idnits 2.17.1 draft-francois-bgp-gshut-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** The document seems to lack a License Notice according IETF Trust Provisions of 28 Dec 2009, Section 6.b.ii or Provisions of 12 Sep 2009 Section 6.b -- however, there's a paragraph with a matching beginning. Boilerplate error? (You're using the IETF Trust Provisions' Section 6.b License Notice from 12 Feb 2009 rather than one of the newer Notices. See https://trustee.ietf.org/license-info/.) Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack separate sections for Informative/Normative References. All references will be assumed normative when checking for downward references. ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 253: '...n of the filters MUST supplant any oth...' RFC 2119 keyword, line 296: '... reconfiguration SHOULD be performed a...' RFC 2119 keyword, line 311: '...tended community, it SHOULD be set non...' RFC 2119 keyword, line 314: '..., this community SHOULD be removed fro...' RFC 2119 keyword, line 316: '... community MAY be removed from the p...' (4 more instances...) Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document seems to contain a disclaimer for pre-RFC5378 work, and may have content which was first submitted before 10 November 2008. The disclaimer is necessary when there are original authors that you have been unable to contact, or if some do not wish to grant the BCP78 rights to the IETF Trust. If you are able to get all authors (current and original) to grant those rights, you can and should remove the disclaimer; otherwise, the disclaimer is needed and you can ignore this comment. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (March 6, 2009) is 5529 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Outdated reference: A later version (-01) exists of draft-marques-idr-best-external-00 == Outdated reference: A later version (-01) exists of draft-decraene-bgp-graceful-shutdown-requirements-00 Summary: 3 errors (**), 0 flaws (~~), 3 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group Pierre Francois 3 Internet-Draft Universite catholique de Louvain 4 Intended status: Informational Bruno Decraene 5 Expires: September 7, 2009 France Telecom 6 Cristel Pelsser 7 NTT Corporation 8 Clarence Filsfils 9 Cisco Systems 10 March 6, 2009 12 Graceful BGP session shutdown 13 draft-francois-bgp-gshut-01 15 Status of this Memo 17 This Internet-Draft is submitted to IETF in full conformance with the 18 provisions of BCP 78 and BCP 79. This document may contain material 19 from IETF Documents or IETF Contributions published or made publicly 20 available before November 10, 2008. The person(s) controlling the 21 copyright in some of this material may not have granted the IETF 22 Trust the right to allow modifications of such material outside the 23 IETF Standards Process. Without obtaining an adequate license from 24 the person(s) controlling the copyright in such materials, this 25 document may not be modified outside the IETF Standards Process, and 26 derivative works of it may not be created outside the IETF Standards 27 Process, except to format it for publication as an RFC or to 28 translate it into languages other than English. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF), its areas, and its working groups. Note that 32 other groups may also distribute working documents as Internet- 33 Drafts. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 The list of current Internet-Drafts can be accessed at 41 http://www.ietf.org/ietf/1id-abstracts.txt. 43 The list of Internet-Draft Shadow Directories can be accessed at 44 http://www.ietf.org/shadow.html. 46 This Internet-Draft will expire on September 7, 2009. 48 Copyright Notice 49 Copyright (c) 2009 IETF Trust and the persons identified as the 50 document authors. All rights reserved. 52 This document is subject to BCP 78 and the IETF Trust's Legal 53 Provisions Relating to IETF Documents in effect on the date of 54 publication of this document (http://trustee.ietf.org/license-info). 55 Please review these documents carefully, as they describe your rights 56 and restrictions with respect to this document. 58 Abstract 60 This draft describes operational procedures aimed at reducing the 61 amount of traffic lost during planned maintenances of routers, 62 involving the shutdown of BGP peering sessions. 64 Table of Contents 66 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 67 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 4 68 3. Packet loss upon manual eBGP session shutdown . . . . . . . . 5 69 4. Practices to avoid packet losses . . . . . . . . . . . . . . . 5 70 4.1. Improving availability of alternate paths . . . . . . . . 6 71 4.2. Graceful shutdown procedures for eBGP sessions . . . . . . 6 72 4.2.1. Outbound traffic . . . . . . . . . . . . . . . . . . . 6 73 4.2.2. Inbound traffic . . . . . . . . . . . . . . . . . . . 7 74 4.3. Graceful shutdown procedures for iBGP sessions . . . . . . 9 75 5. Forwarding modes and forwarding loops . . . . . . . . . . . . 9 76 6. Dealing with Internet policies . . . . . . . . . . . . . . . . 10 77 7. Effect of the g-shut procedure on the convergence . . . . . . 10 78 7.1. Maintenance of an eBGP session . . . . . . . . . . . . . . 10 79 7.1.1. Propagation on the other eBGP sessions of the 80 g-shut initiator . . . . . . . . . . . . . . . . . . . 10 81 7.1.2. Propagation on the other iBGP sessions of the 82 g-shut initiator . . . . . . . . . . . . . . . . . . . 11 83 7.1.3. Propagation of updates in an iBGP full-mesh . . . . . 11 84 7.1.4. Propagation of updates from iBGP to iBGP in a RR 85 hierarchy . . . . . . . . . . . . . . . . . . . . . . 11 86 7.2. Maintenance of an iBGP session . . . . . . . . . . . . . . 12 87 7.3. Applicability of the g-shut procedure . . . . . . . . . . 13 88 7.4. Summary of operations . . . . . . . . . . . . . . . . . . 13 89 7.4.1. Pre-configuration . . . . . . . . . . . . . . . . . . 13 90 7.4.2. Operations at maintenance time . . . . . . . . . . . . 13 91 8. Link Up cases . . . . . . . . . . . . . . . . . . . . . . . . 13 92 8.1. Unreachability local to the ASBR . . . . . . . . . . . . . 13 93 8.2. iBGP convergence . . . . . . . . . . . . . . . . . . . . . 14 94 9. Alternative techniques with limited applicability . . . . . . 15 95 9.1. In-filter reconfiguration . . . . . . . . . . . . . . . . 15 96 9.2. Multi Exit Discriminator tweaking . . . . . . . . . . . . 16 97 9.3. IGP distance Poisoning . . . . . . . . . . . . . . . . . . 16 98 10. IANA considerations . . . . . . . . . . . . . . . . . . . . . 16 99 11. Security Considerations . . . . . . . . . . . . . . . . . . . 17 100 12. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 17 101 13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 17 102 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 17 104 1. Introduction 106 Routing changes in BGP can be caused by planned, manual, maintenance 107 operations. This document discusses operational procedures to be 108 applied in order to reduce or eliminate losses of packets during the 109 maintenance. These losses come from the transient lack of 110 reachability during the BGP convergence following the shutdown of an 111 eBGP peering session between two Autonomous System Border Routers 112 (ASBR). 114 This document presents procedures for the cases where the forwarding 115 plane is impacted by the maintenance, hence when the use of Graceful 116 Restart does not apply. 118 The procedures described in this document can be applied to reduce or 119 avoid packet loss for outbound and inbound traffic flows initially 120 forwarded along the peering link to be shut down. These procedures 121 allow routers to keep using old paths until alternate ones are 122 learned, ensuring that routers always have a valid route available 123 during the convergence process. 125 The goal of the document is to meet the requirements described in 126 [REQS] at best, without changing the BGP protocol or BGP 127 implementations. 129 Still, it explains why reserving a community value for the purpose of 130 BGP session graceful shutdown would reduce the management overhead 131 bound with the solution. It would also allow vendors to provide an 132 automatic graceful shutdown mechanism that does not require any 133 configuration at maintenance time. 135 2. Terminology 137 g-shut initiator : a router on which the session shutdown is 138 performed for the maintenance. 140 g-shut neighbor : a router that peers with the g-shut initiator via 141 (one of) the session(s) to be shut down. 143 Note that for the link-up case, we will refer to these nodes as g-no- 144 shut initiator, and g-no-shut neighbor. 146 Initiator AS : the Autonomous System of the g-shut initiator. 148 Neighbor AS : the Autonomous System of the g-shut neighbor. 150 Affected path / Nominal / pre-convergence path : a BGP path via the 151 peering link(s) undergoing the maintenance. This path will no longer 152 exist after the shutdown. 154 Affected prefix : a prefix initially reached via an affected path. 156 Affected router : a router having an affected prefix. 158 Backup / alternate / post-convergence path : a path toward an 159 affected prefix that will be selected as the best path by an affected 160 router, when the link is shut down and the BGP convergence is 161 completed. 163 Transient alternate path : a path towards an affected prefix that may 164 be transiently selected as best by an affected router during the 165 convergence process but that is not a post-convergence path. 167 Loss of Connectivity (LoC) : the state when a router has no path 168 towards an affected prefix. 170 3. Packet loss upon manual eBGP session shutdown 172 Packets can be lost during a manual shutdown of an eBGP session for 173 two reasons. 175 First, routers involved in the convergence process can transiently 176 lack of paths towards an affected prefix, and drop traffic destined 177 to this prefix. This is because alternate paths can be hidden by 178 nodes of an AS. This happens when the paths are not selected as best 179 by the ASBR that receive them on an eBGP session, or by Route 180 Reflectors that do not propagate them further in the iBGP topology 181 because they do not select them as best. 183 Second, within the AS, routers' FIB can be transiently inconsistent 184 during the BGP convergence and packets towards affected prefixes can 185 loop and be dropped. Note that these loops only happen when ASBR-to- 186 ASBR encapsulation is not used within the AS. 188 This document only addresses the first reason. 190 4. Practices to avoid packet losses 192 This section describes means for an ISP to reduce the transient loss 193 of packets upon a manual shutdown of a BGP session. 195 4.1. Improving availability of alternate paths 197 All solutions that increase the availability of alternate BGP paths 198 in routers performing packet lookups in BGP tables [BestExternal] 199 [AddPath] help in reducing the LoC bound with manual shutdown of eBGP 200 sessions. 202 One solution increasing diversity in such a way that, at any single 203 step of the convergence process following the eBGP session shutdown, 204 a BGP router does not receive a message withdrawing the only path it 205 currently knows for a given NLRI, allows for a simplified g-shut 206 procedure. This simplified procedure would only tackle potential LoC 207 for the inbound traffic. 209 Using advertise-best-external [BestExternal] on ASBRs and RRs helps 210 in avoiding lack of alternate paths in route reflectors upon a 211 convergence. Hence it reduces the LoC duration for the outbound 212 traffic of the ISP upon an eBGP Session shutdown by reducing the iBGP 213 path hunting. 215 Still it does not ensure that BGP routers will always have at least 216 one path towards affected prefixes during the convergence following 217 the event. This property may be verified in future revisions of 218 [BestExternal], notably of its Section 4, hence the current proposal 219 will be updated accordingly. 221 Increasing diversity with [AddPath] might lead to the respect of this 222 property, depending on the path propagation decision process that 223 add-path compliant routers would use. 225 Note that the LoC for the inbound traffic of the maintained router, 226 induced by a lack of alternate path propagation within the iBGP 227 topology of a neighboring AS is not under the control of the operator 228 performing the maintenance, hence the procedure described in 229 Section 4.2.2 should be applied upon the maintenance, even if not 230 required for the outbound traffic. 232 4.2. Graceful shutdown procedures for eBGP sessions 234 This section aims at describing a procedure to be applied to reduce 235 the LoC with readily available BGP features, and without assuming a 236 particular iBGP design in the Initiator and Neighbor ASes. 238 4.2.1. Outbound traffic 240 This section discusses a mean to render the affected paths less 241 desirable by the BGP decision process of affected routers, still 242 allowing these to be used during the convergence while alternate 243 paths are propagated to the affected routers. 245 A decrease of the local-pref value of the affected paths can be 246 issued in order to render the affected paths less preferable, at the 247 highest possible level of the BGP Decision Process. 249 This operation can be performed by reconfiguring the out-filters 250 associated with the iBGP sessions established by the g-shut 251 initiator. 253 The modification of the filters MUST supplant any other rule 254 affecting the local-pref value of the old paths. 256 Compared to using an in-filter of the eBGP session to be shut down, 257 the modification of the out-filters will not let the g-shut initiator 258 switch to another path, as the input to the BGP decision process of 259 that router does not change. As a consequence, the g-shut initiator 260 will not send a withdraw message over its iBGP sessions when it 261 receives an alternate path over an iBGP session. It will however 262 modify the local-pref of the affected paths so that upstream routers 263 will switch to alternate ones. 265 When the actual shutdown of the session is performed, the g-shut 266 initiator will itself switch to the alternate paths. 268 4.2.2. Inbound traffic 270 The solution described for the outbound traffic can be applied at the 271 neighbor AS. This can be done either "manually" or by using a 272 community value dedicated to this task. 274 4.2.2.1. Phone call 276 The operator performing the maintenance of the eBGP session can 277 contact the operator at the other side of the peering link, and let 278 him apply the procedure described above for its own outbound traffic. 280 4.2.2.2. Community tagging 282 A community value (referred to as GSHUT community in this document) 283 can be agreed upon by neighboring ASes. A path tagged with this 284 community must be considered as soon to be affected by a maintenance 285 operation. 287 4.2.2.2.1. Pre-Configuration 289 A g-shut neighbor is pre-configured to set a low local-pref value for 290 the paths received over eBGP sessions which are tagged with the GSHUT 291 community. 293 This rule must supplant any other rule affecting the local-pref value 294 of the paths. 296 This local-pref reconfiguration SHOULD be performed at the out- 297 filters of the iBGP sessions of the g-shut neighbor. That is, the 298 g-shut neighbor does not take into account this low local-pref in its 299 own BGP best path selection. As described in Section 4.2.1 this 300 avoids sending the withdraw messages that can lead to LoC. 302 4.2.2.2.2. Operational action upon maintenance 304 Upon the manual shutdown, the output filter associated with the 305 maintained eBGP session will be modified on the g-shut initiator so 306 as to tag all the paths advertised over the session with the GSHUT 307 community. 309 4.2.2.2.3. Transitivity of the community 311 If the GSHUT community is an extended community, it SHOULD be set non 312 transitive. 314 If a normal community is used, this community SHOULD be removed from 315 the path by the ASBR of the peer receiving it. If not, the GSHUT 316 community MAY be removed from the path by all the ASBRs of the 317 neighboring AS, before propagating the path to other peers. 319 Not propagating the community further in the Internet reduces the 320 amount of BGP churn and avoids rerouting in distant ASes that would 321 also recognize this community value. In other words, it helps 322 concealing the convergence at the maintenance location. 324 There are cases where an interdomain exploration is to be performed 325 to recover the reachability, e.g., in the case of a shutdown in 326 confederations where the alternate paths will be found in another AS 327 of the confederation. In such scenarios, the community value SHOULD 328 be allowed to transit through the confederation but MAY be removed 329 from the paths advertised outside of the confederation. 331 When the local-pref value of a path is conserved upon its propagation 332 from one AS of the confederation to the other, there is no need to 333 have the GSHUT community be propagated throughout that confederation. 335 4.2.2.2.4. Easing the configuration for G-SHUT 337 From a configuration burden viewpoint, it would be much easier to 338 reserve a value for the GSHUT community. 340 First, on the g-shut initiator, an operator would have a single 341 configuration rule to be applied at the maintenance time, which would 342 not depend on the identity of its peer. This would make the 343 maintenance operations less error prone. 345 Second, on the g-shut neighbor, a simple filter related to g-shut can 346 be applied to all iBGP sessions. Additionnaly, this filter doesn't 347 need to be updated each time neighboring ASes are added or removed. 349 4.3. Graceful shutdown procedures for iBGP sessions 351 If the iBGP topology is viable after the maintenance of the session, 352 i.e, if all BGP speakers of the AS have an iBGP signaling path for 353 all prefixes advertised on this g-shut iBGP session, then the 354 shutdown of an iBGP session does not lead to transient 355 unreachability. 357 However, in the case of a shutdown of a router, a reconfiguration of 358 the out-filters of the g-shut initiator should be performed to set a 359 low local-pref value for the paths originated by the g-shut initiator 360 (e.g, BGP aggregates redistributed from other protocols, including 361 static routes). 363 This behavior is equivalent to the recommended behavior for paths 364 "redistributed" from eBGP sessions to iBGP sessions in the case of 365 the shutdown of an ASBR. 367 5. Forwarding modes and forwarding loops 369 If the AS applying the solution does not rely on encapsulation to 370 forward packets from the Ingress Border Router to the Egress Border 371 Router, then transient forwarding loops and consequent packet losses 372 can occur during the convergence process, even if the procedure 373 described above is applied. Hence if zero LoC is required, 374 encapsulation is required between ASBRs of the AS. 376 Using the out-filter reconfiguration avoids the forwarding loops 377 between the g-shut initiator and its directly connected upstream 378 neighbors. Indeed, when this reconfiguration is applied, the g-shut 379 initiator keeps using its own external path and lets the upstream 380 routers converge to the alternate ones. During this phase, no 381 forwarding loops can occur between the g-shut initiator and its 382 upstream neighbors as the g-shut initiator keeps using the affected 383 paths via its eBGP peering links. When all the upstream routers have 384 switched to alternate paths, the transition performed by the g-shut 385 initiator when the session is actually shut down, will be loopfree. 386 Transient forwarding loops between other routers will not be avoided 387 with this procedure. 389 6. Dealing with Internet policies 391 A side gain of the maintenance solution is that it can also reduce 392 the churn implied by a shutdown of an eBGP session. 394 For this, it is recommended to apply the filters modifying the local- 395 pref value of the paths to values strictly lower but as close as 396 possible to the local-pref values of the post-convergence paths. 398 For example, if an eBGP link is shut down between a provider and one 399 of its customers, and another link with this customer remains active, 400 then the value of the local-pref of the old paths SHOULD be decreased 401 to the smallest possible value of the 'customer' local_pref range, 402 minus 1. Thus, routers will not transiently switch to paths received 403 from shared-cost peers or providers, which could lead to the 404 propagation of withdraw messages over eBGP sessions with shared-cost 405 peers and providers. 407 Proceeding like this reduces both BGP churn and traffic shifting as 408 routers will less likely switch to transient paths. 410 In the above example, it also prevents transient unreachabilities in 411 the neighboring AS that are due to the sending of "abrupt" withdraw 412 messages to shared-cost peers and providers. 414 7. Effect of the g-shut procedure on the convergence 416 This section describes the effect of applying the solution. 418 7.1. Maintenance of an eBGP session 420 This section describes the effect of applying the solution for the 421 shutdown of an eBGP session. 423 7.1.1. Propagation on the other eBGP sessions of the g-shut initiator 425 Nothing is propagated on the other eBGP sessions when the out-filters 426 reconfiguration step is applied. The reconfiguration is indeed only 427 defined for its iBGP sessions. 429 The reconfiguration of the iBGP out-filters will trigger the 430 reception of alternate paths at the g-shut initiator. As the eBGP 431 in-filters have not been modified at that step, the old paths are 432 still preferred by the g-shut initiator. 434 7.1.2. Propagation on the other iBGP sessions of the g-shut initiator 436 During the out-filter reconfiguration, path updates are propagated 437 with a reduced local-pref value for the affected paths. As a 438 consequence, Route Reflectors and distant ASBRs select and propagate 439 alternate paths through the iBGP topology as they no longer select 440 the old paths as best. 442 When the shut-down is performed, for each affected prefix, the g-shut 443 initiator propagates on its iBGP sessions: 445 . The alternate path, if the best path was received over an eBGP 446 sessions. 448 . A withdraw, if the best path was received over an iBGP sessions. 450 7.1.3. Propagation of updates in an iBGP full-mesh 452 No transient LoC can occur if a reconfiguration of the iBGP out- 453 filters on the g-shut initiator is performed. 455 7.1.4. Propagation of updates from iBGP to iBGP in a RR hierarchy 457 Upon the reception of the update of a primary path with a lower 458 local-pref value from a client, a Route Reflector RR1 will either 459 propagate the update, or select an alternate path, depending on the 460 fact that the updated primary path is still the best one w.r.t. the 461 state of the Adj-Rib-In of RR1. 463 If the updated primary path is still the best, then the RR will 464 propagate an update for this path to the iBGP neighbors to which it 465 previously advertised the path. Hence it cannot cause transient lack 466 of path in the Adj-Rib-In of its iBGP neighbors. 468 If an alternate path is picked, and this path was also originated by 469 a client of RR1, an update will also be propagated to the same 470 neighbors as the one to which the primary path was initially 471 propagated. Hence it cannot cause transient lack of path in the Adj- 472 Rib-In of its iBGP neighbors. 474 If an alternate path is picked, and this path was received from a 475 member of its Route-Reflector iBGP full-mesh, then a withdraw message 476 is sent. As the alternate path has been sent over each session of 477 the iBGP full-mesh, the propagation of a withdraw for the primary 478 path of RR1 is done to routers that are expected to know the 479 alternate path picked by RR1. 481 The following example describes a situation where some corner case 482 timings could lead to transient unreachability from some members of 483 the iBGP full-mesh. 485 1. A Route Reflector RR1 only knew about the primary path upon 486 the shutdown. 488 2. A member of its RR full-mesh, RR2, propagates an update of 489 the old path with a lower local-pref. 491 3. Another member of its RR full-mesh, RR3 processes the 492 update, selects an alternate path, and propagates an update in 493 the mesh. 495 4. RR2 receives the alternate path, selects it as best, and 496 hence withdraws the updated old path on the iBGP sessions of the 497 mesh. 499 5. If for any reason, RR1 receives and processes the withdraw 500 generated in step 4 before processing the update generated in 501 step 3, RR1 transiently suffers from unreachability for the 502 affected prefix. 504 In such corner cases, the solution improves the iBGP convergence 505 behavior/LoC but does not ensure 0 packet loss, as we cannot define a 506 simple solution relying only on a reconfiguration of the filters of 507 the g-shut initiator. Improving the availability of alternate paths 508 in Route Reflectors, using [BestExternal], or [AddPath], seems to be 509 the most pragmatic solution to these corner cases. 511 The use of [BestExternal] in the iBGP full-mesh between RRs can solve 512 these corner cases by ensuring that within an AS, the advertisement 513 of a new path is not translated into the withdraw of a former path. 515 Indeed, "best-external" ensures that an ASBR does not withdraw a 516 previously advertised (eBGP) path when it receives an additional, 517 preferred path over an iBGP session. Also, "best-intra-cluster" 518 ensures that a RR does not withdraw a previously advertised (iBGP) 519 path to its non clients (e.g. other RRs in a mesh of RR) when it 520 receives a new, preferred path over an iBGP session. 522 7.2. Maintenance of an iBGP session 524 If the shutdown does not temper with the viability of the iBGP 525 topology, the described procedure is sufficient to avoid LoC. 527 7.3. Applicability of the g-shut procedure 529 The applicability of the procedure described in this draft to the 530 cases presented in [REQS] can be shown by combining the effects 531 described in this section. A complete case by case analysis will be 532 provided in the next versions of the draft. 534 7.4. Summary of operations 536 This section summarizes the configurations and actions to be 537 performed to support the g-shut procedure for eBGP peering links. 539 7.4.1. Pre-configuration 541 On each ASBR supporting the g-shut procedure, set-up an out-filter 542 applied on all iBGP sessions of the ASBR, that : 544 . sets the local-pref of the paths tagged with the g-shut community 545 to a low value 547 . removes the g-shut community from the path. 549 7.4.2. Operations at maintenance time 551 On the g-shut initiator : 553 . Apply an in-filter on the maintained BGP session to tag the paths 554 received over the session with the g-shut community. 556 . Apply an out-filter on the maintained BGP session to tag the paths 557 propagated over the session with the g-shut community. 559 . Wait for convergence to happen. 561 . Perform a BGP session shutdown. 563 8. Link Up cases 565 We identify two potential causes for transient packet losses upon an 566 eBGP link up event. The first one is local to the g-shut initiator, 567 the second one is due to the BGP convergence following the injection 568 of new best paths within the iBGP topology. 570 8.1. Unreachability local to the ASBR 572 An ASBR that selects as best a path received over a newly brought up 573 eBGP session may transiently drop traffic. This can typically happen 574 when the nexthop attribute differs from the IP address of the eBGP 575 peer, and the receiving ASBR has not yet resolved the MAC address 576 associated with the IP address of that "third party" nexthop. 578 A BGP speaker implementation could avoid such losses by ensuring that 579 "third party" nexthops are resolved before installing paths using 580 these in the RIB. 582 If the link up event corresponds to an eBGP session that is being 583 manually brought up, over an already up multi-access link, then the 584 operator can ping third party nexthops that are expected to be used 585 before actually bringing the session up, or ping directed broadcast 586 the subnet IP address of the link. By proceeding like this, the MAC 587 addresses associated with these third party nexthops will be resolved 588 by the g-no-shut initiator. 590 8.2. iBGP convergence 592 Similar corner cases as described in Section 7.1.4 for the link down 593 case, can occur during an eBGP link up event. 595 A typical example for such transient unreachability for a given 596 prefix is the following : 598 1. A Route Reflector, RR1, is initially advertising the current 599 best path to the members of its iBGP RR full-mesh. It 600 propagated that path within its RR full-mesh. Another route 601 reflector of the full-mesh, RR2, knows only that path towards 602 the prefix. 603 2. A third Route Reflector of the RR full-mesh, RR3 receives a 604 new best path orginated by the "g-no-shut" initiator, being one 605 of its RR clients. RR3 selects it as best, and propagates an 606 UPDATE within its RR full-mesh, i.e., to RR1 and RR2. 607 3. RR1 receives that path, reruns its decision process, and 608 picks this new path as best. As a result, RR1 withdraws its 609 previously announced best-path on the iBGP sessions of its RR 610 full-mesh. 611 4. If, for any reason, RR3 processes the withdraw generated in 612 step 3, before processing the update generated in step 2, RR3 613 transiently suffers from unreachability for the affected prefix. 615 The use of [BestExternal] among the RR of the iBGP full-mesh can 616 solve these corner cases by ensuring that within an AS, the 617 advertisement of a new route is not translated into the withdraw of a 618 former route. 620 Indeed, "best-external" ensures that an ASBR does not withdraw a 621 previously advertised (eBGP) path when it receives an additional, 622 preferred path over an iBGP session. Also, "best-intra-cluster" 623 ensures that a RR does not withdraw a previously advertised (iBGP) 624 path to its non clients (e.g. other RRs in a mesh of RR) when it 625 receives a new, preferred path over an iBGP session. 627 9. Alternative techniques with limited applicability 629 A few alternative techniques have been considered to provide g-shut 630 capabilities but have been rejected due to their limited 631 applicability. This section describe them for possible reference. 633 9.1. In-filter reconfiguration 635 An In-filter reconfiguration on the eBGP session undergoing the 636 maintenance could be performed instead of out-filter reconfigurations 637 on the iBGP sessions of the g-shut initiator. 639 Upon the application of the maintenance procedure, if the g-shut 640 initiator has an alternate path in its Adj-Rib-In, it will switch to 641 it directly. 643 If this new path was advertised by an eBGP neighbor of the g-shut 644 initiator, the g-shut initiator will send a BGP Path Update message 645 advertising the new path over its iBGP and eBGP sessions. 647 If this new path was received over an iBGP session, the g-shut 648 initiator will select that path and withdraw the previously 649 advertised path over its non-client iBGP sessions. There can be iBGP 650 topologies where the iBGP peers of the g-shut initiator do not know 651 an alternate path, and hence may drop traffic. 653 Also, applying an In-filter reconfiguration on the eBGP session 654 undergoing the maintenance may lead to transient LoC, in full-mesh 655 iBGP topologies if 657 a. An ASBR of the initiator AS, ASBR1 did not initially select 658 its own external path as best, and 660 b. An ASBR of the initiator AS, ASBR2 advertises a new path 661 along its iBGP sessions upon the reception of ASBR1's update 662 following the in-filter reconfiguration on the g-shut initiator, 663 and 665 c. ASBR1 receives the update message, runs its Decision Process 666 and hence withdraws its external path after having selected 667 ASBR2's path as best, and 668 d. An impacted router of the AS processes the withdraw of ASBR1 669 before processing the update from ASBR2. 671 Applying a reconfiguration of the out-filters prevents such transient 672 unreachabilities. 674 Indeed, when the g-shut initiator propagates an update of the old 675 path first, the withdraw from ASBR2 does not trigger unreachability 676 in other nodes, as the old path is still available. Indeed, even 677 though it receives alternate paths, the g-shut initiator keeps using 678 its old path as best as the in-filter of the maintained eBGP session 679 has not been modified yet. 681 Applying the out-filter reconfiguration also prevents packet loops 682 between the g-shut initiator and its direct neighbors when 683 encapsulation is not used between the ASBRs of the AS. 685 9.2. Multi Exit Discriminator tweaking 687 The MED attribute of the paths to be avoided can be increased so as 688 to force the routers in the neighboring AS to select other paths. 690 The solution only works if the alternate paths are as good as the 691 initial ones with respect to the Local-Pref value and the AS Path 692 Length value. In the other cases, increasing the MED value will not 693 have an impact on the decision process of the routers in the 694 neighboring AS. 696 9.3. IGP distance Poisoning 698 The distance to the BGP nexthop corresponding to the maintained 699 session can be increased in the IGP so that the old paths will be 700 less preferred during the application of the IGP distance tie-break 701 rule. However, this solution only works for the paths whose 702 alternates are as good as the old paths with respect to their Local- 703 Pref value, their AS Path length, and their MED value. 705 Also, this poisoning cannot be applied when nexthop self is used as 706 there is no nexthop specific to the maintained session to poison in 707 the IGP. 709 10. IANA considerations 711 Applying the g-shut procedure is rendered much easier with a reserved 712 g-shut community value. Hence this draft suggests to reserve a 713 community value, e.g., 0xFFFF0000, for this purpose. 715 11. Security Considerations 717 By providing the g-shut service to a neighboring AS, an ISP provides 718 means to this neighbor to lower the local-pref value assigned to the 719 paths received from this neighbor. 721 The neighbor could abuse the technique and do inbound traffic 722 engineering by declaring some prefixes as undergoing a maintenance so 723 as to switch traffic to another peering link. 725 If this behavior is not tolerated by the ISP, it SHOULD monitor the 726 use of the g-shut community by this neighbor. 728 12. Acknowledgments 730 The authors wish to thank Olivier Bonaventure and Pradosh Mohapatra 731 for their useful comments on this work. 733 13. References 735 [AddPath] D. Walton, A. Retana, and E. Chen, "Advertisement of 736 Multiple Paths in BGP", draft-walton-bgp-add-paths-06.txt 737 (work in progress). 739 [BestExternal] 740 Marques, P., Fernando, R., Chen, E., and P. Mohapatra, 741 "Advertisement of the best-external route to IBGP", 742 draft-marques-idr-best-external-00.txt, July 2008. 744 [REQS] Decraene, B., Francois, P., Pelsser, C., and Z. Ahmad, 745 "Requirements for the graceful shutdown of BGP sessions", 746 draft-decraene-bgp-graceful-shutdown-requirements-00.txt 747 , December 2007. 749 Authors' Addresses 751 Pierre Francois 752 Universite catholique de Louvain 753 Place Ste Barbe, 2 754 Louvain-la-Neuve 1348 755 BE 757 Email: pierre.francois@uclouvain.be 758 URI: http://inl.info.ucl.ac.be/pfr 759 Bruno Decraene 760 France Telecom 761 38-40 rue du General Leclerc 762 92794 Issi Moulineaux cedex 9 763 FR 765 Email: bruno.decraene@orange-ftgroup.com 767 Cristel Pelsser 768 NTT Corporation 769 9-11, Midori-Cho 3 Chrome 770 Musashino-Shi, Tokyo 180-8585 771 JP 773 Email: pelsser.cristel@lab.ntt.co.jp 775 Clarence Filsfils 776 Cisco Systems 777 De kleetlaan 6a 778 Diegem 1831 779 BE 781 Email: cfilsfil@cisco.com