Individual Contributor

pfrpfr@gmail.com

Orange

bruno.decraene@orange.com

Strasbourg University

pelsser@unistra.fr

Arrcus, Inc.

keyur@arrcus.com

Cisco Systems

cfilsfil@cisco.com

This draft standardizes a new well-known BGP community GRACEFUL_SHUTDOWN to signal the graceful shutdown of paths. This draft also describes operational procedures which use this community to reduce the amount of traffic lost when BGP peering sessions are about to be shut down deliberately, e.g. for planned maintenance.

Routing changes in BGP can be caused by planned maintenance operations. This document defines a well-known community , called GRACEFUL_SHUTDOWN, for the purpose of reducing the management overhead of gracefully shutting down BGP sessions. The well-known community allows implementers to provide an automated graceful shutdown mechanism that does not require any router reconfiguration at maintenance time. This document discusses operational procedures to be applied in order to reduce or eliminate loss of packets during a maintenance operation. Loss comes from transient lack of reachability during BGP convergence which follows the shutdown of an EBGP peering session between two Autonomous System Border Routers (ASBR). This document presents procedures for the cases where the forwarding plane is impacted by the maintenance, hence when the use of Graceful Restart does not apply. The procedures described in this document can be applied to reduce or avoid packet loss for outbound and inbound traffic flows initially forwarded along the peering link to be shut down. These procedures trigger, in both Autonomous Sytems (AS), rerouting to alternate paths if they exist within the AS, while allowing the use of the old path until alternate ones are learned. This ensures that routers always have a valid route available during the convergence process. The goal of the document is to meet the requirements described in at best, without changing the BGP protocol. Other maintenance cases, such as the shutdown of an IBGP session or the establishement of an EBGP session, are out of scope of this document. For information, they are briefly discussed in . The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [BCP14] when, and only when, they appear in all capitals, as shown here.

graceful shutdown initiator: a router on which the session shutdown is performed for the maintenance. graceful shutdown receiver: a router that has a BGP session, to be shutdown, with the graceful shutdown initiator.

Packets can be lost during the BGP convergence following a manual shutdown of an EBGP session for two reasons. First, some routers can have no path toward an affected prefix, and drop traffic destined to this prefix. This is because alternate paths can be hidden by nodes of an AS. This happens when is not used and the paths are not selected as best by the ASBR that receive them on an EBGP session, or by Route Reflectors that do not propagate them further in the IBGP topology because they do not select them as best. Second, the FIB can be inconsistent between routers within the AS, and packets toward affected prefixes can loop and be dropped unless encapsulation is used within the AS. This document only addresses the first reason.

This section describes configurations and actions to be performed for the graceful shutdown of EBGP peering links. The goal of this procedure is to retain the paths to be shutdown between the peers, but with a lower LOCAL_PREF value, allowing the paths to remain in use while alternate paths are selected and propagated, rather than simply withdrawing the paths. The LOCAL_PREF value SHOULD be lower than any of the alternative paths. The RECOMMENDED value is 0. Note that some alternative techniques with limited applicability are discussed for information in .

On each ASBR supporting the graceful shutdown receiver procedure, an inbound BGP route policy is applied on all EBGP sessions of the ASBR, that: matches the GRACEFUL_SHUTDOWN community. sets the LOCAL_PREF attribute of the paths tagged with the GRACEFUL_SHUTDOWN community to a low value. For information purpose, example of configurations are provided in .

On the graceful shutdown initiator, at maintenance time, the operator: applies an outbound BGP route policy on the EBGP session to be shutdown. This policy tags the paths propagated over the session with the GRACEFUL_SHUTDOWN community. This will trigger the BGP implementation to re-advertise all active routes previously advertised, and tag them with the GRACEFUL_SHUTDOWN community. applies an inbound BGP route policy on the EBGP session to be shutdown. This policy tags the paths received over the session with the GRACEFUL_SHUTDOWN community and sets LOCAL_PREF to a low value. wait for route readvertisement over the EBGP session, and BGP routing convergence on both ASBRs. shutdown the EBGP session, optionally using to communicate the reason of the shutdown. In the case of a shutdown of the whole router, in addition to the graceful shutdown of all EBGP sessions, there is a need to gracefully shutdown the routes originated by this router (e.g, BGP aggregates redistributed from other protocols, including static routes). This can be performed by tagging these routes with the GRACEFUL_SHUTDOWN community and setting LOCAL_PREF to a low value.

BGP Implementers SHOULD provide configuration knobs that utilize the GRACEFUL_SHUTDOWN community to drain BGP neighbors in preparation of impending neighbor shutdown. Implementation details are outside the scope of this document.

The IANA has assigned the community value 0xFFFF0000 to the planned-shut community in the "BGP Well-known Communities" registry. IANA is requested to change the name planned-shut to GRACEFUL_SHUTDOWN and set this document as the reference.

By providing the graceful shutdown service to a neighboring AS, an ISP provides means to this neighbor and possibly its downstream ASes to lower the LOCAL_PREF value assigned to the paths received from this neighbor. The neighbor could abuse the technique and do inbound traffic engineering by declaring some prefixes as undergoing a maintenance so as to switch traffic to another peering link. If this behavior is not tolerated by the ISP, it SHOULD monitor the use of the graceful shutdown community.

The authors wish to thank Olivier Bonaventure, Pradosh Mohapatra, Job Snijders, John Heasley, and Christopher Morrow for their useful comments.

A few alternative techniques have been considered to provide graceful shutdown capabilities but have been rejected due to their limited applicability. This section describes them for possible reference.

The MED attribute of the paths to be avoided can be increased so as to force the routers in the neighboring AS to select other paths. The solution only works if the alternate paths are as good as the initial ones with respect to the LOCAL_PREF value and the AS Path Length value. In the other cases, increasing the MED value will not have an impact on the decision process of the routers in the neighboring AS.

The distance to the BGP NEXT_HOP corresponding to the maintained session can be increased in the IGP so that the old paths will be less preferred during the application of the IGP distance tie-break rule. However, this solution only works for the paths whose alternates are as good as the old paths with respect to their LOCAL_PREF value, their AS Path length, and their MED value. Also, this poisoning cannot be applied when BGP NEXT_HOP self is used as there is no BGP NEXT_HOP specific to the maintained session to poison in the IGP.

This appendix is non-normative. Example routing policy configurations to honor the GRACEFUL_SHUTDOWN well-known BGP community.

For the shutdown of an IBGP session, provided the IBGP topology is viable after the maintenance of the session, i.e, if all BGP speakers of the AS have an IBGP signaling path for all prefixes advertised on this graceful shutdown IBGP session, then the shutdown of an IBGP session does not lead to transient unreachability. As a consequence, no specific graceful shutdown action is required.

We identify two potential causes for transient packet losses upon the establishment of an EBGP session. The first one is local to the startup initiator, the second one is due to the BGP convergence following the injection of new best paths within the IBGP topology.

An ASBR that selects as best a path received over a newly established EBGP session may transiently drop traffic. This can typically happen when the NEXT_HOP attribute differs from the IP address of the EBGP peer, and the receiving ASBR has not yet resolved the MAC address associated with the IP address of that "third party" NEXT_HOP. A BGP speaker implementation MAY avoid such losses by ensuring that "third party" NEXT_HOPs are resolved before installing paths using these in the RIB. Alternatively, the operator (script) MAY ping third party NEXT_HOPs that are expected to be used before establishing the session. By proceeding like this, the MAC addresses associated with these third party NEXT_HOPs are resolved by the startup initiator.

During the establishment of an EBGP session, in some corner cases a router may have no path toward an affected prefix, leading to loss of connectivity. A typical example for such transient unreachability for a given prefix is the following: Let's consider three Route Reflectors (RR): RR1, RR2, RR3. There is a full mesh of IBGP sessions between them. 1. RR1 is initially advertising the current best path to the members of its IBGP RR full-mesh. It propagated that path within its RR full-mesh. RR2 knows only that path toward the prefix. 2. RR3 receives a new best path originated by the startup initiator, being one of its RR clients. RR3 selects it as best, and propagates an UPDATE within its RR full-mesh, i.e., to RR1 and RR2. 3. RR1 receives that path, reruns its decision process, and picks this new path as best. As a result, RR1 withdraws its previously announced best-path on the IBGP sessions of its RR full-mesh. 4. If, for any reason, RR3 processes the withdraw generated in step 3, before processing the update generated in step 2, RR3 transiently suffers from unreachability for the affected prefix. The use of or among the RR of the IBGP full-mesh can solve these corner cases by ensuring that within an AS, the advertisement of a new route is not translated into the withdraw of a former route. Indeed, "best-external" ensures that an ASBR does not withdraw a previously advertised (EBGP) path when it receives an additional, preferred path over an IBGP session. Also, "best-intra-cluster" ensures that a RR does not withdraw a previously advertised (IBGP) path to its non clients (e.g. other RRs in a mesh of RR) when it receives a new, preferred path over an IBGP session.