
include tunnel components and round trip delay across each
component.
Justification: As they discover IP forwarding details, operators
may need to reveal tunneling details.
7) Support the following tunneling technologies: GRE, MPLS, IPSEC,
GMPLS, IP-in-IP, L2TP. Be easily extensible to support new tunnel
technologies.
Justification: Operators will use the generic route-tracing
application to discover how an IP network forwards datagrams. As
many tunnel types may support the IP network, the generic
route-tracing application must detect and reveal details
concerning multiple tunnel types.
8) Trace through nested, heterogeneous tunnels (e.g., IP-in-IP
over MPLS).
Justification: Operators will use the generic route-tracing
application to discover how an IP network forwards datagrams. As
nested, heterogeneous tunnels may support the IP network, the
generic route-tracing application must detect and reveal details
concerning nested, heterogeneous tunnels.
9) At the users request, trace through the forwarding plane, the
control plane or both.
Justification: Operators need to identify the root cause of
forwarding plane failures. Control plane information is sometimes
useful in determining the cause of forwarding plane failure.
10) Support control plane tracing for all tunnel types. When
tracing through the control plane, the hop ingress device reports
hop details. The hop ingress device is the device that originates
the hop.
Justification: Control plane information is available regarding
all tunnel types.
11) Support tracing through forwarding plane for all tunnel types
that implement TTL decrement (or some similar mechanism). When
tracing through the forwarding plane, the hop egress device
reports hop details. The hop egress device is the device that
terminates the hop.
Justification: Forwarding plane information may not be available
for tunnels that do not support TTL decrement.
12) Support tracing through the forwarding plane for all tunnel
types that implement TTL decrement, regardless of whether the
tunnel engages in TTL propagation. (That is, support tunnel
tracing regardless of whether the TTL value is copied from an
inner header to an outer header at tunnel ingress.)
Justification: Forwarding plane information is always available,
regardless of whether the tunnel engages in TTL propagation.
13) When tracing through the control plane, display the MTU
associated with each interface that forwards datagrams through the
traced path.
Justification: MTU information is sometimes useful in identifying
the root cause of forwarding and control plane failures.
14) When tracing through the forwarding plane, display the MTU
associated with each interface that receives datagrams along the
traced path.
Justification: MTU information is sometimes useful in identifying
the root cause of forwarding and control plane failures.
15) Support partial traces through paths containing devices that
do not provide protocol support for generic route tracing. When
the application encounters such a device, it should inform the
user and attempt to discover details regarding the next interface
downstream.
Justification: The application must provide useful information
even if the supporting protocol is not universally deployed.
4. Protocol Requirements
Implementors require a new protocol that supports the generic
route-tracing application. This protocol reveals the path between
two points in an IP network. When access policy permits, the
protocol also reveals tunnel details.
4.1. Information Requirements
The protocol consists of probes and probe responses. Each probe
elicits exactly one response. Each response represents a hop that
contributes to the path between two interfaces. A hop can be either
a top-level IP hop or lower-level hop that is contained by a tunnel.
Justification: Because the generic route-tracing application must
trace through broken paths, the required protocol must use a separate
response message to deliver details regarding each hop. The protocol
must use a separate probe to elicit each response because the
alternative approach, using the single probe with the IP Router Alert
Option, is unacceptable. Many networks forward datagrams that
specify IP options differently than they would forward datagrams that
do not specify IP options. Therefore, the introduction of IP options
would cause the application to trace a forwarding path other than the
path that its user intended to trace.
4.2. Transport Layer Requirements
UDP should carry all protocol messages to their destinations. Other
transport mechanisms may be considered when protocol details are
specified.
Justification: Because the probe/response scheme described above is
stateless, a stateless transport is required. Candidate transports
included UDP over IP, IP and ICMP. ICMP was disqualified because
carrying MPLS information in an ICMP datagram would constitute a
layer violation. IP was disqualified in order to conserve protocol
identifiers.
4.3. Stateless Protocol
The protocol must be stateless. That is, nodes should not have to
maintain state between successive traceroute messages.
Justification: Statelessness is required to support scaling and to
prevent denial of service attacks.
4.4. Routing Requirements
The device that hosts the route-tracing application must maintain an
IP route to the ingress of the traced path. It must also maintain an
IP route to the ingress of each tunnel for which it is requesting
tunnel details. The device that hosts the tunnel tracing application
need not maintain a route to any other device that supports the
traced path.
All of the devices to which the route-tracing application must
maintain a route must maintain a route back to the route-tracing
application.
In order for the protocol to provide tunnel details, all devices
contained by a tunnel must maintain an IP route to the tunnel
ingress.
Justification: The protocol must be sufficiently robust to operate
when tunnel interior devices do not maintain a route back to the
device that hosts the route tracing application.
5. Security Considerations
A configurable access control policy determines the degree to which
features described herein are delivered. The access control policy
requires user identification and authorization.
The new protocol must not introduce security holes nor consume
excessive resources (e.g., CPU, bandwidth). It also must not be
exploitable by those launching DoS attacks or replaying messages.
6. Informative References
[RFC-2151] Kessler, G. and S. Shepard, "A Primer On Internet and
TCP/IP Tools and Utilities", FYI 30, RFC 2151, June 1997.
[RFC-2925] White, K., "Definitions of Managed Objects for Remote
Ping, Traceroute, and Lookup Operations", RFC 2925,
September 2000.
7. Acknowledgements
Thanks to Randy Bush and Steve Bellovin for their comments.
8. Authors' Addresses
Ronald P. Bonica
MCI
22001 Loudoun County Pkwy
Ashburn, Virginia, 20147
EMail: ronald.p.bonica@mci.com
Kireeti Kompella
Juniper Networks, Inc.
1194 N. Mathilda Ave.
Sunnyvale, California 94089
EMail: kireeti@juniper.net
David Meyer
EMail: dmm@maoz.com
9. Full Copyright Statement
Copyright (C) The Internet Society (2003). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assignees.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS 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.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
AI ELI5
On this page
§1Why Tunnels Need Special Tracing
Tunnels are a powerful way to carry traffic through networks. An IP tunnel wraps IP datagrams inside other datagrams, hiding the inner network. MPLS Label Switched Paths tunnel traffic by adding labels. GRE, IPsec, L2TP, and GMPLS all tunnel traffic for different purposes.
The problem is that traceroute, the standard tool for discovering the path through an IP network, cannot see inside tunnels. Depending on how the tunnel is configured, traceroute might show the tunnel as a single hop (when the tunnel does not copy the TTL from the inner header to the outer header), or it might show many hops without indicating they are inside a tunnel (when TTL is copied). Either way, an operator loses visibility into the tunnel itself.
This memo specifies what a better generic route-tracing application must do: discover the forwarding path, reveal tunnel details when permitted, show tunnel types and endpoints, and help diagnose broken tunnels.
Tunnels hide what is inside them. When your traceroute tool goes through a tunnel, it either shows nothing or shows many stops without saying they are inside a tunnel. You need a tool that says where the tunnel is and what is happening inside it.
§2Application Requirements
The memo lists 15 detailed requirements for the tracing application. Key requirements include:
Security awareness: The application must support a security token that identifies the tracer's privileges. Network elements will use this to decide whether to reveal tunnel details or hide them from unauthorized parties.
In-line and third-party traces: An in-line trace discovers the path from the host running the tool to any interface in the network. A third-party trace discovers the path between any two interfaces, and does not require running the tool on a router in the path. Unlike existing tools, it should not rely on IP options (which change how the network forwards the datagram) or SNMP access.
Tunnel visibility: When tracing through a tunnel, the user can choose to see the tunnel as a single hop or in detail (revealing tunnel type, name, identifier, endpoints, and component details with round-trip delays). The application must support nested tunnels (for example, IP-in-IP over MPLS) and multiple tunnel types: GRE, MPLS, IPsec, GMPLS, IP-in-IP, and L2TP.
Robustness: The tool must work even when some routers do not support the protocol, and provide useful information about partial paths through broken tunnels. It must also display the MTU (Maximum Transmission Unit) at each interface, which is important for diagnosing fragmentation issues.
The tool needs to ask permission to peek inside tunnels, trace from anywhere to anywhere without needing to sit on a router, show you the tunnel details if you have permission, and tell you what went wrong even if the path is broken.
§3Protocol Design Principles
The protocol uses a simple probe-and-response model. The tracing application sends a probe, and each hop in the path responds with details about itself. This is stateless—each hop does not need to remember previous probes—and uses UDP for transport.
Why stateless: Stateless operation prevents denial of service attacks where an attacker floods the network with probes, and allows the protocol to scale to millions of probes without overwhelming individual routers.
Why UDP: The protocol needs a stateless transport. ICMP was rejected because carrying MPLS label information in an ICMP datagram would violate the ISO layer model. Plain IP was rejected to conserve protocol identifiers. UDP sits at the right level and provides the needed statelessness.
Routing requirements: The device running the tool must have IP routes to the tunnel ingressess it is tracing. All devices in the path must have routes back to the tracer so they can send responses. This is a practical constraint that makes the protocol work in real networks.
The protocol asks one question per probe and gets one answer back, without needing to remember anything. This keeps it simple and fast, and uses UDP, which works well for quick questions and answers.
§4Security and Access Control
The application is designed with access control in mind. A configurable policy determines what information each user can see. The policy requires user identification and authorization before granting access to tunnel details or control plane information.
The protocol itself must be resistant to abuse. It must not create security holes, consume excessive resources (CPU or bandwidth), or be exploitable by attackers launching denial of service attacks or replaying old messages. The stateless design helps here: because each message is independent, there is no session state to corrupt or hijack.
The security token mentioned in the application requirements provides a way to encode authorization information in each probe, so that intermediate routers can make access control decisions without needing to maintain per-user state.
Only people with permission can see tunnel details. The protocol should not let attackers crash routers or trick them into answering the same question twice.