idnits 2.17.1 draft-ietf-nfsv4-migration-issues-15.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. 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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 2211 has weird spacing: '... pieces of th...' -- The document date (May 19, 2018) is 2166 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 5661 (Obsoleted by RFC 8881) == Outdated reference: A later version (-05) exists of draft-ietf-nfsv4-mv0-trunking-update-00 == Outdated reference: A later version (-04) exists of draft-ietf-nfsv4-mv1-msns-update-00 -- Obsolete informational reference (is this intentional?): RFC 3530 (Obsoleted by RFC 7530) Summary: 1 error (**), 0 flaws (~~), 4 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 NFSv4 D. Noveck, Ed. 3 Internet-Draft NetApp 4 Intended status: Informational P. Shivam 5 Expires: November 20, 2018 IBM 6 C. Lever 7 B. Baker 8 ORACLE 9 May 19, 2018 11 NFSv4 Migration and Trunking: Implementation and Specification Issues 12 draft-ietf-nfsv4-migration-issues-15 14 Abstract 16 This document discusses a range of implementation and specification 17 issues concerning features related to the use of location-related 18 attributes in NFSv4. These include migration, which transfers 19 responsibility for a file system from one server to another, and 20 trunking which deals with the discovery and control of the set of 21 server endpoints to use to access a file system. The focus of the 22 discussion, which relates to multiple minor versions, is on defining 23 the appropriate clarifications and corrections for existing 24 specifications. 26 Status of This Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at https://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on November 20, 2018. 43 Copyright Notice 45 Copyright (c) 2018 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (https://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 Table of Contents 60 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 61 2. Language . . . . . . . . . . . . . . . . . . . . . . . . . . 4 62 2.1. Requirements Language . . . . . . . . . . . . . . . . . . 4 63 2.2. Use of Normative Terms . . . . . . . . . . . . . . . . . 4 64 2.3. Terminology Used in this Document . . . . . . . . . . . . 5 65 3. Issues that Apply to Multiple Versions and their Resolution . 7 66 3.1. Issue Summary . . . . . . . . . . . . . . . . . . . . . . 7 67 3.2. Resolution of Multi-Version Issues . . . . . . . . . . . 8 68 3.2.1. Providing Trunking Discovery . . . . . . . . . . . . 9 69 3.2.2. Interaction of Trunking and Migration . . . . . . . . 10 70 3.2.3. Dealing with Multiple Connection Types . . . . . . . 12 71 4. NFSv4.0 Issues . . . . . . . . . . . . . . . . . . . . . . . 13 72 4.1. Core NFSv4.0 Migration Issues . . . . . . . . . . . . . . 13 73 4.2. Resolution of Core Migration Protocol Difficulties in 74 NFSv4.0 . . . . . . . . . . . . . . . . . . . . . . . . . 14 75 4.3. Additional NFSv4.0 Issues . . . . . . . . . . . . . . . . 15 76 4.4. Resolution of Additional NFSv4.0 Issues . . . . . . . . . 15 77 4.4.1. Resolution of NFSv4.0 Issues with Multiple Connection 78 Types . . . . . . . . . . . . . . . . . . . . . . . . 17 79 5. Issues for NFSv4.1 and Beyond . . . . . . . . . . . . . . . . 18 80 5.1. Issues to Address for NFSv4.1 . . . . . . . . . . . . . . 18 81 5.1.1. Addressing State Merger in NFSv4.1 . . . . . . . . . 19 82 5.1.2. Addressing pNFS Relationship with Migration . . . . . 20 83 5.1.3. Addressing Server_owner Changes in NFSv4.1 . . . . . 20 84 5.1.4. Addressing Confirmation Status of Migrated 85 Client IDs in NFSv4.1 . . . . . . . . . . . . . . . . 21 86 5.1.5. Addressing Changes in Trunking Configuration . . . . 23 87 5.1.6. Addressing Session Migration in NFSv4.1 . . . . . . . 23 88 5.1.7. Dealing with Multiple Connection Types in NFSv4.1 . . 23 89 5.2. Possible Resolutions for NFSv4.1 Protocol Issues . . . . 24 90 5.2.1. Client ID Confirmation Issues . . . . . . . . . . . . 25 91 5.2.2. Dealing with Multiple Location Entries . . . . . . . 26 92 5.2.3. Migration and pNFS . . . . . . . . . . . . . . . . . 28 93 5.3. Defining Server Responsibilities for NFSv4.1 . . . . . . 29 94 5.3.1. Server Responsibilities in Effecting Transparent 95 State Migration . . . . . . . . . . . . . . . . . . . 29 97 5.3.2. Synchronizing Session Transfer . . . . . . . . . . . 30 98 5.4. Defining Client Responsibilities for NFSv4.1 . . . . . . 33 99 5.4.1. Client Recovery from Migration Events . . . . . . . . 33 100 5.4.2. The Migration Discovery Process . . . . . . . . . . . 35 101 5.4.3. Overview of Client Response to NFS4ERR_MOVED . . . . 37 102 5.4.4. Obtaining Access to Sessions and State after 103 Migration . . . . . . . . . . . . . . . . . . . . . . 39 104 5.4.5. Obtaining Access to Sessions and State after Network 105 Address Transfer . . . . . . . . . . . . . . . . . . 40 106 5.5. Resolution of NFSv4.1 Issues . . . . . . . . . . . . . . 41 107 5.6. Potential Protocol Extensions . . . . . . . . . . . . . . 43 108 6. Evolution of Issue Handling . . . . . . . . . . . . . . . . . 44 109 6.1. History of this Document . . . . . . . . . . . . . . . . 44 110 6.2. Further Work Needed . . . . . . . . . . . . . . . . . . . 46 111 7. Security Considerations . . . . . . . . . . . . . . . . . . . 49 112 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 50 113 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 50 114 9.1. Normative References . . . . . . . . . . . . . . . . . . 50 115 9.2. Informative References . . . . . . . . . . . . . . . . . 51 116 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 52 117 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 52 119 1. Introduction 121 This is an informational document that discusses a number of related 122 issues in multiple versions of NFSv4. 124 Many of these relate to the migration feature of NFSv4, which 125 provides for moving responsibility for a single filesystem from one 126 server to another, without disruption to clients. A number of 127 problems in the specification of this feature in NFSv4.0 were 128 resolved by the publication of [RFC7931], which added trunking 129 detection to NFSV4.0. However, NFSv4.0 remains without an 130 appropriate discussion of trunking discovery, which has many 131 important connections with migration. As a result, NFSv4.0 requires 132 clarification of how the client is to respond to changes in the 133 trunking arrangements to use, both when migration occurs and when it 134 does not. 136 In addition, there are specification issues to be resolved with 137 regard to the NFSv4.1 version of these features which are discussed 138 in this document. 140 All of the issues discussed relate to the handling and interpretation 141 of the location-related attributes fs_locations and fs_locations_info 142 and to the proper client and server handling of changes in the values 143 of these attributes 144 These issues are all related to the protocol features for effecting 145 file system migration, or to trunking discovery but it is not 146 possible to treat each of these features in isolation. These 147 features are inherently linked because migration needs to deal with 148 the possibility of multiple server addresses in location attributes 149 and because location attributes, which provide trunking-related 150 information, may change, which might or might not involve migration. 152 2. Language 154 2.1. Requirements Language 156 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 157 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 158 document are to be interpreted as described in [RFC2119]. 160 2.2. Use of Normative Terms 162 This document, which deals with existing issues/problems in 163 standards-track documents, is in the informational category, and 164 while the facts it reports may have normative implications, any such 165 normative significance reflects the readers' preferences. For 166 example, we may report that the existing definition of migration for 167 NFSv4.1 does not properly describe how migrating state is to be 168 merged with existing state for the destination server. While it is 169 to be expected that client and server implementers will judge this to 170 be a situation that it would be appropriate to resolve, the judgment 171 as to how pressing this issue should be considered is a judgment for 172 the reader, and eventually the nfsv4 working group to make. 174 We do explore possible ways in which such issues can be dealt with, 175 with minimal negative effects, given that the working group has 176 decided to address these issues, but the choice of exactly how to 177 address these is best given effect in one or more standards-track 178 documents and/or errata. 180 In the context of this informational document, these normative 181 keywords will generally occur in the context of a quotation, most 182 often direct but sometimes indirect. The context will make it clear 183 whether the quotation is from: 185 o The base definition of the NFSv4.0 protocol [RFC7530]. 187 o The document updating the handling of migration in NFSv4.0 188 [RFC7931]. 190 o The current definition of the NFSv4.1 protocol [RFC5661]. 192 An additional possibility is that these terms may appear in a 193 proposed or possible text to serve as a replacement for a current 194 protocol specification. Sometimes, a number of possible alternative 195 texts may be listed and benefits and detriments of each examined in 196 turn. 198 2.3. Terminology Used in this Document 200 In this document the phrase "client ID" always refers to the 64-bit 201 shorthand identifier assigned by the server (a clientid4) and never 202 to the structure which the client uses to identify itself to the 203 server (called an nfs_client_id4 or client_owner in NFSv4.0 and 204 NFSv4.1 respectively). The opaque identifier within those structures 205 is referred to as a client id string". 207 Regarding the discussion of potential network endpoints, we use the 208 following terminology: 210 o The phrase "connection type" denotes the use of an existing or 211 potential connection to support NFSv4 layered on top of the RPC 212 stream transport as described in [RFC5531] or on top of RPC-over- 213 RDMA as described in [RFC8166]. Establishing a connection of a 214 particular type requires that the client and server support that 215 connection type given the particular client and server network 216 addresses used. 218 o Each connection is established between a client and a specfic 219 server endpoint. Two endpoints are considered distinct if they 220 differ in either network address or connection type. Multiple 221 connections may be established to the same endpoint or to 222 different endpoints. 224 o The phrase "network endpoint specification" refers to the 225 combination of a network address and a connection type. 227 Regarding trunking of connections to server network endpoints, we use 228 the following terminology: 230 o Trunking detection refers to ways of deciding whether two specific 231 network endpoints are connected to the same NFSv4 server. The 232 means available to make this determination depends on the protocol 233 version, and, in some cases, on the client implementation. 235 o Two network endpoints connected to the same server are said to be 236 server-trunkable. 238 o Two network endpoints connected to the same server such that 239 connections to those endpoints can be used to support a single 240 common session are referred to as session-trunkable. Note NFSv4.1 241 allows two endpoints to be server-trunkable without being session- 242 trunkable, while in NFSv4.0 no addresses are session-trunkable, 243 since there are no sessions. 245 o Trunking discovery is a process by which a client using one 246 network endpoint can obtain the network addresses for endpoints 247 that are trunkable (either server-trunkable or session-trunkable) 248 with it. 250 Regarding terminology relating to attributes used in trunking 251 discovery and other multi-server namespace features: 253 o Location attributes include the fs_locations and fs_locations_info 254 attributes. 256 o Location entries are the individual file system locations in the 257 location attributes. 259 o Location elements are derived from location entries. If a 260 location entry specifies an IP address there is only a single 261 corresponding location element. Location entries that contain a 262 host name, are resolved using DNS, and may result in one or more 263 location elements. All location elements consist of a location 264 address which is the IP address of an interface to a server and an 265 fs name which is the location of the file system within the 266 server's pseudo-fs. The fs name is empty if the server has no 267 pseudo-fs and only a single exported file system at the root 268 filehandle. 270 o Two location elements are said to be server-trunkable if they 271 specify the same fs name and the location addresses are such that 272 the location addresses are server-trunkable. 274 o Two location elements are said to be session-trunkable if they 275 specify the same fs name and the location addresses are such that 276 the location addresses are session-trunkable. 278 Each set of server-trunkable location elements defines the available 279 access paths to a particular file system. When there are multiple 280 such file systems, each of these, which contains the same data, is a 281 replica of the others. Logically, such replication is symmetric, 282 since the fs currently in use and an alternate fs are replicas of 283 each other. Often, in other documents, the term "replica" is not 284 applied to the fs currently in use, despite the fact that the 285 replication relation is inherently symmetric. 287 3. Issues that Apply to Multiple Versions and their Resolution 289 Although there are a common set of issues that need to be addressed, 290 the diffences between NFSv4.0 and NFSv4.1 means that the detailed 291 handling of these issues will be significantly different in each 292 prototol. 294 In order to accommodate this situation, this section will deal with 295 the commonalities across protocol version while the specfics 296 appropriate to each protocol version are dealt with in Sections 4 and 297 5 respectively. 299 3.1. Issue Summary 301 Many of these issues arise from a lack of clarity regarding the 302 meaning of and proper handling for location attributes that specify 303 more than a single server address. Such situations can arise as a 304 result of multiple entries in the same attribute or because a single 305 entry has a server name which, when processed by DNS, is mapped to 306 multiple server addresses. 308 Another set of issues arises from the fact that many of the 309 facilities thst must deal with multiple network addresses assume 310 there is only a single connection type shared by all of the 311 addresses. It is necessary to deal with a mixture of connection 312 types. 314 Both [RFC7530] and [RFC5661] indicate that multiple addresses may be 315 present and that these addresses may be different paths to the same 316 server as well as different copies of the same data. However, the 317 following issues have, for both protocols, interfered with the 318 recognition of the existing location attributes as a way of providing 319 a trunking discovery function: 321 o There is no discussion of the use of these attributes when a file 322 system is first accessed, giving the impression that they are only 323 to be used as a way of overcoming access difficulties. 325 o The treatment of migration (and in the case of NFSv4.1 of file 326 system transitions in general) is written as if only a single 327 server address will be accessed. 329 o Although location attributes can contain the addresses of 330 migration targets and of additional replicas as well, the issues 331 that arise when both of these are specified are not clearly 332 discussed. 334 In addition, there are factors regarding trunking that relate to 335 specific protocol versions and documents: 337 o In NFSv4.0, as described solely by [RFC7530], trunking is treated 338 as a problem to be avoided, making the whole matter moot. 340 o In NFSv4.0, as described by [RFC7530] together with [RFC7931], the 341 situation is different. There is a means of trunking detection 342 suggested in [RFC7931] but it is a suggestion only valid when the 343 client chooses to use the uniform client id string model. 345 o For NFSv4.1, as described by [RFC5661], there is a standard method 346 of trunking detection, which can be relied upon. 348 The issues that need to be addressed for both versions are: 350 o Provision of a trunking discovery facility to allow a client to 351 find out about other addresses that may be used to access the 352 current server. 354 o Discussion of how the appropriate connection type for a given 355 client-server connection is to be arrived at and of trunking 356 issues between endpoints of multiple connection types using the 357 same network address. 359 o Better integration of migration with trunking changes, including 360 situations in which the set of endpoints usable to access the same 361 server changes (without migration) and those in which there is a 362 shift to a different server, but trunking of endpoints on either 363 the source or destination is involved 365 Note that although these issues need to be addressed for both 366 protocols, the resolutiions need not be the same and the protocol 367 facilities within each protocol may limit the completeness of the 368 resolutions provided. 370 3.2. Resolution of Multi-Version Issues 372 Although the specifics of addressing these issues will be different 373 for different versions, there are some common aspects discussed in 374 the subsections below: 376 o The trunking discovery function is to be addressed in 377 substantially the same way in both versions, as explained in 378 Section 3.2.1. The only version-related differences are the 379 inclusion of the fs_locations_info attribute in NFSv4.1 and the 380 potential addition of further per-endpoint information within 381 extensions to be defined for use in later versions of NFSv4. 383 o The interaction of trunking and migration is discussed in general 384 terms in Section 3.2.2. However, the specifics of the NFSv4.1 385 client's response to NFS4ERR_MOVED are discussed in Sections 386 5.4.3, 5.4.4, and 5.4.5. 388 3.2.1. Providing Trunking Discovery 390 A client can discover a set network addresses to use to access a file 391 system using an NFSv4 server in a number of ways: 393 o If the client is accessing a server using its name, that name can 394 be mapped to a set of IP addresses using DNS and if multiple 395 addresses are available, those addresses can generally be used 396 together to access the server. 398 o A client connected to a server without knowledge of its name can 399 obtain the value of a location attribute (e.g. fs_locations). 400 Where an entry within that attribute specifies a server name, DNS 401 can be used to obtain one or more network addresses corresponding 402 to that name. In cases in which one of those is the address being 403 used, the others that corresponding to that name can also be used 404 to access the server. 406 o A client can obtain the value of a location attribute (e.g. 407 fs_locations) and use location entries that specify network 408 addresses. When there is a means of trunking detection available 409 (see below), all of addresses that are determined to correspond to 410 the same server can be used to access the server. 412 Note that the last two of these are usable in situations in which 413 NFS4ERR_MOVED was returned. Note that this does not necessarily mean 414 that migration has occurred since there may be a shift in the set of 415 network addresses to be used without changing to a different server. 416 See Section 3.2.2 for further discussion. 418 Which of the above means of providing trunking information is 419 appropriate to use in a given environment will depend on security 420 considerations, the possible need for the server to direct different 421 clients to different sets of addresses, and the availability of 422 trunking detection facilities on the clients. 424 With regard to security, the possibility that requests to determine 425 the set of network addresses corresponding to a given server might be 426 interfered with or have their responses corrupted needs to be taken 427 account of. As a result, when use of DNSSEC is not available, it 428 might not be advisable to present server names in location attributes 429 and present the network addresses directly, eliminating the need to 430 use DNS to effect this translation. Fetching of location attributes 431 should be done with integrity protection. 433 In many cases, the server will provide all the network addresses to 434 be used to access a given server, allowing the client to select the 435 address or set of addresses most suited to its purposes. However, in 436 some situations, the server will want to direct clients to use 437 specific sets of network addresses to effect load balancing, to meet 438 quality-of-service goals, or to optimize use of clustered servers by 439 directing traffic to the cluster element most able to handle it 440 efficiently. In such environments, presentation of network addresses 441 directly in the location attribute can help give the server the 442 necessary control over the paths to be used when accessing particular 443 file systems. When such techniques are used, servers typically 444 present their own network addresses in the location attribute while 445 adding the names of other servers, such as those used to access 446 replicas. 448 Trunking detection allows the client to determine whether two network 449 addresses can be used to access the same server. The availability of 450 trunking detection depends on the protocol version, and, in some 451 case, on client implementation choices: 453 o For NFSv4.0, a means by which it can be determined if two network 454 addresses correspond to the same server is suggested in [RFC7931]. 455 However is it is optional and only available to clients using the 456 uniform client id string approach. 458 o For NFSv4.1, the client can compare the server_owner returned in 459 the response to EXCHANGE_ID to determine if two network addresses 460 correspond to the same server. 462 As a result, direct presentation of network addresses in location 463 entries may be problematic for NFSv4.0, since some clients might not 464 have the trunking detection facilities that allow them to take 465 advantage of this information. For further discussion of issues 466 related to NFSv4.0, see Section 4.4. 468 3.2.2. Interaction of Trunking and Migration 470 When the set of network addresses designated by a location attribute 471 changes, NFS4ERR_MOVED may or may not result, and in some of the 472 cases in which it is returned migration will occur, while in others 473 there a shift in the network addresses used to access a particular 474 file system with no migration. 476 o When the list of networks addresses is a superset of that 477 previously in effect, there is no need for migration or any other 478 sort of client adjustment. Nevertheless, the client is free to 479 use an additional address if it provides another path to the same 480 server. If, on the other hand, it does not do so, the client may 481 treat it as it does a replica, to be used if the current server 482 addresses become unavailable. 484 o When the list of networks addresses is a subset of that previously 485 in effect, immediate action is not needed if the address was not 486 being used. The client should avoid using it in the future, 487 whether the address is for a replica or a potential additional 488 path to the server being used. 490 o When an address being removed is one of a number of paths to the 491 current server, the client can cease to use it but it can continue 492 to use it until NFS4ERR_MOVED is received. This is not considered 493 a migration event, unless it is the last available path to the 494 server that has become unusable. 496 When migration does occur, multiple addresses may be in use on the 497 server previous to migration and multiple addresses may be available 498 for use on the destination server. 500 With regard to the server in use, it may be that return of 501 NFS4ERR_MOVED indicates that a particular network address is no 502 longer to be used, without implying that migration of the file system 503 to a different server is needed In light of this possibility, clients 504 are best off not concluding that migration has occurred until 505 concluding that all the network addresses known to be associated with 506 the server are not usable. 508 It should be noted that the need to defer this determination is not 509 absolute. If a client is not aware of all network addresses for any 510 reason, if may conclude that migration has occurred when it has not 511 and treat a switch to a different server address as if it were a 512 migration event. This is generally harmless since the use of the 513 same server via a new address will appear as a successful Transparent 514 State Migration. 516 While significant harm will not arise from this misapprehension, it 517 can give rise to disconcerting situations. For example, if a lock 518 has been revoked during the address shift, it will appear to the 519 client as if the lock has been lost during migration, normally 520 calling for it to be recoverable via an fs-specific grace period 521 associated with the migration event. 523 With regard to the destination server, it is desirable for the client 524 to be aware of all the valid network addresses that can be used to 525 access the destination server. However, there is no need for this to 526 be done immediately. Implementations can process the additional 527 location elements in parallel with normal use of the first valid 528 location entry found to access the destination. 530 Because a location attribute may include entries relating to the 531 current server, the migration destination and possible replicas to 532 use, scanning for available network addresses could potentially be a 533 long process. The following list of helpful practices, here 534 presented as suggestions, could become RECOMMENDATIONs or 535 REQUIREMENTs in future standards-track documents 537 o Servers are well advised to place location entries that represent 538 addresses usable with the current server or a migration target 539 before those associated with replicas. 541 o A client can cease scanning for trunkable location entries once it 542 encounters one whose fs_name differs from the current fs name. 544 o A client can cease scanning for trunkable location entries once it 545 encounters a location element whose address in not server- 546 trunkable with the one it is using. 548 3.2.3. Dealing with Multiple Connection Types 550 Because of the use of RPC-over-RDMA [RFC8166] as an inderlying 551 transport for NFSV4, as described in [RFC8267], a client may have 552 mutiple connection types to the same server network address. This 553 gives rise to a number of issues with regard to NFSv4 multi-server 554 namespace features. 556 o In the case of migration or referral, the client is directed to 557 one or more server network addresses and faces the problem of 558 selecting the appropriate connection type. 560 o When trunking multiple connections, the client might be directed 561 to use the same server network address with a different set of of 562 potential connection types leaving the client to choose the 563 connection type to be used when a set with multiple connection 564 types is provided. 566 Although the situation is similar for both protocol versions, 567 differences in the attributes supported may result in important 568 differences in how connection types are selected. 570 o In the case of NFSv4.0, the fs_locations attribute has no ability 571 to indicate valid connection types. Only the network address is 572 provided, either directly in the location entry or as a result of 573 a server name being mapped to a set of network addresses. As a 574 result, the client may have to attempt connection with multiple 575 connection types, making its own selection of the subset to be 576 used when more than one connection type is available. 578 o NFSv4.1 has some facilities to aid in the selection of connection 579 types. The entries within the fs_locations_info attribute may 580 indicate the availability of RDMA connection support using the 581 FSLI4TF_RDMA flag. In addition, for RDMA implementation which 582 allow conversion (i.e. step-up) between non-RDMA and RDMA modes 583 within the scope of a single connection, the 584 CREATE_SESSION4_FLAG_CONN_RDMA flag may be used as part of 585 detecting whether RDMA support is present. When that flag is not 586 present, step-up is not supported, but the client may use the 587 FSLI4TF_RDMA flag to determine if RDMA support is available, and 588 establish a new connection to use to obtain RDMA support. 590 In addition to the selection of an appropriate connection type to use 591 when multiple connection types are available, the simultaneous 592 availability of multiple connection types raises issues related to 593 trunking, in the same way as the availability of multiple network 594 addresses connected to the same server. These issues, including the 595 relationship of such trunking to migration might be dealt with could 596 potentially be dealt sdifferently within NFSv4.0 and NFSv4.1, 597 although similar treatment is desirabble. The treatment of these 598 issues is discused in Sections 4.4.1 and 5.1.7 respectively. 600 Note that the handling of trunking for NFSv4.0 and for an NFSv4.1 601 metadata server differs from that for an NFSv4.1 data server. In 602 that latter case, specification of trunking patterns including the 603 connection type of endpoints is under the control of the metadata 604 server and the client simply uses the information presented by the 605 metadata server to guide selection of the endpoints to be accessed. 607 One potential difference between the versions that needs to be 608 resolved concerns the issue of the trunking of multiple connections 609 directed to endpoints that share a network address while differing as 610 to connection type. While NFSv4.1 is specified in [RFC5661] as 611 requring that such connections be trunkable, neither [RFC7530] nor 612 [RFC7931] contains a corresponding statement. 614 4. NFSv4.0 Issues 616 4.1. Core NFSv4.0 Migration Issues 618 Many of the problems seen with Transparent State Migration derived 619 from the inability of NFSv4.0servers to determine whether two client 620 IDs, issued on different servers, corresponded to the same client. 621 This difficulty derived in turn from the common practice, recommended 622 by [RFC7530], in which each client presented different client 623 identification strings to different servers, rather than presenting 624 the same identification string to all servers. 626 This practice, later referred to as the "non-uniform" client id 627 string approach, derived from concern that, since NFSv4.0 provided no 628 means to determine whether two IP addresses correspond to the server, 629 a single client connected to both might be confused by the fact that 630 state changes made via one IP address might unexpectedly affect the 631 state maintained with respect to the second IP address, thought of as 632 a separate server 634 To avoid this unexpected behavior, clients used the non-uniform 635 client id string approach. By doing so, a client connected to two 636 different servers (or to two IP addresses connected to the same 637 server) appeared to be two different servers. Since the server is 638 under the impression that two different clients are involved, state 639 changes made on each distinct IP address cannot be reflected on 640 another. 642 However, by doing things in that way, state migrated from server to 643 server cannot be referred to the actual client which generated it, 644 leading to confusion. 646 In addition to this core problem, the following issues with regard to 647 Transparent State Migration needed to be addressed: 649 o Clarification regarding the ability to merge state from different 650 leases even though their expiration times might not be precisely 651 synchronized. 653 o Clarifying the treatment of client IDs since it is not always 654 clear when clientid4 and when nfs_client_id4 was intended. 656 o Clarifying the logic of returning NFS4ERR_LEASE_MOVED. 658 o Clarifying the handling NFS4ERR_CLID_INUSE. 660 4.2. Resolution of Core Migration Protocol Difficulties in NFSv4.0 662 The client string identification issue was addressed in [RFC7931] as 663 follows: 665 o Defining both the uniform and non-uniform client id string 666 approaches as valid choices but indicating that the latter posed 667 difficulties for Transparent Stare Migration. 669 o Providing a way that clients using the uniform approach could use 670 to determine whether two IP addresses are connected to the same 671 server. 673 o Allowing clients using the uniform approach to avoid negative 674 consequences due to otherwise unexpected behavior since behavior 675 that is a consequence of known trunking relationships is not 676 unexpected. 678 o As a result, servers migrating state can be aware of the fact that 679 the same client is associated with two different items of state 680 even when that state was originally created on two different 681 servers. 683 Since all of the other issues noted in Section 4.1 were also 684 addressed by [RFC7931], publication of that document updating 685 [RFC7530] addressed all issues with Transparent State Migration in 686 NFSv4.0 known at that time. 688 4.3. Additional NFSv4.0 Issues 690 In light of the fact that a large set of migration-specific issues 691 were addressed by the publication of [RFC7931], the remaining issues 692 derive from those mentioned in Section 3.1. These include: 694 o Introducing facilities for trunking discovery. 696 o Clarifying the handling of multiple connection types, including 697 issues related to the trunking of multiple connections of 698 different types to the same network address. 700 o Clarifying the relationship between migration and trunking, 701 including trunking among multple server endpoints sharing a server 702 network address. 704 4.4. Resolution of Additional NFSv4.0 Issues 706 One possible approach to addressing these issues would entail 707 publication of an additional standards-track document updating 708 [RFC7530]. 710 Fortunately, it appears that all of the material to be updated 711 appears in Section 8 of that document, whether it concerns the 712 provision of trunking discovery or the interaction of trunking and 713 migration. It also appears that none of the material to be updated 714 is in sections updated by [RFC7931]. 716 A review of the existing Section 8 of [RFC7530], shows the following 717 sections as requiring significant attention: 719 o The existing Section 8.1 requires a considerable expansion to 720 explain the various uses of the fs_locations and the possible 721 interactions among them. 723 o The existing Section 8.4 may require substantial re-organization 724 to reflect the facts that fs_locations has multiple functions and 725 may be referenced on multiple occasions. 727 o The existing Section 8.5 follows the previous approach for NFSv4.0 728 in assuming that trunking simply cannot and should not happen. 729 For example, the last paragraph says: 731 If a single location entry designates multiple server IP 732 addresses, the client should choose a single one to use. When 733 two server addresses are designated by a single location entry 734 and they correspond to different servers, this normally 735 indicates some sort of misconfiguration, and so the client 736 should avoid using such location entries when alternatives are 737 available. When they are not, clients should pick one of the 738 IP addresses and use it, without using others that are not 739 directed to the same server. 741 In addition, no part of the existing Section 8, mentions the 742 possibility of multiple connection types, which completes the 743 exclusion of the possibility of multiple trunked server 744 endpoints from the existing description of NFSv4.0. 746 As written, this section seems to foreclose any use of trunking in 747 connection with migration. In retrospect, it appears that this 748 section should have been revised as part of [RFC7931], but since 749 that was not done then, the issue needs to be addressed now. 751 Overall, it appears that, in addition to the revision of Sections 8.1 752 and 8.5, Section 8.4 need to be reorganized. One possible approach 753 is to divide the material into sub-sections as follows: 755 o A replacement introductory subsection describing all the uses of 756 location information. 758 o A new subsection describing trunking discovery and detection, 759 based on use of the existing entries within the fs_locations 760 attribute. 762 o A new subsection describing the handling of multiple connection 763 types. For a discussion of issues to be addressed, see 764 Section 4.4.1. 766 o A replacement subsection dealing with replication and trunking. 768 o A replacement subsection dealing with migration. 770 o A new subsection dealing with the interaction of trunking, 771 replication, and migration. 773 4.4.1. Resolution of NFSv4.0 Issues with Multiple Connection Types 775 The existence of multiple connection types raises issues regarding 776 how the connection type to be used is determied by the client. Such 777 issues need to be addressed when a new server is accessed and also 778 when NFS4ERR_MOVED is returned and a server endpoint is to be 779 selected to access the current file system. 781 The absence of explicit support for multiple connection types within 782 NFSv4.0 means that the client has a great deal of freedom in making 783 this determination, although some implementation guidance could be 784 provided. A client could attempt to establish a connection of each 785 connection type and the connection type (or types) that it chooses. 786 To make this an efficient process, servers which do not provide 787 support for a particular connection type should promptly indicate 788 that non-support. It should be the case that all server endpoints 789 sharing a particular network address are to be considered trunkable,, 790 even though currently neither [RFC7530] nor [RFC7931] explicitly 791 states that. 793 The approach mentioned above should, in general, be usable in the 794 cases of migration and referral, as well as for initial mount. 795 Clients might well treat these situations differently, for example by 796 using the type of the current connection as the initial type to try 797 in the migration case, while not doing in other cases. 799 Situations in which NFS4ERR_MOVED is returned without requiring any 800 shift in target network address require special attention, in order 801 to allow a shift in the network endpoint to be used to be indicated 802 even if there is no corresponding shift in network address. In the 803 absence of multiple connection types, receiving NFS4ERR_MOVED when 804 acessing one file system serves as an indication that that address is 805 not to be used to access that file system subsequently, making it 806 necessary to use other network addresses to access the file system, 807 after migration or a shift in trunking patterns without migration. 809 Since NFSv4.0 does not provide any way for the server to specify the 810 use of particular connection types, it might seem that there is no 811 way for the server to direct such a shift. However, when 812 NFS4ERR_MOVED is returned and the network address on which it was 813 returned is still present in the location entries returned, a client 814 may reasonably conclude that: 816 o The endpoint from whic NFS4ERR_MOVED was returned is not to be 817 used to acess the file system in question. 819 o Other endpoints using the same network address but different 820 connection types could be used to access the filesystem. 822 This gives the client a set of server endpoints to test for access to 823 the filsystem. In cases in which there is already a connection 824 established to that endpoint, file system access can be tested using 825 a PUTFH within the target file system followed by a GETFH, which will 826 either succeed or return NFS4ERR_MOVED depending on whether the 827 endpoint used can validly access the file system. In other cases a 828 connection will need to be established before such a test can be 829 performed. 831 5. Issues for NFSv4.1 and Beyond 833 5.1. Issues to Address for NFSv4.1 835 Because NFSv4.1 embraces the uniform client-string approach, as 836 advised by section 2.4 of [RFC5661], addressing migration issues is 837 simpler, in that a shift in client id string models is not required. 838 Instead, NFSv4 returns information in the EXCHANGE_ID response to 839 enable trunking relationships to be determined by the client. 841 Despite this simplification, there are substantial issues that need 842 to be dealt with: 844 o The other necessary part of addressing migration issues, providing 845 for the server's merger of leases that relate to the same client, 846 is not currently addressed by [RFC5661] and changes need to be 847 made to make it clear that state needs to be appropriately merged 848 as part of migration, to avoid multiple client IDs between a 849 client-server pair. 851 o The current discussion (in [RFC5661]), of the possibility of 852 server_owner changes is incomplete and confusing. 854 o As with NFSV4.0, the interaction of trunking with migration and 855 other aspects of multi-server namespace needs to be clarified. 857 Addressing migration in NFSv41 will also require adaptation of the 858 approaches used in [RFC7931] to the NFSv4.1 environment including: 860 o The use of EXCHANGE_ID needs to be accommodated including issues 861 associated with the expected confirmation status of client IDs 862 transferred by Transparent State Migration. 864 o The use of sessions needs to be addressed including discussion of 865 the proper use of the status bits returned by the SEQUENCE 866 operation. 868 In addition, there are a number of new features within NFSv4.1 whose 869 relationship with migration needs to be clarified. Some examples: 871 o There needs to be some clarification of how migration, and 872 particularly Transparent State Migration, should interact with 873 pNFS layouts. 875 o There are a number of issues related to the migration of sessions 876 that need to be addressed. 878 Discussion of how to resolve these issues will appear in the sections 879 below. 881 5.1.1. Addressing State Merger in NFSv4.1 883 The existing treatment of state transfer in [RFC5661], has similar 884 problems to that in [RFC7530] in that it assumes that the state for 885 multiple filesystems formerly on different servers will not be merged 886 so that it appears under a single common client ID. We've already 887 seen the reasons that this is a problem with regard to NFSv4.0. 889 Although we don't have the problems stemming from the non-uniform 890 client-string approach, there are a number of complexities in the 891 existing treatment of state management in the section entitled "Lock 892 State and File System Transitions" in [RFC5661] that make this non- 893 trivial to address: 895 o Migration is currently treated together with other sorts of 896 filesystem transitions including transitioning between replicas 897 without any NFS4ERR_MOVED errors. 899 o There is separate handling and discussion of the cases of matching 900 and non-matching server scopes. 902 o In the case of matching server scopes, the text calls for an 903 unrealistic degree of transparency, suggesting that the source and 904 destination servers need to cooperate in stateid and client ID 905 assignment. 907 o In the case of non-matching server scopes, the text does not 908 mention the possibility of the transparent migration of state at 909 all, resulting in a functional regression from NFSV4.0 911 5.1.2. Addressing pNFS Relationship with Migration 913 This is made difficult because, within the pNFS framework, migration 914 might mean any of several things: 916 o Transfer of the MDS, leaving DS's as they are. 918 This would be minimally disruptive to those using layouts but 919 would require the pNFS control protocol being used to support the 920 DS being directed to a new MDS. 922 o Transfer of a DS, leaving everything else in place. 924 Such a transfer can be handled without using migration at all. 925 The server can recall/revoke layouts, and issue new ones, as 926 appropriate. 928 o Transfer of the filesystem to a new filesystem with both MDS and 929 DS's moving. 931 In such a transfer, an entirely different set of DS's will be at 932 the target location. There may even be no pNFS support on the 933 destination filesystem at all. 935 Migration needs to support both the first and last of these models. 937 5.1.3. Addressing Server_owner Changes in NFSv4.1 939 Section 2.10.5 of [RFC5661] states the following. 941 The client should be prepared for the possibility that 942 eir_server_owner values may be different on subsequent EXCHANGE_ID 943 requests made to the same network address, as a result of various 944 sorts of reconfiguration events. When this happens and the 945 changes result in the invalidation of previously valid forms of 946 trunking, the client should cease to use those forms, either by 947 dropping connections or by adding sessions. For a discussion of 948 lock reclaim as it relates to such reconfiguration events, see 949 Section 8.4.2.1. 951 While this paragraph is literally true in that such reconfiguration 952 events can happen and clients have to deal with them, it is confusing 953 in that it can be read as suggesting that clients have to deal with 954 them without disruption, which in general is impossible. 956 A clearer alternative would be: 958 It is always possible that, as a result of various sorts of 959 reconfiguration events, eir_server_scope and eir_server_owner 960 values may be different on subsequent EXCHANGE_ID requests made to 961 the same network address. 963 In most cases such reconfiguration events will be disruptive and 964 indicate that an IP address formerly connected to one server is 965 now connected to an entirely different one. 967 Some guidelines on client handling of such situations follow: 969 o When eir_server_scope changes, the client has no assurance that 970 any id's it obtained previously (e.g. file handles) can be 971 validly used on the new server, and, even if the new server 972 accepts them, there is no assurance that this is not due to 973 accident. Thus it is best to treat all such state as lost/ 974 stale although a client may assume that the probability of 975 inadvertent acceptance is low and treat this situation as 976 within the next case. 978 o When eir_server_scope remains the same and 979 eir_server_owner.so_major_id changes, the client can use 980 filehandles it has and attempt reclaims. It may find that 981 these are now stale but if NFS4ERR_STALE is not received, he 982 can proceed to reclaim his opens. 984 o When eir_server_scope and eir_server_owner.so_major_id remain 985 the same, the client has to use the now-current values of 986 eir_server-owner.so_minor_id in deciding on appropriate forms 987 of trunking. 989 5.1.4. Addressing Confirmation Status of Migrated Client IDs in NFSv4.1 991 When a client ID is transferred between systems as a part of 992 migration, it has never been clear whether it should be considered 993 confirmed or unconfirmed on the target server. In the case in which 994 an associated session is transferred together with the client ID, it 995 is clear that the transferred client ID needs to be considered 996 confirmed, as the existence of an associated session is incompatible 997 with an unconfirmed client ID. 999 The case in which a client ID is transferred without an associated 1000 session is less clear-cut, particularly since the treatment of 1001 EXCHANGE_ID in [RFC5661] assumes that CREATE_SESSION is the only 1002 means by which a client id may be confirmed. While this assumption 1003 is valid in the absence of Transparent State Migration, 1004 implementation of migration means that if this assumption is 1005 maintained, it is not clear how migrated client IDs can be a 1006 accommodated. If this assumption were maintained, we would have to 1007 choose between the following two alternatives, regarding whether the 1008 client ID to be reported as confirmed when EXCHANGE_ID is used to 1009 register an already-known client_owner with the server. 1011 o Report the client ID unconfirmed, because of the lack of an 1012 associated session. This makes it simpler for the client to 1013 determine whether there is an associated session transferred at 1014 the same time. However, it is inconsistent with the fact there 1015 are stateids which have been transferred with the client ID. 1017 o Report the client ID as confirmed, because it was confirmed on the 1018 source server and the transfer is not considered to have affected 1019 that. Given the current description of EXCHANGE_ID in [RFC5661], 1020 some modification in the treatment of client id confirmation is 1021 called for. In particular, provision would have to be made to 1022 enable the client id slot sequence id to be used by the client to 1023 be determined. 1025 Although the first approach makes it simpler for the client to 1026 determine whether there is an associated session transferred at the 1027 same time, it makes it more difficult to determine whether 1028 Transparent State Migration has occurred. Section 5.1.6. 1030 In any case, adjustments will be required to deal with the fact that 1031 [RFC5661] currently assumes that a client id can only be confirmed by 1032 issuing a CREATE_SESSION. In order to properly deal with the status 1033 of migrated client ids, we have to distinguish among: 1035 o The confirmation status as reported by EXCHANGE_ID. 1037 o Whether the client id is considered confirmed as that term is used 1038 in the many other cases in which the confirmation status of a 1039 client ID affects how requests are handled. 1041 o How the client is to determine the initial sequence id to be used 1042 when doing operations such as CREATE_SESSION. 1044 In [RFC5661] as it currently stands all of these are tied together 1045 and it is not obvious how migrated client IDs could be accommodated 1046 in this structure, and what changes are necessary to make this 1047 possible. For more discussion of this issue, see Section 5.2.1. 1049 5.1.5. Addressing Changes in Trunking Configuration 1051 When the client us capable of finding out a set of network addresses 1052 to use in accessing a server, it is always possible for that set to 1053 change. 1055 This sometimes requires that a network address previously used to 1056 access a server becomes invalid for that purpose. This requires a 1057 way of notifying the client and a way for the client to adapt to this 1058 change by using a new set of network addresses to access the server. 1059 his will involve recovery much like hat for migration although the 1060 same server and file system is used throughout. 1062 5.1.6. Addressing Session Migration in NFSv4.1 1064 Some issues that need to be addressed regard the migration of 1065 sessions, in addition to client IDs and stateids 1067 o It needs to be made clearer how the client can deal with the 1068 possibility that sessions might or might not be transferred as 1069 part of Transparent State Migration. 1071 o Rules need to be clarified regarding possible transfer of sessions 1072 when either the source session is being used to access other file 1073 systems on source server or there is already a session connecting 1074 the client to the destination server. 1076 o There needs to be more detail regarding how the protocol avoids 1077 situations in which the same session is subject to concurrent 1078 changes on two different servers at the same time. 1080 5.1.7. Dealing with Multiple Connection Types in NFSv4.1 1082 The existence of multiple connection types raises issues regarding 1083 how the connection type to be used is determied by the client. Such 1084 issues need to be addressed when a new server is accessed and also 1085 when NFS4ERR_MOVED is returned and a server endpoint is to be 1086 selected to access the current file system. 1088 The limited support for multiple connection types within NFSv4.1 1089 means that a client can make this determination by first establishing 1090 a non-RDMA connection and then using the the FSLI4TF_RDMA flag in the 1091 fs_locations_info attribute for the root file system to determine if 1092 an RDMA conection should be established. Such a connection can then, 1093 at the client's option, replace or remain truked with the original 1094 connection. As an alternative, where support is provided, the xxxx 1095 flag returned by EXCHANGE_ID can be used to guide a trnsfer of the 1096 existing connection to RDMA mode. 1098 The approach mentioned above should, in general, be usable in the 1099 cases of migration and referral, as well as for initial mount. 1101 Situations in which NFS4ERR_MOVED is returned without requiring any 1102 shift in target network address require special attention, in order 1103 to allow a shift in the network endpoint to be used to be indicated 1104 even if there is no corresponding shift in network address. In the 1105 absence of multiple connection types, receiving NFS4ERR_MOVED when 1106 acessing one file system serves as an indication that that address is 1107 not to be used to access that file system subsequently, making it 1108 necessary to use other network addresses to access the file system, 1109 after migration or a shift in trunking patterns without migration. 1111 Since NFSv4.1 only limited facilities for the server to specify the 1112 use of particular connection types, there are difficulties in 1113 directing such a shift. When NFS4ERR_MOVED is returned and the 1114 network address on which it was returned is still present in the 1115 location entries returned, a client may reasonably conclude that: 1117 o The usability of the aassociated RDMA endpoint can be determined 1118 based on the status of the the FSLI4TF_RDMA in the 1119 fs_locations_info attribute for the file system being accessed. 1121 o The endpoint returning NFS4ERR_MOVED is not to be used to acess 1122 the file system in question. 1124 o Other endpoints using the same network address but different 1125 connection types could be used to access the filesystem. 1127 This generally allows client to determine set of server endpoints to 1128 be used to access the filsystem. In cases in which there is some 1129 ambiguity file system access can be tested by establishing a 1130 connection if not already present and then using a PUTFH within the 1131 target file system followed by a GETFH, which will either succeed or 1132 return NFS4ERR_MOVED depending on whether the endpoint used can 1133 validly access the file system. 1135 5.2. Possible Resolutions for NFSv4.1 Protocol Issues 1137 The subsections below explore some ways of dealing with clarifying 1138 the protocol to address issues discussed in Section 5.1 1140 5.2.1. Client ID Confirmation Issues 1142 As mentioned previously [RFC5661], makes no provision for client IDs 1143 that are confirmed other than through the use of CREATE_SESSION. For 1144 example Section 18.35 of [RFC5661] states: 1146 The client uses the EXCHANGE_ID operation to register a particular 1147 client owner with the server. The client ID returned from this 1148 operation will be necessary for requests that create state on the 1149 server and will serve as a parent object to sessions created by 1150 the client. In order to confirm the client ID it must first be 1151 used, along with the returned eir_sequenceid, as arguments to 1152 CREATE_SESSION. If the flag EXCHGID4_FLAG_CONFIRMED_R is set in 1153 the result, eir_flags, then eir_sequenceid MUST be ignored, as it 1154 has no relevancy. 1156 In deciding how to address the status of migrated client IDs in the 1157 case of Transparent State Migration, we should avoid giving undue 1158 weight to the last sentence of the above simply because it is stated 1159 in the form of a normative requirement. We should instead focus on 1160 the reasons such terms (i.e. those defined by [RFC2119]) are to be 1161 used, to state interoperability constraints. In this case, the 1162 "MUST" applies to a conclusion based on the premise that a 1163 CREATE_SESSION must have been done to assure that the client ID is 1164 reliably known to the server. 1166 In that light, let us consider a possible replacement, that treats 1167 confirmation by means of CREATE_SESSION as one of a number of 1168 possible means and avoids some the undesirable consequences of 1169 adherence to the current approach, originally conceived without 1170 taking state migration into account. 1172 The client uses the EXCHANGE_ID operation to register a particular 1173 client_owner with the server. However, when the client_owner has 1174 been already been registered by other means (e.g. Transparent 1175 State Migration), the client may still use EXCHANGE_ID to obtain 1176 the client ID assigned previously. 1178 The client ID returned from this operation will be associated with 1179 the connection on which the EXHANGE_ID is received and will serve 1180 as a parent object for sessions created by the client on this 1181 connection or to which the connection is bound. As a result of 1182 using those sessions to make requests involving the creation of 1183 state, that state will become associated with the client ID 1184 returned. 1186 In situations in which the registration of the client_owner has 1187 not occurred previously, the client ID must first be used, along 1188 with the returned eir_sequenceid, in creating an associated 1189 session using CREATE_SESSION. 1191 If the flag EXCHGID4_FLAG_CONFIRMED_R is set in the result, 1192 eir_flags, then it is an indication that the registration of the 1193 client_owner has already occurred and that a further 1194 CREATE_SESSION is not needed to confirm it. Of course, subsequent 1195 CREATE_SESSION operations may be needed for other reasons. 1197 The value eir_seqenceid is used to establish an initial sequence 1198 value associate with the client ID returned. In cases in which a 1199 CREATE_SESSION has already been done, there is no need for this 1200 value, since sequencing of such request has already been 1201 established and the client has no need for this value and will 1202 ignore it 1204 5.2.2. Dealing with Multiple Location Entries 1206 The possibility that more than one server address may be present in 1207 location attributes requires further clarification. This is 1208 particularly the case, given the potential role of trunking for 1209 NFSv4.1, whose connection to migration needs to be clarified. 1211 The description of the location attributes in [RFC5661], while it 1212 indicates that multiple address entries in these attributes may be 1213 used to indicate alternate paths to the file system, does so mainly 1214 in the context of replication and does so without mentioning 1215 trunking. The discussion of migration does not discuss the 1216 possibility of multiple location entries or trunking, which we will 1217 explore here. 1219 We will cover cases in which multiple addresses appear directly in 1220 the attributes as well as those in which the multiple addresses 1221 result because a single location entry is expanded into multiple 1222 location elements using addresses provided by DNS. 1224 When the set of valid location elements by which a file system may be 1225 accessed changes, migration need not be involved. Some cases to 1226 consider: 1228 o When the set of location elements expands, migration is not 1229 involved. In the case in which the additional elements are not 1230 trunkable with ones previously being used, the new elements serve 1231 as additional access locations, available in case of the failure 1232 of server addresses being used. When additional elements are 1233 trunkable with those currently being used the client may use the 1234 additional addresses just as they might have if they had been 1235 available when use of the file system began. 1237 There is no current mechanism by which the client can be notified 1238 of a change in the set of available location for an fs. Given the 1239 client has at least one IP address available to access the 1240 filesystem in question, periodic polling is an adequate mechanism 1241 for the client to find additional server addresses to use to 1242 access the file system. 1244 o When the set of location elements contracts but none of the 1245 elements no longer usable were in fact being used by the client, 1246 then no migration is involved and no change in network addresses 1247 is needed. Only if the client were to start using one of the 1248 unavailable elements would the client be notified (via 1249 NFS4ERR_MOVED) of the need to not use those elements and to use 1250 others provided by a location attribute. 1252 When a specific server address being used becomes unavailable to 1253 service a particular file system, NFS4ERR_MOVED will be returned, and 1254 the client will respond based on the available locations. Whether 1255 continuity of locking state will be available depends on a number of 1256 factors: 1258 o If there are still elements in use trunkable with the element that 1259 has become unavailable, there will still be a continuity of 1260 locking state, even though Transparent State Migration per se has 1261 not occurred. If the in-use addresses are session-trunkable with 1262 the address becoming unavailable, only one connection is lost and 1263 all existing sessions will remain available. If, on the other 1264 hand, the in-use addresses are only clientid-trunkable with the 1265 address becoming unavailable, a session can be lost. However, 1266 that session can be made available on those other nodes, just as 1267 they it would have been if Transparent State Migration were in 1268 effect, even though no migration has occurred. 1270 o Otherwise, if there are available addresses trunkable with the one 1271 that has become unavailable, the client has access to existing 1272 locking state once it establishes a connection with the new 1273 addresses, using a new or existing session depending on the type 1274 of trunking in effect. This is also similar to the case in which 1275 Transparent State Migration has occurred, even though there is no 1276 migration, with the state remaining on the existing server. 1278 Note that this case, as well as the previous one, can be expected 1279 in the case in which the server seeks to direct traffic with 1280 regard to particular file systems to choose addresses, in the 1281 interest of load balancing, to adjust to hardware availability 1282 constraints, or for other reasons. 1284 o In other cases, migration has occurred and the client can 1285 determine whether Transparent State Migration occurred and whether 1286 any locking state was lost during the transfer. 1288 Whether migration has occurred or not, the client can use the 1289 procedure described in Section 5.4.3 to recover access to existing 1290 locking state and, in some cases, sessions. 1292 One should note the following differences between migration with 1293 Transparent State Migration and the similar cases in which there is a 1294 continuity of locking state with no change in the server. 1296 o When locks are lost (as indicated when using them or via the 1297 SEQ4_STATUS flags) and migration has not been done, they are not 1298 to be reclaimed, except when SEQ4_STATUS_RESTART_RECLAIM_NEEDED is 1299 set. Instead such losses are treated as lock revocations and 1300 acknowledged using FREE_STATEID. 1302 o When migration has not been done, there is no need for a 1303 RECLAIM_COMPLETE (with rca_one_fs set to true). 1305 5.2.3. Migration and pNFS 1307 When pNFS is involved, the protocol is capable of supporting: 1309 o Migration of the MDS, leaving DS's in place. 1311 o Migration of the file system as a whole, including the MDS and 1312 associated DS's. 1314 o Replacement of one DS by another. 1316 o Migration of a pNFS file system to one in which pNFS is not used. 1318 o Migration of a file system not using pNFS to one in which layouts 1319 are available. 1321 Migration of the MDS function is directly supported by Transparent 1322 State Migration. Layout state will normally be transparently 1323 transferred, just as other state is. As a result, Transparent State 1324 Migration provides a framework in which, given appropriate inter-MDS 1325 data transfer, one MDS can be substituted for another. 1327 Migration of the file system function as a whole can be accomplished 1328 by recalling all layouts as part of the initial phase of the 1329 migration process. As a result, IO will be done through the MDS 1330 during the migration process, and new layouts can be granted once the 1331 client is interacting with the new MDS. An MDS can also effect this 1332 sort of transition by revoking all layouts as part of Transparent 1333 State Migration, as long as the client is notified about the loss of 1334 state. 1336 In order to allow migration to a file system on which pNFS is not 1337 supported, clients need to be prepared for a situation in which 1338 layouts are not available or supported on the destination file system 1339 and so direct IO requests to the destination server, rather than 1340 depending on layouts being available. 1342 Replacement of one DS by another is not addressed by migration as 1343 such but can be effected by an MDS recalling layouts for the DS to be 1344 replaced and issuing new ones to be served by the successor DS. 1346 Migration may transfer a file system from a server which does not 1347 support pNFS to one which does. In order to properly adapt to this 1348 situation, clients which support pNFS, but function adequately in its 1349 absence, should check for pNFS support when a file system is migrated 1350 and be prepared to use pNFS when support is available. 1352 5.3. Defining Server Responsibilities for NFSv4.1 1354 The subsections below discuss the responsibilities of source and 1355 destination servers in effecting the necessary transfer of 1356 information to support Transparent State Migration. 1358 5.3.1. Server Responsibilities in Effecting Transparent State Migration 1360 The basic responsibility of the source server in effecting 1361 Transparent State Migration is to make available to the destination 1362 server a description of each piece of locking state associated with 1363 the file system being migrated. In addition to client id string and 1364 verifier, the source server needs to provide, for each stateid: 1366 o The stateid including the current sequence value. 1368 o The associated client ID. 1370 o The handle of the associated file. 1372 o The type of the lock, such as open, byte-range lock, delegation, 1373 layout. 1375 o For locks such as opens and byte-range locks, there will be 1376 information about the owner(s) of the lock. 1378 o For recallable/revocable lock types, the current recall status 1379 needs to be included. 1381 o For each lock type there will by type-specific information, such 1382 as share and deny modes for opens and type and byte ranges for 1383 byte-range locks and layouts. 1385 A further server responsibility concerns locks that are revoked or 1386 otherwise lost during the process of file system migration. Because 1387 locks that appear to be lost during the process of migration will be 1388 reclaimed by the client, the servers have to take steps to ensure 1389 that locks revoked soon before or soon after migration are not 1390 inadvertently allowed to be reclaimed in situations in which the 1391 continuity of lock possession cannot be assured. 1393 o For locks lost on the source but whose loss has not yet been 1394 acknowledged by the client (by using FREE_STATEID), the 1395 destination must be aware of this loss so that it can deny a 1396 request to reclaim them. 1398 o For locks lost on the destination after the state transfer but 1399 before the client's RECLAIM_COMPLTE is done, the destination 1400 server should note these and not allow them to be reclaimed. 1402 An additional responsibility of the cooperating servers concerns 1403 situations in which a stateid cannot be transferred transparently 1404 because it conflicts with an existing stateid held by the client and 1405 associated with a different file system. In this case there are two 1406 valid choices: 1408 o Treat the transfer, as in NFSv4.0, as one without Transparent 1409 State Migration. In this case, conflicting locks cannot be 1410 granted until the client does a RECLAIM_COMPLETE, after reclaiming 1411 the locks it had, with the exception of reclaims denied because 1412 they were attempts to reclaim locks that had been lost. 1414 o Implement Transparent State Migration, except for the lock with 1415 the conflicting stateid. In this case, the client will be aware 1416 of a lost lock (through the SEQ4_STATUS flags) and be allowed to 1417 reclaim it. 1419 5.3.2. Synchronizing Session Transfer 1421 When transferring state between the source and destination, the 1422 issues discussed in Section 7.2 of [RFC7931] must still be attended 1423 to. In this case, the use of NFS4ERR_DELAY is still necessary in 1424 NFSv4.1, as it was in NFSv4.0, to prevent locking state changing 1425 while it is being transferred. 1427 There are a number of important differences in the NFS4.1 context: 1429 o The absence of RELEASE_LOCKOWNER means that the one case in which 1430 an operation could not be deferred by use of NFS4ERR_DELAY no 1431 longer exists. 1433 o Sequencing of operations is no longer done using owner-based 1434 operation sequences numbers. Instead, sequencing is session- 1435 based 1437 As a result, when sessions are not transferred, the techniques 1438 discussed in [RFC7931] are adequate and will not be further 1439 discussed. 1441 When sessions are transferred, there are a number of issues that pose 1442 challenges since, 1444 o A single session may be used to access multiple file systems, not 1445 all of which are being transferred. 1447 o Requests made on a session may, even if rejected, affect the state 1448 of the session by advancing the sequence number associated with 1449 the slot used. 1451 As a result, when the filesystem state might otherwise be considered 1452 unmodifiable, the client might have any number of in-flight requests, 1453 each of which is capable of changing session state, which may be of a 1454 number of types: 1456 1. Those requests that were processed on the migrating file system, 1457 before migration began. 1459 2. Those requests which got the error NFS4ERR_DELAY because the file 1460 system being accessed was in the process of being migrated. 1462 3. Those requests which got the error NFS4ERR_MOVED because the file 1463 system being accessed had been migrated. 1465 4. Those requests that accessed the migrating file system, in order 1466 to obtain location or status information. 1468 5. Those requests that did not reference the migrating file system. 1470 It should be noted that the history of any particular slot is likely 1471 to include a number of these request classes. In the case in which a 1472 session which is migrated is used by filesystems other than the one 1473 migrated, requests of class 5 may be common and be the last request 1474 processed, for many slots. 1476 Since session state can change even after the locking state has been 1477 fixed as part of the migration process, the session state known to 1478 the client could be different from that on the destination server, 1479 which necessarily reflects the session state on the source server, at 1480 an earlier time. In deciding how to deal with this situation, it is 1481 helpful to distinguish between two sorts of behavioral consequences 1482 of the choice of initial sequence ID values. 1484 o The error NFS4ERR_SEQ_MISORDERED is returned when the sequence ID 1485 in a request is neither equal to the last one seen for the current 1486 slot nor the next greater one. 1488 In view of the difficulty of arriving at a mutually acceptable 1489 value for the correct last sequence value at the point of 1490 migration, it may be necessary for the server to show some degree 1491 of forbearance, when the sequence ID is one that would be 1492 considered unacceptable if session migration were not involved. 1494 o Returning the cached reply for a previously executed request when 1495 the sequence ID in the request matches the last value recorded for 1496 the slot. 1498 In the cases in which an error is returned and there is no 1499 possibility of any non-idempotent operation having been executed, 1500 it may not be necessary to adhere to this as strictly as might be 1501 proper if session migration were not involved. For example, the 1502 fact that the error NFS4ERR_DELAY was returned may not assist the 1503 client in any material way, while the fact that NFS4ERR_MOVED was 1504 returned by the source server may not be relevant when the request 1505 was reissued, directed to the destination server. 1507 One part of adapting to these sorts of issues would restrict 1508 enforcement of normal slot sequence enforcement semantics until the 1509 client itself, by issuing a request using a particular slot on the 1510 destination server, established the new starting sequence for that 1511 slot on the migrated session. 1513 An important issue is that the specification needs to take note of 1514 all potential COMPOUNDs, even if they might be unlikely in practice. 1515 For example, a COMPOUND is allowed to access multiple file systems 1516 and might perform non-idempotent operations in some of them before 1517 accessing a file system being migrated. Also, a COMPOUND may return 1518 considerable data in the response, before being rejected with 1519 NFS4ERR_DELAY or NFS4ERR_MOVED, and may in addition be marked as 1520 sa_cachethis. 1522 Some possibilities that need to be considered to address the issues: 1524 o Do not enforce any sequencing semantics for a particular slot 1525 until the client has established the starting sequence for that 1526 slot on the destination server. 1528 o For each slot, do not return a cached reply returning 1529 NFS4ERR_DELAY or NFS4ERR_MOVED until the client has established 1530 the starting sequence for that slot on the destination server. 1532 o Until the client has established the starting sequence for a 1533 particular slot on the destination server, do not report 1534 NFS4ERR_SEQ_MISORDERED or return a cached reply returning 1535 NFS4ERR_DELAY or NFS4ERR_MOVED, where the reply consists solely of 1536 a series of operations where the response is NFS4_OK until the 1537 final error. 1539 5.4. Defining Client Responsibilities for NFSv4.1 1541 The subsections below discuss the responsibilities of the client in 1542 dealing with transition to a new server (migration) and to use of new 1543 network addresses in accessing existing servers. 1545 5.4.1. Client Recovery from Migration Events 1547 When a file system is migrated, there a number of migration-related 1548 status indications with which clients need to deal: 1550 o If an attempt is made to use or return a filehandle within a file 1551 system that has been migrated away from the server on which it was 1552 previously available, the error NFS4ERR_MOVED is returned. 1554 This condition continues on subsequent attempts to access the file 1555 system in question. The only way the client can avoid the error 1556 is to cease accessing the filesystem in question at its old server 1557 location and access it instead on the server to which it has been 1558 migrated. 1560 o Whenever a SEQUENCE operation is sent by a client to a server 1561 which generated state held on that client which is associated with 1562 a file system that has been migrated away from the server on which 1563 it was previously available, the status bit 1564 SEQ4_STATUS_LEASE_MOVED is set in the response. 1566 This condition continues until the client acknowledges the 1567 notification by fetching a location attribute for the migrated 1568 file system. When there are multiple migrated file systems, a 1569 location attribute for each such migrated file system needs to be 1570 fetched, in order to clear the condition. Even after the 1571 condition is cleared, the client needs to respond by using the 1572 location information to access the destination server to ensure 1573 that leases are not needlessly expired. 1575 Unlike the case of NFSv4.0 in which the corresponding conditions are 1576 distinct errors and thus mutually exclusive, in NFSv4.1 the client 1577 can, and often will, receive both indications on the same request. 1578 As a result, implementations need to address the question of how to 1579 co-ordinate the necessary recovery actions when both indications 1580 arrive simultaneously. It should be noted that when the server 1581 decides whether SEQ4_STATUS_LEASE_MOVED is to be set, it has no way 1582 of knowing which file system will be referenced or whether 1583 NFS4ERR_MOVED will be returned. 1585 While it is true that, when only a single migrated file system is 1586 involved, a single set of actions will clear both indications, the 1587 possibility of multiple migrated file systems calls for an approach 1588 in which there are separate recovery actions for each indication. In 1589 general, the response to neither indication can be subsumed within 1590 the other since: 1592 o If the client were to respond only to the MOVED indication, there 1593 would be no effective client response to a situation in which a 1594 file system was not being actively accessed at the time migration 1595 occurred. As a result, leases on the destination server might be 1596 needlessly expired. 1598 o If the client were to respond only to the LEASE_MOVED indication, 1599 recovery for migrated file systems in active use could be deferred 1600 in order to accomplish recovery for others not being actively 1601 accessed. The consequences of this choice can pose particular 1602 problems when there are a large number of file systems supported 1603 by a particular server, or when it happens that some servers, 1604 after receiving migrated file systems have periods of 1605 unavailability, such as occur as a result of server reboot. This 1606 can result in recovery for actively accessed migrated file systems 1607 being unnecessarily delayed for long periods of time. 1609 Similar considerations apply to other arrangements in which one of 1610 the indications, while not ignored per se, is subsumed within a 1611 single recovery process focused on recovery for the other indication. 1613 Although clients are free to decide on their own approaches to 1614 recovery, we will explore below an approach with the following 1615 characteristics: 1617 o All instances of the MOVED indication, whether they involve 1618 migration or not, should be dealt with promptly, either by doing 1619 the necessary recovery directly, providing that it be done 1620 asynchronously, or ensuring that it is already under way. 1622 o All instances of the LEASE_MOVED indication should be dealt with 1623 asynchronously, in a migration discovery thread whose job is to 1624 clear that indication by fetching the appropriate location 1625 attribute. Because this thread will only be fetching a location 1626 attribute and the fs_status attribute for the file systems 1627 referenced by the client, it cannot receive MOVED indications. 1628 Some useful guidance regarding possible implementation of a 1629 migration discovery thread can be found in Section 5.4.2. 1631 o When a migration discovery thread happens upon a migrated file 1632 system (i.e. not present and not a referral), the thread is likely 1633 to have cleared one (out of an unknown number) of file systems 1634 whose migration needs to be responded to. The discovery thread 1635 needs to schedule the appropriate migration recovery (as described 1636 in Section 5.4.3). This is necessary to ensure that migrated file 1637 systems will be referenced on the destination server in order to 1638 avoid unnecessary lease expiration. 1640 For many of the migrated file systems discovered in this way, the 1641 client has not received any MOVED indication. In such cases, 1642 lease recovery needs to be scheduled but it should not interfere 1643 with continuation of the migration discovery function. 1645 o When a migration discovery thread receives a LEASE_MOVED 1646 indication, it takes no special action but continues its normal 1647 operation. On the other hand, if a LEASE_MOVED indication is not 1648 received, it indicates that the thread has completed its work 1649 successfully. 1651 5.4.2. The Migration Discovery Process 1653 As noted above, LEASE_MOVED indications are best dealt with in a 1654 migration discovery thread. Because of this structure, 1656 o No action needs to be taken for such indications received by the 1657 migration discovery threads, since continuation of that thread's 1658 work will address the issue. 1660 o For such indications received in other contexts, the generally 1661 appropriate response is to initiate or otherwise provide for the 1662 execution of a migration discovery thread for file systems 1663 associated with the server IP address returning the indication. 1665 o In all cases in which the appropriate migration discovery thread 1666 is running, nothing further needs to be done to respond to 1667 LEASE_MOVED indications. 1669 This leaves a potential difficulty in situations in which the 1670 migration discovery thread is near to completion but is still 1671 operating. One should not ignore a LEASE_MOVED indication if the 1672 discovery thread is not able to respond to migrated file system 1673 without additional aid. A further difficulty in addressing such 1674 situation is that a LEASE_MOVED indication may reflect the server's 1675 state at the time the SEQUENCE operation was processed, which may be 1676 different from that in effect at the time the response is received. 1678 A useful approach to this issue involves the use of separate 1679 externally-visible discovery thread states representing non- 1680 operation, normal operation, and completion/verification of migration 1681 discovery processing. 1683 Within that framework, discovery thread processing would proceed as 1684 follows. 1686 o While in the normal-operation state, the thread would fetch, for 1687 successive file systems known to the client on the server being 1688 worked on, a location attribute plus the fs_status attribute. 1690 o If the fs_status attribute indicates that the file system is a 1691 migrated one (i.e. fss_absent is true and fss_type != 1692 STATUS4_REFERRAL) and thus that it is likely that the fetch of the 1693 location attribute has cleared one the file systems contributing 1694 to the LEASE_MOVED indication. 1696 o In cases in which that happened, the thread cannot know whether 1697 the LEASE_MOVED indication has been cleared and so it enters the 1698 completion/verification state and proceeds to issue a COMPOUND to 1699 see if the LEASE_MOVED indication has been cleared. 1701 o When the discovery thread is in the completion/verification state, 1702 if others get a LEASE_MOVED indication they note this fact and it 1703 is used when the request completes, as described below. 1705 When the request used in the completion/verification state completes: 1707 o If a LEASE_MOVED indication is returned, the discovery thread 1708 resumes its normal work. 1710 o Otherwise, if there is any record that other requests saw a 1711 LEASE_MOVED indication, that record is cleared and the 1712 verification request retried. The discovery thread remains in 1713 completion/verification state. 1715 o If there has been no LEASE_MOVED indication, the work of the 1716 discovery thread is considered completed and it enters the non- 1717 operating state. 1719 5.4.3. Overview of Client Response to NFS4ERR_MOVED 1721 This section outlines a way in which a client that receives 1722 NFS4ERR_MOVED can respond by using a new server or network address if 1723 one is available. As part of that process, it will determine: 1725 o Whether the NFS4ERR_MOVED indicates migration has occurred, or 1726 whether it indicates another sort of file system transition as 1727 discussed in Section 5.2.2. 1729 o In the case of migration, whether Transparent State Migration has 1730 occurred. 1732 o Whether any state has been lost during the process of Transparent 1733 State Migration. 1735 o Whether sessions have been transferred as part of Transparent 1736 State Migration. 1738 During the first phase of this process, the client proceeds to 1739 examine location entries to find the initial network address it will 1740 use to continue access to the file system or its replacement. For 1741 each location entry that the client examines, the process consists of 1742 five steps: 1744 1. Performing an EXCHANGE_ID is directed at the location address. 1745 This operation is used to register the client-owner with the 1746 server, to obtain a client ID to be use subsequently to 1747 communicate with it, to obtain tat client ID's confirmation 1748 status and, to determine server_owner and scope for the purpose 1749 of determining if the entry is trunkable with that previously 1750 being used to access the file system (i.e. that it represents 1751 another path to the same file system and can share locking state 1752 with it). 1754 2. Making an initial determination of whether migration has 1755 occurred. The initial determination will be based on whether the 1756 EXCHANGE_ID results indicate that the current location element is 1757 server-trunkable with that used to access the file system when 1758 access was terminated by receiving NFS4ERR_MOVED. If it is, then 1759 migration has not occurred and the transition is dealt with, at 1760 least initially, as one involving continued access to the same 1761 file system on the same server through a new network address. 1763 3. Obtaining access to existing session state or creating new 1764 sessions. How this is done depends on the initial determination 1765 of whether migration has occurred and can be done as described in 1766 Section 5.4.4 in the case of migration or as described in 1767 Section 5.4.5 in the case of a network address transfer without 1768 migration. 1770 4. Verification of the trunking relationship assumed in step 2 as 1771 discussed in Section 2.10.5.1 of [RFC5661]. Although this step 1772 will generally confirm the initial determination, it is possible 1773 for verification to fail with the result that an initial 1774 determination that a network address shift (without migration) 1775 has occurred may be invalidated and migration determined to have 1776 occurred. There is no need to redo step 3 above, since it will 1777 be possible to continue use of the session established already. 1779 5. Obtaining access to existing locking state and/or reobtaining it. 1780 How this is done depends on the final determination of whether 1781 migration has occurred and can be done as described in 1782 Section 5.4.4 in the case of migration or as described in 1783 Section 5.4.5 in the case of a network address transfer without 1784 migration. 1786 Once the initial address has been determined, clients are free to 1787 apply an abbreviated process to find additional addresses trunkable 1788 with it (clients may seek session-trunkable or server trunkable 1789 addresses depending on whether they support clientid trunking). 1790 During this later phase of the process, further location entries are 1791 examined using the abbreviated procedure specified below: 1793 1. Before the EXCHANGE_ID, the fs_name field is examined and if it 1794 does not match that currently being used, the entry is ignored. 1795 otherwise, one proceeds as specified by step 1 above,. 1797 2. In the case that the network address is session-trunkable with 1798 one used previously a BIND_CONN_TO_SESSION is used to access that 1799 session using new network address. Otherwise, or if the bind 1800 operation fails, a CREATE_SESSION is done. 1802 3. The verification procedure referred to in step 4 above is used. 1803 However, if it fails, the entry is ignored and the next available 1804 entry is used. 1806 5.4.4. Obtaining Access to Sessions and State after Migration 1808 In the event that migration has occurred, the determination of 1809 whether Transparent State Migration has occurred is driven by the 1810 client ID returned by the EXCHANGE_ID and the reported confirmation 1811 status. 1813 o If the client ID is an unconfirmed client ID not previously known 1814 to the client, then Transparent State Migration has not occurred. 1816 o If the client ID is a confirmed client ID previously known to the 1817 client, then any transferred state would have been merged with an 1818 existing client ID representing the client to the destination 1819 server. In this state merger case, Transparent State Migration 1820 might or might not have occurred and a determination as to whether 1821 it has occurred is deferred until sessions are established and we 1822 are ready to begin state recovery. 1824 o If the client ID is a confirmed client ID not previously known to 1825 the client, then the client can conclude that the client ID was 1826 transferred as part of Transparent State Migration. In this 1827 transferred client ID case, Transparent State Migration has 1828 occurred although some state may have been lost. 1830 Once the client ID has been obtained, it is necessary to obtain 1831 access to sessions to continue communication with the new server. In 1832 any of the cases in which Transparent State Migration has occurred, 1833 it is possible that a session was transferred as well. To deal with 1834 that possibility, clients can, after doing the EXCHANGE_ID, issue a 1835 BIND_CONN_TO_SESSION to connect the transferred session to a 1836 connection to the new server. If that fails, it is an indication 1837 that the session was not transferred and that a new session needs to 1838 be created to take its place. 1840 In some situations, it is possible for a BIND_CONN_TO_SESSION to 1841 succeed without session migration having occurred. If state merger 1842 has taken place then the associated client ID may have already had a 1843 set of existing sessions, with it being possible that the sessionid 1844 of a given session is the same as one that might have been migrated. 1845 In that event, a BIND_CONN_TO_SESSION might succeed, even though 1846 there could have been no migration of the session with that 1847 sessionid. 1849 Once the client has determined the initial migration status, and 1850 determined that there was a shift to a new server, it needs to re- 1851 establish its lock state, if possible. To enable this to happen 1852 without loss of the guarantees normally provided by locking, the 1853 destination server needs to implement a per-fs grace period in all 1854 cases in which lock state was lost, including those in which 1855 Transparent State Migration was not implemented. 1857 Clients need to be deal with the following cases: 1859 o In the state merger case, it is possible that the server has not 1860 attempted Transparent State Migration, in which case state may 1861 have been lost without it being reflected in the SEQ4_STATUS bits. 1862 To determine whether this has happened, the client can use 1863 TEST_STATEID to check whether the stateids created on the source 1864 server are still accessible on the destination server. Once a 1865 single stateid is found to have been successfully transferred, the 1866 client can conclude that Transparent State Migration was begun and 1867 any failure to transport all of the stateids will be reflected in 1868 the SEQ4_STATUS bits. Otherwise. Transparent State Migration has 1869 not occurred. 1871 o In a case in which Transparent State Migration has not occurred, 1872 the client can use the per-fs grace period provided by the 1873 destination server to reclaim locks that were held on the source 1874 server. 1876 o In a case in which Transparent State Migration has occurred, and 1877 no lock state was lost (as shown by SEQ4_STATUS flags), no lock 1878 reclaim is necessary. 1880 o In a case in which Transparent State Migration has occurred, and 1881 some lock state was lost (as shown by SEQ4_STATUS flags), existing 1882 stateids need to be checked for validity using TEST_STATEID, and 1883 reclaim used to re-establish any that were not transferred. 1885 For all of the cases above, RECLAIM_COMPLETE with an rca_one_fs value 1886 of true should be done before normal use of the file system including 1887 obtaining new locks for the file system. This applies even if no 1888 locks were lost and needed to be reclaimed. 1890 5.4.5. Obtaining Access to Sessions and State after Network Address 1891 Transfer 1893 The case in which there is a transfer to a new network address 1894 without migration is similar to that described in Section 5.4.4 in 1895 that there is a need to obtain access to needed sessions and locking 1896 state. However, the details are simpler and will vary depending on 1897 the type of trunking between the address receiving NFS4ERR_MOVED and 1898 that to which the transfer is to be made 1900 To make a session available for use, a BIND_CONN_TO_SESSION should be 1901 used to obtain access to the session previously in use. Only if this 1902 fails, should a CREATE_SESSION be done. While this procedure mirrors 1903 that in Section 5.4.4, there is an important difference in that 1904 preservation of the session is not purely optional but depends on the 1905 type of trunking. 1907 Access to appropriate locking state should need no actions beyond 1908 access to the session. However. the SEQ4_STATUS bits should be 1909 checked for lost locking state, including the need to reclaim locks 1910 after a server reboot. 1912 5.5. Resolution of NFSv4.1 Issues 1914 One possibility is that addressing all of the NFSv4.1 issues would 1915 entail publication of a standards-track document updating [RFC5661]. 1917 Such a document would have three major elements: 1919 o A considerable expansion of the existing Section 11.4 explaining 1920 the various uses of the location attribute and the possible 1921 interactions among these various uses. This, like the 1922 corresponding replacement section for NFSv4.0 would be based on 1923 our Section 3.2 above. Information regarding the specifics of 1924 trunking discovery might appear in this section, in a new sub- 1925 section. As part of this revision, the existing Section 11.4.2 1926 would need to be revised to explain all the possible results of 1927 NFS4ERR_MOVED including migration and a possible transparent 1928 transition in which the network address changes but the server 1929 does not. 1931 o A revision of the existing section 18.35 (dealing with 1932 EXCHANGE_ID) addressing the issues discussed in Section 5.2.1. 1934 o A major replacement of the existing Section 11.7, entitled 1935 "Effecting File System Transitions", as discussed below. 1937 In addition, there is a set of smaller changes necessary 1939 o Update the existing Section 2.10.5 to clarify the proper response 1940 to server_owner changes, as described in our Section 5.1.3. 1942 o Replacement of the existing Section 15.1.2.4 to reflect the fact 1943 that NFS4ERR_MOVED can occur when a file system is now accessible 1944 at a different network address. A possible replacement text might 1945 be: 1947 The file system that contains the current filehandle object is 1948 not accessible using the network address which has been used. 1949 It may have been relocated, migrated to another server, be 1950 accessible using another network address on the current server, 1951 or it may have never been present. The client may obtain the 1952 new file system location by obtaining the "fs_locations" or 1953 "fs_locations_info" attribute for the current filehandle. For 1954 further discussion, refer to Section 11.4.2 1956 The replacement for the existing section 11.7 would maintain most 1957 sections essentially as they are, only making minor changes to 1958 include server-trunking in the discussion. However, in some cases 1959 involving more significant changes to existing sub-sections, and 1960 potential new sub-sections are listed below: 1962 o The existing Section 11.7.1 needs to be modified to refer 1963 explicitly to the previous discussion of trunking discovery. 1965 In addition, the term "multi-home single-server namespace", used 1966 nowhere else in [RFC5661], poses difficulties. From the 1967 description given it appears that the case being referred to in 1968 one in which two network addresses return server_owners with the 1969 same major_id and different minor_id values, making the network 1970 addresses server-trunkable without being session trunkable. 1972 A better approach would be to refer to "server-trunking" as used 1973 elsewhere in this document and use the replacement for the 1974 existing Section 18.35 to identify clientid trunking as the means 1975 to adapt to network addresses which are server-trunkable without 1976 being session-trunkable and session trunking as the means to adapt 1977 to network addresses which are session-trunkable. 1979 o The existing Section 11.7.2 needs to be better connected to 1980 trunking discovery. By calling these "transparent" transitions, 1981 it obscures the fact that some (or all) of the "transitions" it is 1982 discussing are not in fact transitions between servers or file 1983 systems but merely changes the set of communication paths in use. 1985 o The existing Section 11.7.2.1, needs to address more clearly the 1986 case of server-trunkable addresses which are not session- 1987 trunkable. As it is, it mentions the related concept of 1988 clustering, but only deals explicitly with the case in which two 1989 distinct servers share access to one or more file systems and does 1990 not mention the case in which the network addresses can be used to 1991 access a shared stateid space without being session-trunkable. 1993 o The existing Section 11.7.2.2, while correct, needs to be part of 1994 a general re-organization since the characteristics it lists as 1995 necessary for a transparent transition will be of use in other 1996 contexts, particularly as they apply to Transparent State 1997 Migration as well. It make sense to move these to a new sub- 1998 section within the equivalent of the Existing Section 11.7. 2000 o The existing Section 11.7.7, needs the a major rework to deal with 2001 its basic assumption, that existing state can only be made 2002 available on the destination server if the source and destination 2003 co-operate in state management and maintain a common client id 2004 space. It is not clear how this can be done, other than for 2005 servers working together so as to provide clientid trunking, a 2006 case that is probably considered as a "transparent transition". 2007 The section needs to modified to allow something along the lines 2008 of NFSv4.0-style Transparent State Migration with the details 2009 provided by a later section (see below). 2011 A related issues concerns the sentence, "In the case of migration, 2012 the servers involved in the migration of a file system SHOULD 2013 transfer all server state from the original to the new server. It 2014 is unclear why this is a "SHOULD" as the rest of the paragraph 2015 essentially tells the client that it needs to be prepared for the 2016 server not to do this. The equivalent is a "should" in [RFC7931], 2017 and there is no reason to add to confusion by making a "SHOULD" in 2018 NFSv4.1. also, there is no mention of the need to provide a fs- 2019 specific grace period in the cases in which Transparent State 2020 Migration is not made available. 2022 o Adding a new section (at level of the existing Section 11.7.7) 2023 about state transfer during migration. Although the phrase 2024 "Transparent State Migration" is well established in the context 2025 of NFSv4.0, the word "transparent" could cause confusion given the 2026 existing use of the phrase "transparent transitions". A possible 2027 title for the new section is "State Transfer during Migration" 2029 The new section would present the NFSv4.1-equivalent of Transparent 2030 State Migration as described in [RFC7931]. This would address the 2031 issues presented in Section 5.1 along the lines suggested in Sections 2032 5.2, 5.3, and 5.4. 2034 5.6. Potential Protocol Extensions 2036 There are a number possibilities to provides additional facilities 2037 related to issues discussed in this document using the protocol 2038 extension mechanisms described in [RFC8178]. These facilities relate 2039 to the handling of multiple connection types. 2041 The possibility of additional connection types was not addressed in 2042 NFSv4.0, either in [RFC3530] or [RFC7530]. While the use of mutiple 2043 connection types is allowed, facilities to determine the connection 2044 type to be used are sub-optimal and are expected to remain so. 2046 In the case of NFSv4.1, there are facilities to aid in the 2047 determination of connection types that can be used. However, such 2048 facilities are limited to the two connection types already defined 2049 and may have weaknesses in dealing with changes in the set of 2050 connection types to be used and in selecting connections to be used, 2051 particularly in clustered server environments, in which the set of 2052 potential trunked server endpoints can be large. 2054 In light of this situation, it appears that a number of potential 2055 extensions to NFSv4 might be considered, as provided for by 2056 [RFC8178]. Such extensions could take the form of additional 2057 OPTIONAL attributes. While these attributes would be part of 2058 NFSv4.2, the fact that there is no change in the set of REQUIRED 2059 features between NFSv4.1 and NFSv4.2 means that the upgrade path for 2060 clients and servers can be made relatively simple. 2062 The additonal attributes sketched out below would provide a more 2063 complete way of addressing the possibility of trunking of a large set 2064 of server endpoints, of multiple connection types: 2066 o A new fs-scoped attribute, fs_location_endpoints, could provide 2067 potential locations of a file system by using location entries 2068 specifying each potential endpoint, rather than specifing, as do 2069 fs_locations and fs_locations_info, the nework address applicable 2070 to all potenia endpoints. 2072 o A new server-scoped attribute, server_endpoints, could provide a 2073 set of trunkable endpoints to be used to access the current 2074 server, together with additional performance-related information 2075 useful for endpoint selection. 2077 A fuller elaboration of these proposals would require the writing of 2078 one or more standards-track documents, assuming sufficient interest 2079 in proceeding along this route. Any such work would be separate from 2080 other work suggested to resolve existing protocol issues and will not 2081 be mentioned in Section 6.2 2083 6. Evolution of Issue Handling 2085 6.1. History of this Document 2087 The contents of successive versions of this document have changed 2088 because new issues have been discovered, because there have been 2089 changes in our understanding of how these features should interact, 2090 and because some of the issues have been adequately addressed with 2091 regard to certain protocol versions. 2093 As a result, it may be helpful to understand the history of these 2094 issues, which is complicated because multiple NFSv4 protocols have 2095 been involved. 2097 This history can be summarized as follows 2099 o Initially, the focus was on the difficulties seen in NFSv4.0 2100 implementations of Transparent State Migration, and on identifying 2101 possible corrections to [RFC7530] that might address these issues. 2103 At this point, treatment of NFSv4.1 was minimal. 2105 o As examination of the issues continued, it became clear that the 2106 use of the non-uniform client string model was a critical element 2107 of the problem and further work proceeded on that basis. 2109 During the period, treatment of NFSv4.1 was expanded but the fact 2110 that NFSv4.1 had existing facilities for trunking detection was 2111 taken as an indication that the problems would not be difficult to 2112 address.. 2114 o As work proceeded on a standards-track document addressing the 2115 NFSv4.0 issues, material that proposed changes to address the 2116 issues became less relevant, since the effective vehicle for 2117 addressing these issues became the standards-track document 2118 eventually published as [RFC7931]. 2120 During this period, and subsequently, treatment of NFSv4.1 2121 remained essentially unchanged. 2123 o With the publication of [RFC7931], material regarding fixes for 2124 the NSV4.0 became vestigial but the material was retained for a 2125 while together with a shift from proposing those changes to 2126 reporting that they had been made. 2128 o Later, in response to experiences testing existing NFSv4.1 2129 implementations of migration, the focus of the document shifted 2130 decisively to NFSv4.1. As part of the analysis of migration 2131 within NFSv4.1, it was realized that issues related to the 2132 appearance of multiple addresses were fundamental to clearly 2133 describing how migration would work and that changes in the set of 2134 such addresses might or might not involve migration. 2136 At this point, discussion of NFSV4.0 issues was further limited. 2137 The issues seen were noted but the discussion of the resolution 2138 was limited to explaining that they had been addressed by the 2139 publication of [RFC7931]. 2141 o Finally, based on the results of work to provide NFSv4 with 2142 trunking discovery facilities, a decision was made that this work 2143 was most appropriately dealt with together with migration, for 2144 reasons noted previously. 2146 Since the trunking discovery facilities apply to all NFSv4 minor 2147 versions, work was needed to define those for NFSv4.0as well, 2148 together with the necessary interactions with migration. 2150 Although there is a need for further working group discussion and 2151 review, it appears that the issues to be dealt with have been 2152 identified and that most work to address these issues need to take 2153 place as part of the construction of one or more standards-track 2154 documents. See Section 6.2 for further information about possible 2155 approaches to providing the necessary documents. 2157 6.2. Further Work Needed 2159 The following table classifies issues in this area and indicates 2160 which are currently adequately addressed and where the protocol 2161 specifications need further correction or clarification. Where the 2162 topic is adequately addressed, a reference is given to the RFC 2163 providing support for the issue. In other cases, an area name 2164 (explained below) is given. 2166 +-------+-----------+----------+-----------+----------+-------------+ 2167 | Vers. | Trunking | Trunking | Xparent | Multiple | Interaction | 2168 | | Detection | Disc. | State | Conn. | of Trunking | 2169 | | | | Migration | Types | and | 2170 | | | | | | Migration | 2171 +-------+-----------+----------+-----------+----------+-------------+ 2172 | v4.0 | [RFC7931] | TrDisc-0 | [RFC7931] | Mct-0 | Int-0 | 2173 | v4.1+ | [RFC5661] | TrDisc-1 | TSM-1 | Mct-1 | Int-1 | 2174 +-------+-----------+----------+-----------+----------+-------------+ 2176 The following table explains the work that needs to be done 2177 corresponding to each area name above. 2179 +----------+--------------------------------------------------------+ 2180 | Area | Description | 2181 | Name | | 2182 +----------+--------------------------------------------------------+ 2183 | TrDisc-0 | Although it is possible for there to be multiple | 2184 | | location entries for a given file system, the | 2185 | | possibility of using these to enable trunking | 2186 | | discovery was not addressed in [RFC7530], most likely | 2187 | | because trunking was considered a problem to be | 2188 | | avoided (rather than a helpful feature) at that time. | 2189 | | This situation could have been addressed by the | 2190 | | publication of [RFC7931] but unfortunately that did | 2191 | | not happen. | 2192 +----------+--------------------------------------------------------+ 2193 | TrDisc-1 | Despite the fact that [RFC5661] provides a means of | 2194 | | trunking detection, trunking discovery was not | 2195 | | addressed. This problem was compounded by confusion | 2196 | | regarding multiple file system replicas arising from | 2197 | | the fact that multiple network addresses connected to | 2198 | | the same server were treated as if they were referring | 2199 | | to distinct sets of replicas. | 2200 +----------+--------------------------------------------------------+ 2201 | TSM-1 | Unlike [RFC7530], which mishandled Transparent State | 2202 | | Migration because of confusion arising from the lack | 2203 | | of appropriate trunking support, [RFC5661] simply | 2204 | | neglected to provide any description of this feature. | 2205 | | It appears likely that confusion between the needs of | 2206 | | migration and those of dealing with shifts in | 2207 | | responsibility for clustered file system access had | 2208 | | significant role in allowing this issue to be ignored. | 2209 | | Rectifying this situation along the lines of [RFC7931] | 2210 | | is complicated by the need to rewrite significant | 2211 | | pieces of the section about multi-server namespace to | 2212 | | address this confusion. Beyond this, the necessary | 2213 | | treatment will need to reflect changes required by the | 2214 | | use of the sessions model and related changes in | 2215 | | NFSv.1 and also address migration-related issues | 2216 | | raised by optional features such as pNFS and the | 2217 | | fs_locations_info attribute. | 2218 +----------+--------------------------------------------------------+ 2219 | Mct-0 | Even though protocol support for multiple connection | 2220 | | types is quite limited in NFSv4.0, there still are | 2221 | | multiple connection types specified and implemented. | 2222 | | As a result, some guidance has to be given to allow | 2223 | | interoperable implementations to be developed, and | 2224 | | used, without extensive user configuration effort. | 2225 | | This should include some treatment of situtions in | 2226 | | which the set of connection types to be used to access | 2227 | | a given file system changes, requiring appropriate | 2228 | | recovery from an NFS4ERR_MOVED error. | 2229 +----------+--------------------------------------------------------+ 2230 | Mct-1 | Even though protocol support for multiple connection | 2231 | | types is more limited than one might like, there are | 2232 | | helpful facilities that can be used to simplify the | 2233 | | process of determining the connection type(s) to be | 2234 | | uused. The proper use of the available facilities | 2235 | | needs to be clarified including examination of cases | 2236 | | in which the set of connection types to be used to | 2237 | | access a given file system changes, requiring | 2238 | | appropriate recovery from an NFS4ERR_MOVED error. | 2239 +----------+--------------------------------------------------------+ 2240 | Int-0 | The need to provide trunking-related information puts | 2241 | | additional focus on the issue of dealing with changes | 2242 | | in the value of location-related attributes. This | 2243 | | applies when trunking configurations change and at | 2244 | | other times as well. In addition, the existence of | 2245 | | multiple network addresses connected to the same | 2246 | | server requires clarification when migration and | 2247 | | replication features are used. | 2248 +----------+--------------------------------------------------------+ 2249 | Int-1 | This requires similar handling to the case above. | 2250 | | However, further work is made necessary by the fact | 2251 | | that shifts between different sets of network | 2252 | | addresses are erroneously treated as instances of | 2253 | | migration in [RFC5661]. | 2254 +----------+--------------------------------------------------------+ 2256 There are number of possible ways of packaging the necessary changes 2257 into RFCs. Some of these are impractical for various reasons: 2259 o While it possible to treat each area in its own RFC, writing seven 2260 RFCs would increase the work required, and delay needed 2261 corrections to both versions. Further, it would result in a 2262 situation in which in which someone needing to understand the 2263 specification of NFS version 4.0 or 4.1 would need to be familiar 2264 with a large set of RFCs. 2266 o One could have a document addressing all of the areas above Such a 2267 document would update both [RFC7530] and [RFC5661]. That would 2268 result in a confusing document given how different the v4.0 and 2269 v4.1 protocols are, since most readers will want a clear 2270 description of one or the other. 2272 o It is also possible to produce separate documents addressing 2273 Trdisc-*, TSM-1., Mct-*, and Int-*. This would be subject many of 2274 the difficulties of the two approaches above. 2276 The alternative, of organizing the changes by minor version, is being 2277 actively pursued by work on following Standards Track working group 2278 documents: 2280 o [I-D.ietf-nfsv4-mv0-trunking-update] addresses the issues within 2281 TrDisc-0 and Int-0 by providing updates to [RFC7530], the vast 2282 majority of which are within Section 8 of that document. Work to 2283 include Mct-0 needs to be added. 2285 o [I-D.ietf-nfsv4-mv1-msns-update] addresses the issues within 2286 TrDisc-1, TSM-1, and Int-1 providing updates to [RFC5661], the 2287 vast majority of which are within Section 11 of that document. 2288 Work to include Mct-1 is underway. 2290 These two documents will require additional review and discussion 2291 before proceeding to publication as Proposed standards, updating 2292 [RFC7530] and [RFC5661] respectively. 2294 If the working group decides to continue along this path, it may be 2295 desirable to consolidate the changes currently specified in these 2296 documents. Currently, these document replace individual sub-sections 2297 of Section 8 (of [RFC7530]) or Section 11 (of [RFC5661]). While this 2298 is helpful in explaining what is changing and why, things might be 2299 different when the eventual RFC is published. At that point, it is 2300 could be judged more important to have simply understood 2301 specifications of NFS versions 4.0 and 4.1. At that point, a full 2302 replacement section of the affected section might be more desirable 2303 as the basis of the RFC to be published. Alternatively, that 2304 consolidation might be delayed and done later as part of publication 2305 of rfc7530bis and rfc5661bis documents. 2307 7. Security Considerations 2309 In general, the Security Considerations sections of existing 2310 specifications for NFS versions 4.0 and 4.1 provide recommendations 2311 for appropriate handling of requests obtaining location-related 2312 information. In particular, it is recommended that integrity 2313 protection be used when fetching location-related attributes: 2315 o With regard to NFSv4.0, this is done in Section 8.6 of [RFC7931] 2316 which updates the Security Considerations section of [RFC7530]. 2318 o With regard to NFSv4.1, this is done in the Security 2319 Considerations section of [RFC5661]. 2321 Despite this however, there is a need for further changes in the 2322 Security Considerations with regard to both minor versions dealt with 2323 here. The following issues need to be addressed: 2325 o Because of the potential use of DNS to convert server names to a 2326 set of server network addresses, such translations are subject to 2327 the same sorts of potential interference with trunking discovery 2328 that would occur when trunking discovery is provided using network 2329 addresses returned in the location-related attributes. 2331 To address this issue, specifications for both minor versions need 2332 to mention the issue and indicate that use of DNSSEC [RFC4033] is 2333 appropriate. When it is not available, the server should allow 2334 use of DNS for trunking discovery to be avoided by presenting 2335 network addresses in the location-related attributes, with these 2336 values subject to RPCSECGSS integrity protection. 2338 o Although use of RPCSEC_GSS ([RFC2203], [RFC5403], [RFC7861]) with 2339 integrity protection is RECOMMENDED and "implementations" are 2340 REQUIRED to provide support. However, the possibility that a 2341 particular client may be unable to use RPCSEC_GSS when accessing a 2342 particular server cannot be excluded. As a result, it is 2343 necessary to discuss how such situations affect trunking 2344 discovery, referral, replication, and migration. 2346 o In the case of replication, referral, and migration, it is 2347 necessary to discuss how RPCSEC_GSS mutual authentication on the 2348 destination can be used to make sure that the network addresses 2349 provided by trunking discovery have not been interfered with and 2350 correspond to the server names provided by the location attributes 2351 on the server to which the client was directed. 2353 8. IANA Considerations 2355 This document does not require actions by IANA. 2357 9. References 2359 9.1. Normative References 2361 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 2362 Requirement Levels", BCP 14, RFC 2119, 2363 DOI 10.17487/RFC2119, March 1997, 2364 . 2366 [RFC2203] Eisler, M., Chiu, A., and L. Ling, "RPCSEC_GSS Protocol 2367 Specification", RFC 2203, DOI 10.17487/RFC2203, September 2368 1997, . 2370 [RFC4033] Arends, R., Austein, R., Larson, M., Massey, D., and S. 2371 Rose, "DNS Security Introduction and Requirements", 2372 RFC 4033, DOI 10.17487/RFC4033, March 2005, 2373 . 2375 [RFC5403] Eisler, M., "RPCSEC_GSS Version 2", RFC 5403, 2376 DOI 10.17487/RFC5403, February 2009, 2377 . 2379 [RFC5531] Thurlow, R., "RPC: Remote Procedure Call Protocol 2380 Specification Version 2", RFC 5531, DOI 10.17487/RFC5531, 2381 May 2009, . 2383 [RFC5661] Shepler, S., Ed., Eisler, M., Ed., and D. Noveck, Ed., 2384 "Network File System (NFS) Version 4 Minor Version 1 2385 Protocol", RFC 5661, DOI 10.17487/RFC5661, January 2010, 2386 . 2388 [RFC7530] Haynes, T., Ed. and D. Noveck, Ed., "Network File System 2389 (NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530, 2390 March 2015, . 2392 [RFC7861] Adamson, A. and N. Williams, "Remote Procedure Call (RPC) 2393 Security Version 3", RFC 7861, DOI 10.17487/RFC7861, 2394 November 2016, . 2396 [RFC7931] Noveck, D., Ed., Shivam, P., Lever, C., and B. Baker, 2397 "NFSv4.0 Migration: Specification Update", RFC 7931, 2398 DOI 10.17487/RFC7931, July 2016, 2399 . 2401 [RFC8166] Lever, C., Ed., Simpson, W., and T. Talpey, "Remote Direct 2402 Memory Access Transport for Remote Procedure Call Version 2403 1", RFC 8166, DOI 10.17487/RFC8166, June 2017, 2404 . 2406 [RFC8178] Noveck, D., "Rules for NFSv4 Extensions and Minor 2407 Versions", RFC 8178, DOI 10.17487/RFC8178, July 2017, 2408 . 2410 [RFC8267] Lever, C., "Network File System (NFS) Upper-Layer Binding 2411 to RPC-over-RDMA Version 1", RFC 8267, 2412 DOI 10.17487/RFC8267, October 2017, 2413 . 2415 9.2. Informative References 2417 [I-D.ietf-nfsv4-mv0-trunking-update] 2418 Lever, C. and D. Noveck, "NFS version 4.0 Trunking 2419 Update", draft-ietf-nfsv4-mv0-trunking-update-00 (work in 2420 progress), January 2018. 2422 [I-D.ietf-nfsv4-mv1-msns-update] 2423 Noveck, D. and C. Lever, "NFSv4.1 Update for Multi-Server 2424 Namespace", draft-ietf-nfsv4-mv1-msns-update-00 (work in 2425 progress), January 2018. 2427 [RFC3530] Shepler, S., Callaghan, B., Robinson, D., Thurlow, R., 2428 Beame, C., Eisler, M., and D. Noveck, "Network File System 2429 (NFS) version 4 Protocol", RFC 3530, DOI 10.17487/RFC3530, 2430 April 2003, . 2432 Acknowledgments 2434 The editor and authors of this document gratefully acknowledge the 2435 contributions of Trond Myklebust of Primary Data, Robert Thurlow of 2436 Oracle, and Andy Adamson of NetApp. We also thank Tom Haynes of 2437 Primary Data and Spencer Shepler of Microsoft for their guidance and 2438 suggestions. 2440 Special thanks go to members of the Oracle Solaris NFS team, 2441 especially Rick Mesta and James Wahlig who were then part of that 2442 team, for their work implementing an NFSv4.0 migration prototype and 2443 identifying many of the issues documented here. Also, the work of 2444 Xuan Qi for Oracle using NFSv4.1 client and server prototypes was 2445 helpful. 2447 Authors' Addresses 2449 David Noveck (editor) 2450 NetApp 2451 1601 Trapelo Road 2452 Waltham, MA 02451 2453 US 2455 Phone: +1 781 572 8038 2456 Email: davenoveck@gmail.com 2458 Piyush Shivam 2459 IBM Corporation 2460 11501 Burnet Road 2461 Austin, TX 78758 2462 US 2464 Email: piyush.shivam@ibm.com 2465 Charles Lever 2466 Oracle Corporation 2467 1015 Granger Avenue 2468 Ann Arbor, MI 48104 2469 US 2471 Phone: +1 248 614 5091 2472 Email: chuck.lever@oracle.com 2474 Bill Baker 2475 Oracle Corporation 2476 5300 Riata Park Ct. 2477 Austin, TX 78727 2478 US 2480 Phone: +1 512 401 1081 2481 Email: bill.baker@oracle.com