NFSv4 Working Group J. Lentini Internet-Draft C. Everhart Intended status: Standards Track NetApp Expires: August 24, 2009 D. Ellard BBN Technologies R. Tewari M. Naik IBM Almaden February 20, 2009 NSDB Protocol for Federated Filesystems draft-ietf-nfsv4-federated-fs-protocol-01 Status of this Memo This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on August 24, 2009. Copyright Notice Copyright (c) 2009 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Lentini, et al. Expires August 24, 2009 [Page 1] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 Abstract This document describes a file system federation protocol that enables file access and namespace traversal across collections of independently administered fileservers. The protocol specifies a set of interfaces by which fileservers with different administrators can form a fileserver federation that provides a namespace composed of the filesystems physically hosted on and exported by the constituent fileservers. Table of Contents 1. Requirements notation . . . . . . . . . . . . . . . . . . . . 4 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.1. Protocol Goals . . . . . . . . . . . . . . . . . . . . . . 5 3. Overview of Features and Concepts . . . . . . . . . . . . . . 7 3.1. Namespace . . . . . . . . . . . . . . . . . . . . . . . . 7 3.2. Fileset and Fileset Name (FSN) . . . . . . . . . . . . . . 7 3.3. Fileset Location (FSL) . . . . . . . . . . . . . . . . . . 8 3.3.1. Mutual Consistency across Fileset Locations . . . . . 8 3.4. Namespace Database (NSDB) . . . . . . . . . . . . . . . . 9 3.5. Mount Points, Junctions and Referrals . . . . . . . . . . 10 3.6. Unified Namespace and the Root Fileset . . . . . . . . . . 10 3.7. Fileservers . . . . . . . . . . . . . . . . . . . . . . . 11 3.8. File-access Clients . . . . . . . . . . . . . . . . . . . 11 4. Interaction with client protocols . . . . . . . . . . . . . . 12 4.1. Interaction with NFSv4 and NFSv4.1 . . . . . . . . . . . . 12 4.2. Interaction with other distributed file system protocols . . . . . . . . . . . . . . . . . . . . . . . . 12 5. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 5.1. Create a Fileset and its FSL(s) . . . . . . . . . . . . . 13 5.1.1. Creating a Fileset and an FSN . . . . . . . . . . . . 13 5.1.2. Adding a Replica of a Fileset . . . . . . . . . . . . 14 5.2. Junction Resolution . . . . . . . . . . . . . . . . . . . 14 5.3. Example use case for fileset annotations . . . . . . . . . 15 6. Mapping the NSDB onto LDAP . . . . . . . . . . . . . . . . . . 16 6.1. Basic LDAP Configuration . . . . . . . . . . . . . . . . . 16 6.2. LDAP Attributes . . . . . . . . . . . . . . . . . . . . . 16 6.2.1. fedfsUuid . . . . . . . . . . . . . . . . . . . . . . 16 6.2.2. fedfsNetAddr . . . . . . . . . . . . . . . . . . . . . 17 6.2.3. fsnUuid . . . . . . . . . . . . . . . . . . . . . . . 17 6.2.4. nsdbName . . . . . . . . . . . . . . . . . . . . . . . 18 6.2.5. fslHost . . . . . . . . . . . . . . . . . . . . . . . 18 6.2.6. fslPath . . . . . . . . . . . . . . . . . . . . . . . 18 6.2.7. fslUuid . . . . . . . . . . . . . . . . . . . . . . . 19 6.2.8. type . . . . . . . . . . . . . . . . . . . . . . . . . 19 6.2.9. currency . . . . . . . . . . . . . . . . . . . . . . . 19 Lentini, et al. Expires August 24, 2009 [Page 2] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 6.2.10. info . . . . . . . . . . . . . . . . . . . . . . . . . 20 6.2.11. annotation . . . . . . . . . . . . . . . . . . . . . . 20 6.2.12. childFsnUuid . . . . . . . . . . . . . . . . . . . . . 20 6.2.13. childNsdbName . . . . . . . . . . . . . . . . . . . . 21 6.3. LDAP Objects . . . . . . . . . . . . . . . . . . . . . . . 21 6.3.1. FsnObject . . . . . . . . . . . . . . . . . . . . . . 21 6.3.2. FslObject . . . . . . . . . . . . . . . . . . . . . . 22 7. NSDB Protocol Operations . . . . . . . . . . . . . . . . . . . 23 7.1. Administrative NSDB Operations . . . . . . . . . . . . . . 23 7.1.1. Creating an FSN . . . . . . . . . . . . . . . . . . . 24 7.1.2. Deleting an FSN . . . . . . . . . . . . . . . . . . . 25 7.1.3. Create an FSL . . . . . . . . . . . . . . . . . . . . 25 7.1.4. Delete an FSL . . . . . . . . . . . . . . . . . . . . 26 7.1.5. Update an FSL . . . . . . . . . . . . . . . . . . . . 26 7.2. Fileserver to NSDB Operations . . . . . . . . . . . . . . 27 7.2.1. Looking up FSLs for an FSN . . . . . . . . . . . . . . 27 8. Security Considerations . . . . . . . . . . . . . . . . . . . 28 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 29 10. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 30 11. Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 12. References . . . . . . . . . . . . . . . . . . . . . . . . . . 34 12.1. Normative References . . . . . . . . . . . . . . . . . . . 34 12.2. Informational References . . . . . . . . . . . . . . . . . 35 Appendix A. Acknowledgments . . . . . . . . . . . . . . . . . . . 36 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 37 Lentini, et al. Expires August 24, 2009 [Page 3] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 1. Requirements notation The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119]. Lentini, et al. Expires August 24, 2009 [Page 4] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 2. Introduction A federated filesystem enables file access and namespace traversal in a uniform, secure and consistent manner across multiple independent fileservers within an enterprise (and possibly across multiple enterprises) with reasonably good performance. Traditionally, building a namespace that spans multiple fileservers has been difficult for two reasons. First, the fileservers that export pieces of the namespace are often not in the same administrative domain. Second, there is no standard mechanism for the fileservers to cooperatively present the namespace. Fileservers may provide proprietary management tools and in some cases an administrator may be able to use the proprietary tools to build a shared namespace out of the exported filesystems. Relying on vendor- proprietary tools does not work in larger enterprises or when collaborating across enterprises because it is likely that the system will contain fileservers running different software, each with their own protocols, with no common protocol to manage the namespace or exchange namespace information. The filesystem federation protocol described below allows fileservers from different vendors and/or with different administrators to cooperatively build a namespace. The scope of the filesystem federation protocol is currently limited to NFSv4 [RFC3530] or NFSv4.1 [NFSv4.1] capable fileservers. The federation protocols have been designed to accommodate other file access protocols in the future. The requirements for federated namespaces are described in [FEDFS-REQTS]. 2.1. Protocol Goals The objective of this draft is to specify a set of protocols by which fileservers, possibly with different administrators, can form a fileserver federation that provides a namespace composed of the filesystems physically hosted on and exported by the fileservers of the federation. It should be possible, using systems that implement the federation protocols, to share a common namespace across all the fileservers in the federation. It should also be possible for the federation to project multiple namespaces and enable clients to traverse each one. Such a federation may contain an arbitrary number of namespace repositories, each belonging to a different administrative entity, and each rendering a part of the namespace. Such a federation may also have an arbitrary number of administrative entities responsible for administering disjoint subsets of the Lentini, et al. Expires August 24, 2009 [Page 5] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 fileservers. In the rest of the document the term fileserver implies a fileserver that is part of the federation. Lentini, et al. Expires August 24, 2009 [Page 6] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 3. Overview of Features and Concepts 3.1. Namespace The goal of a unified namespace is to make all managed data available to all clients via the same path in a common filesystem-like namespace. This should be achieved with minimal or zero client configuration. In particular, updates to the common namespace should not require configuration changes at the client. Filesets, which are the unit of data management, are a set of files and directories. From the perspective of the clients, the common namespace is constructed by mounting filesets that are physically located on different fileservers. The namespace, which is defined in terms of fileset definitions, fileset identifiers, the location of each fileset in the namespace, and the physical location of the implementation(s) of each fileset, is stored in a set of namespace repositories, each managed by an administrative entity. The namespace schema defines the model used for populating, modifying, and querying the namespace repositories. It is not required by the federation that the namespace be common across all fileservers. It should be possible to have several independently rooted namespaces. 3.2. Fileset and Fileset Name (FSN) A fileset is defined to be a container of data and is the basic unit of data management. Depending on the implementation, they may be anything between an individual directory of an exported filesystem to an entire exported filesystem at a fileserver. A fileset is uniquely represented by its fileset name (FSN). An FSN is considered unique across the federation. An FSN contains information sufficient to locate the namespace database (NSDB) that holds authoritative information about it and an identifier, called the FsnUuid, that identifies it on that NSDB. After an FSN is created, it is associated with a fileset location (FSL) on a fileserver. A fileset can be implemented by one or more FSLs. The attributes of an FSN are: NsdbName: the fully qualified domain name of an NSDB location that contains authoritative information for this FSN. FsnUuid: a 128-bit UUID (universally unique identifier), conforming to [RFC4122], that is used to uniquely identify an FSN. An NSDB SHOULD ensure that no two FSNs it stores have the same FsnUuid. Lentini, et al. Expires August 24, 2009 [Page 7] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 3.3. Fileset Location (FSL) An FSL represents the location where the fileset data resides. Each FSL contains a host:path pair on a file server. An FSL may also have additional attributes. Each location has an associated type that determines the protocol(s) that may be used to access its data. Type information can be used to decide the list of locations that will be returned to the client. Other attributes associated with an FSL are based on the NFSv4.1 fs_locations_info attribute [NFSv4.1]. Each FSL consists of: FslHost: the fully qualified domain name of the host fileserver storing the physical data FslPathname: the exported pathname at that host fileserver FslUuid: the 128-bit UUID of the FSL Type: the protocol that should be used to access this FSL (e.g. NFSv4 or NFSv4.1) Currency: the time lag of this FSL represented as the number of time units it lags the latest version as defined by the NFSv4.1 fs_locations_server's fls_currency field. A currency value of 0 represents the latest version. Currency values are less than or equal to zero Info: as defined in the NFSv4.1 fs_locations_info attribute Annotations: a list of name/value pairs that can be interpreted by a fileserver and used to generate a referral. The semantics of the name/value pair is not defined by this protocol and is intended to be used by higher-level protocols. This field MAY be used to store the NFSv4.1 fl_locations_server's fls_info values 3.3.1. Mutual Consistency across Fileset Locations All of the FSLs that have the same FSN (thereby reference the same fileset) are equivalent from the point of view of client access; the different locations of a fileset represent the same data, though potentially at different points in time. Fileset locations are equivalent but not identical. Locations may either be read-only or read-write. Typically, multiple read-write locations are backed by a clustered filesystem while read-only locations are replicas created by a federation-initiated or external replication. Read-only locations may represent consistent point-in-time copies of a read- write location. The federation protocols, however, cannot prevent Lentini, et al. Expires August 24, 2009 [Page 8] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 subsequent changes to a read-only location nor guarantee point-in- time consistency of a read-only location if the read-write location is changing. Regardless of the type, all locations exist at the same mount point in the namespace and, thus, one client may be referred to one location while another is directed to a different location. Since updates to each fileset location are not controlled by the federation protocol, it is the responsibility of administrators to guarantee the functional equivalence of the data. The federation protocol does not guarantee that the different locations are mutually consistent in terms of the currency of the data. It relies on the client file-access protocol (i.e., NFSv4) to contain sufficient information to help the clients determine the currency of the data at each location in order to ensure that the clients do not revert back in time when switching locations. 3.4. Namespace Database (NSDB) The NSDB service is a federation-wide service that provides interfaces to define, update, and query FSN information and FSN to FSL mapping information. An individual repository of namespace information is called an NSDB location. Each NSDB location is managed by a single administrative entity. A single admin entity can manage multiple NSDB locations. The difference between the NSDB service and an NSDB location is analogous to that between the DNS service and a particular DNS server. Each NSDB location stores the definition of the FSNs for which it is authoritative. It also stores the definitions of the FSLs associated with those FSNs. An NSDB location is authoritative for the filesets that it defines. An NSDB location can cache information from a peer NSDB location. The fileserver can always contact a local NSDB location (if it has been defined) or directly contact any NSDB location to resolve a junction. Each NSDB location supports an LDAP [RFC4510] interface and can be accessed by an LDAP client. An NSDB MAY be replicated throughout the federation. If an NSDB is replicated, the NSDB MUST exhibit loose, converging consistency as defined in [RFC3254]. The mechanism by which this is achieved is outside the scope of this document. Many LDAP implementations support replication. These features MAY be used to replicate the NSDB. Lentini, et al. Expires August 24, 2009 [Page 9] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 3.5. Mount Points, Junctions and Referrals A mount point is a directory in a parent fileset where a target fileset may be attached. If a client traverses the path leading from the root of the namespace to the mount point of a target fileset it should be able to access the data in that target fileset (assuming appropriate permissions). The directory where a fileset is mounted is represented by a junction in the underlying filesystem. In other words, a junction can be viewed as a reference from a directory in one fileset to the root of the target fileset. A junction can be implemented as a special marker on a directory that is interpreted by the fileserver as a mount point, or by some other mechanism in the underlying file system. What data is used by the underlying file system to represent the junction is not defined by this protocol. The essential property is that the server must be able to find, given the junction, the FSN for the target fileset. The mechanism by which the server maps a junction to an FSN is outside the scope of this document. The FSN (as described earlier) contains both the the authoritative NSDB location and the FsnUuid (a UUID for the fileset). When a client traversal reaches a junction, the client is referred to a list of FSLs associated with the FSN that was the target of the junction. The client can then redirect its connection to one of the FSLs. This act is called a referral. For NFSv4 clients, the FSL information is returned in the fs_locations or fs_locations_info attributes. The federation protocols do not limit where and how many times a fileset is mounted in the namespace. Filesets can be nested; a fileset can be mounted under another fileset. 3.6. Unified Namespace and the Root Fileset The root fileset, when defined, is the top-level fileset of the federation-wide namespace. The root of the unified namespace is the top level directory of this fileset. A set of designated fileservers in the federation can export the root fileset to render the federation-wide unified namespace. When a client mounts the root fileset from any of these designated fileservers it can view a common federation-wide namespace. The properties and schema definition of the root fileset and the protocol details that describe how to configure and replicate the root fileset are available in a companion draft [To be Published]. Lentini, et al. Expires August 24, 2009 [Page 10] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 3.7. Fileservers Fileservers are NFSv4 or NFSv4.1 servers that store the physical fileset data or fileservers that refer the client to other fileservers. A fileserver can be implemented in a number of different ways, including a single system, a cluster of systems, or some other configuration. 3.8. File-access Clients File access clients are standard off-the-shelf NAS clients that access file data using the NFSv4 protocol, the NFSv4.1 protocol, or some other protocol. Lentini, et al. Expires August 24, 2009 [Page 11] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 4. Interaction with client protocols 4.1. Interaction with NFSv4 and NFSv4.1 The federation protocol is compatible with the requirements of NFSv4 referral mechanisms as defined in [RFC3530] and the NFSv4.1 referral mechanisms as defined in [NFSv4.1]. 4.2. Interaction with other distributed file system protocols The federation protocol does not mandate a specific format for pathnames. Therefore it is possible to store the pathnames used by a variety of distributed file systems, including CIFS. Lentini, et al. Expires August 24, 2009 [Page 12] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 5. Examples In this section we provide examples and discussion of the basic operations facilitated by the federated file system protocol: creating a fileset, adding a replica of a fileset, resolving a junction, and creating a junction. 5.1. Create a Fileset and its FSL(s) A fileset is the abstraction of a set of files and their containing directory tree. The fileset abstraction is the fundamental unit of data management in the federation. This abstraction is implemented by an actual directory tree whose root location is specified by a fileset location (FSL). In this section, we describe the basic requirements for starting with a directory tree and creating a fileset that can be used in the federation protocols. Note that we do not assume that the process of creating a fileset requires any transformation of the files or the directory hierarchy. The only thing that is required by this process is assigning the fileset a fileset name (FSN) and expressing the location(s) of the implementation of the fileset as FSL(s). There are many possible variations to this procedure, depending on how the FSN that binds the FSL is created, and whether other replicas of the fileset exist, are known to the federation, and need to be bound to the same FSN. It is easiest to describe this in terms of how to create the initial implementation of the fileset, and then describe how to add replicas. 5.1.1. Creating a Fileset and an FSN 1. Choose the NSDB node that will keep track of the FSL(s) and related information for the fileset. 2. Request that the NSDB node register a new FSN for the fileset. The FSN UUID is chosen by the administrator or generated automatically by administration software. The former case is used if the fileset is being restored, perhaps as part of disaster recovery, and the administrator wishes to specify the FSN UUID in order to permit existing junctions that reference that FSN to work again. At this point, the FSN exists, but its fileset locations are unspecified. Lentini, et al. Expires August 24, 2009 [Page 13] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 3. Send the FSN, the hostname, the export path, the type, the currency, info, and annotations for the fileset to the NSDB node. The NSDB node records this info and creates the initial FSL for the fileset. 5.1.2. Adding a Replica of a Fileset Adding a replica is straightforward: the NSDB node and the FSN are already known. The only remaining step is to add another FSL. Note that the federation protocols do not include methods for creating or managing replicas: this is assumed to be a platform- dependent operation (at least at this time). The only interface required is the ability to register or remove the registration of replicas for a fileset. 5.2. Junction Resolution A fileset may contain references to other filesets. These references are represented by junctions. If a client requests access to a fileset object that is a junction, the server resolves the junction to discover the FSL(s) that implements the referenced fileset. There are many possible variations to this procedure, depending on how the junctions are represented and how the information necessary to perform resolution is represented by the server. Step 4 is the only step that interacts directly with the federation protocols. The rest of the steps may use platform-specific interfaces. 1. The server determines that the object being accessed is a junction. 2. The server does a local lookup to find the FSN of the target fileset. 3. Using the FSN, the server finds the NSDB node responsible for the target object. 4. The server contacts that NSDB node and asks for the set of FSLs that implement the target FSN. The NSDB node responds with a set of FSLs. Lentini, et al. Expires August 24, 2009 [Page 14] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 5.3. Example use case for fileset annotations The fileset annotations can be used to define relationships between filesets that can be used by an auxiliary replication protocol. Consider the scenario where a fileset is created and mounted at some point in the namespace. A snapshot of the read-write FSL of that fileset is taken periodically at different frequencies say a daily snapshot or a weekly snapshot. The different snapshots are mounted at different locations in the namespace. The daily snapshots are considered as a different fileset from the weekly ones but both are related to the source fileset. For this we can define an annotation labeling the filesets as source and replica. The replication protocol can use this information to copy data from one or more FSLs of the source fileset to all the FSLs of the replica fileset. The replica filesets are read-only while the source fileset is read- write. This follows the traditional AFS model of mounting the read-only volume at a path in the namespace different from that of the read- write volume. The federation protocol does not control or manage the relationship among filesets. It merely enables annotating the filesets with user- defined relationships. Lentini, et al. Expires August 24, 2009 [Page 15] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 6. Mapping the NSDB onto LDAP This section describes the LDAP schema used to define the LDAP implementation of the NSDB service. The first part of the section describes the basic properties of the LDAP configuration that MUST be used in order to ensure compatibility between different implementations. The second section defines the new LDAP attribute types and the subsequent sections describe the new object types and specify how the distinguished name (DN) of each object instance MUST be constructed. 6.1. Basic LDAP Configuration The base name (or suffix) for all DNs used by the NSDB schema is "dc=fed-fs,dc=com". The DN of the privileged LDAP user is, by convention, "cn=admin,dc=fed-fs,dc=com". This user is able to modify the contents of the LDAP database. It is permitted to use a different DN (or add additional privileged users) but if a different DN is used then every admin entity that needs to modify the contents of the database or view privileged information must be made aware of the new DN. It MUST be possible for the anonymous (unauthenticated) user to perform LDAP queries that access the NSDB data. All implementations SHOULD use the same schema, or, at minimum, a schema that includes all of the objects, with each of the attributes, named in the following sections. 6.2. LDAP Attributes This section describes the required attributes of the NSDB LDAP schema. 6.2.1. fedfsUuid A fedfsUuid is the base type for all of the universally unique identifiers (UUIDs) used by the federated file system protocols. This SHOULD be defined in terms of the text representation of the standard UUID (as defined in [RFC4122]). It MAY also be useful, for purposes of debugging or annotation, to permit a fedfsUuid to include members of a more general class of strings. Lentini, et al. Expires August 24, 2009 [Page 16] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 A fedfsUuid is a single-valued LDAP attribute. It is formally defined as follows: attributetype ( 1.3.6.1.4.1.31103.1.1 NAME 'fedfsUuid' DESC 'a UUID used by NSDB' SUP name SINGLE-VALUE ) 6.2.2. fedfsNetAddr A fedfsNetAddr is the locative name of a network service. It MUST be a UTF-8 string and represent a network location in either IPv4, IPv6, or DNS host name notation. The format is the same as that specified for an fs_location4's server array elements in section 11.9 of [NFSv4.1]. This attribute is single-valued. It is formally defined as follows: attributetype ( 1.3.6.1.4.1.31103.1.2 NAME 'fedfsNetAddr' DESC 'a network name of a host or service' SUP name SINGLE-VALUE ) 6.2.3. fsnUuid A fsnUuid represents the fsnUuid component of an FSN. The fsnUuid is a subclass of fedfsUuid. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.3 NAME 'fsnUuid' DESC 'the FSN UUID component of an FSN' SUP fedfsUuid SINGLE-VALUE ) Lentini, et al. Expires August 24, 2009 [Page 17] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 6.2.4. nsdbName An nsdbName is the NSDB component of an FSN. The nsdbName attribute is a subclass of fedfsNetAddr. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.4 NAME 'nsdbName' DESC 'the NSDB location component of an FSN' SUP fedfsNetAddr SINGLE-VALUE ) 6.2.5. fslHost An fslHost is the hostname/port component of an FSL. The fslHost attribute is a subclass of fedfsNetAddr. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.5 NAME 'fslHost' DESC 'service location for a fileset server' SUP fedfsNetAddr SINGLE-VALUE ) 6.2.6. fslPath The path component of an FSL. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.6 NAME 'fslPath' DESC 'server-local path to a fileset' SUP name SINGLE-VALUE ) Lentini, et al. Expires August 24, 2009 [Page 18] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 6.2.7. fslUuid Each FSL must have a UUID associated with it, which serves as part of its DN. The fslUuid attribute is a subclass of fedfsUuid. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.7 NAME 'fslUuid' DESC 'UUID of an FSL' SUP fedfsUuid SINGLE-VALUE ) 6.2.8. type The type of an FSL. This attribute is used to specify the distribute file system protocol that can be used to access an FSL. The following values are defined for this field: nfsv4 : the FSL is accessible via the NFSv4 protocol. Values for other protocols may be defined at a later time. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.8 NAME 'type' DESC 'CIFS, NFS, etc' SUP name ) 6.2.9. currency The currency of an FSL. This attribute is used to populate the NFSv4.1 fs_locations_info's currency field. This attribute is single-valued. Lentini, et al. Expires August 24, 2009 [Page 19] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 attributetype ( 1.3.6.1.4.1.31103.1.9 NAME 'currency' DESC 'up-to-date measure of the data' SUP name ) 6.2.10. info Information about the FSL. This attribute is used to populate the NFSv4.1 fs_locations_info's info field. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.10 NAME 'info' DESC 'information about the file system' SUP name ) 6.2.11. annotation An annotation of an NSDB object. This attribute is multi-valued; an object type that permits annotations may have any number of annotations per instance. attributetype ( 1.3.6.1.4.1.31103.1.11 NAME 'annotation' DESC 'annotation of an NSDB object' SUP name ) 6.2.12. childFsnUuid The fsnUuid of the target of a junction. The childFsnUuid attribute is a subclass of fsnUuid. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.12 NAME 'childFsnUuid' DESC 'the fsnUuid of a Junction's target' Lentini, et al. Expires August 24, 2009 [Page 20] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 SUP fsnUuid SINGLE-VALUE ) 6.2.13. childNsdbName The nsdbName of the target of a junction. The childNsdbName attribute is a subclass of nsdbName. This attribute is single-valued. attributetype ( 1.3.6.1.4.1.31103.1.13 NAME 'childNsdbName' DESC 'the nsdbName of a Junction's target' SUP nsdbName SINGLE-VALUE ) 6.3. LDAP Objects 6.3.1. FsnObject An FsnObject represents an FSN. The required attributes of an FsnObject are an nsdbName and fsnUuid. The DN of an FSN is assumed to take the following form: "fsnUuid=FSNUUID,dc=fed-fs,dc=com", where fsnUuid is the UUID of the FSN. An FsnObject MAY also have additional attributes, but these attributes MUST NOT be referenced by any part of this draft. objectclass ( 1.3.6.1.4.1.31103.1.1001 NAME 'FsnObject' DESC 'Representing a Fed-fs Fileset' SUP top STRUCTURAL MUST ( fsnUuid $ nsdbName ) MAY ( descr $ annotation )) Lentini, et al. Expires August 24, 2009 [Page 21] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 6.3.2. FslObject An FslObject represents an FSL. The required attributes of an FslObject are an nsdbName, fsnUuid, fslHost, fslPath, fslUuid, type, currency, info, and annotations. An FslObject's currency and annotations attributes MAY be null. The DN of an FSL is required to take the following form: "fslUuid=UUID,fsnUuid=FSNUUID,dc=fed-fs,dc=com". To find all the FSLs that match a given FSN, query for the children of the object with DN "fsnUuid=FSNUUID,dc=fed-fs,dc=com" with a filter for "objectType = fslObject". (If you want to be doubly careful, you can also filter by the nsdbName.) objectclass ( 1.3.6.1.4.1.31103.1.1002 NAME 'FslObject' DESC 'A physical instance of a fileset' SUP fsnObject STRUCTURAL MUST ( fsnUuid $ nsdbName $ fslHost $ fslPath $ fslUuid ) MAY ( descr $ annotation )) Lentini, et al. Expires August 24, 2009 [Page 22] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 7. NSDB Protocol Operations The operations defined by the protocol can be described as several sub-protocols that are used by entities within the federation to perform different roles. The first of these sub-protocols defines how the state of an NSDB location can be initialized and updated. The primary use of this sub-protocol is by an administrator to add, edit, or delete filesets, their properties, and their fileset locations. The second of these sub-protocols defines the queries that are sent to an NSDB location in order to perform resolution (or find other information about the information stored within that NSDB location) and the responses returned by the NSDB location. The primary use of this sub-protocol is by a fileset server in order to perform resolution, but it may also be used by an administrator to query the state of the system. The first and second sub-protocols are defined as LDAP operations, using the schema defined in the previous section. If each NSDB location is a standard LDAP server, then, in theory, it is unnecessary to describe the LDAP operations in detail, because the operations are ordinary LDAP operations to query and update records. However, we do not require that an NSDB location implement a complete LDAP service, and therefore we define in these sections the minimum level of LDAP functionality required to implement an NSDB location. The NSDB sub-protocols are defined in the next two sub-sections. The third sub-protocol defines the queries or other requests that are sent to a fileset server in order to get information from it or to modify the state of the fileset server in a manner related to the federation protocols. The primary purpose of this protocol is for an administrator to create or delete a junction or fileset or discover related information about a particular fileset server. The third sub-protocol is defined as ONC RPC operations. The reason for using a different RPC mechanism (instead of mapping these operations onto LDAP) is to minimize the changes required to the fileset server. The ONC RPC administration protocol is defined in [FEDFS-ADMIN]. 7.1. Administrative NSDB Operations The admin entity initiates and controls the commands to manage fileset and namespace information. The admin entity, however, is Lentini, et al. Expires August 24, 2009 [Page 23] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 stateless. All state is maintained at the NSDB locations or at the fileserver. We require that each NSDB location be able to act as an LDAP server and that the protocol used for communicating between the admin entity and each NSDB is LDAP. The names we assign to these operations are entirely for the purpose of exposition in this document, and are not part of the LDAP dialogs. In the description of the LDAP messages and LDIF, we use the following notation: constant strings and literal names are specified in lower or mixed case, while variables or values are specified in uppercase. One important exception to this rule is that the names of the error codes follow the convention (used widely in other protocols, including NFS) of having names that are entirely uppercase. 7.1.1. Creating an FSN The administrator uses this operation to create a new FSN by requesting the NSDB to create a new FsnObject in its LDAP database with an fsnUuid of FSNUUID and an NsdbName of NSDB. The NSDB location that receives the request SHOULD check that the NSDB matches its own value and return an error if it does not. This is to ensure that an FSN is always created by the NSDB location encoded within the FSN as its owner. The NSDB location that receives the request SHOULD check all of the attributes for validity and consistency, but this is not generally possible for LDAP servers because the consistency requirements cannot be expressed in the LDAP schema (although many LDAP servers can be extended, via plug-ins or other mechanisms, to add functionality beyond the strict definition of LDAP). 7.1.1.1. LDAP Request The admin chooses the fsnUuid and NsdbName of the FSN. The fsnUuid is a UUID and should be chosen via a standard process for creating a UUID (described in [RFC4122]). The NsdbName is the name of the NSDB location that will serve as the source of definitive information about an FSN for the life of that FSN. In the example below, the admin server chooses a fsnUuid of FSNUUID and the NsdbName of NSDB and then sends an LDAP ADD request, described by the LDIF below, to the NSDB location NSDB. This will create a new FsnObject on that NSDB location with the given attributes in the LDAP database. Lentini, et al. Expires August 24, 2009 [Page 24] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 dn: fsnUuid=FSNUUID,dc=fed-fs,dc=com changeType: add objectClass: FsnObject fsnUuid: FSNUUID nsdbName: NSDB 7.1.2. Deleting an FSN Deletes the given fileset name. This assumes that all the FSLs related to that FSN have already been deleted. If FSL records for this FSN still exist in the database of the NSDB that receives this request, then this function MUST return an error. Note that the FSN delete function only removes the fileset from the namespace (by removing the records for that FSN from the NSDB location that receives this request). The fileset and its data are not deleted. Any junction that has this FSN as its target may continue to point to this non-existent FSN. A dangling reference may be detected when a client tries to resolve the target of a junction that refers to the deleted FSN and the NSDB returns an error. 7.1.2.1. LDAP Request The admin sends an LDAP DELETE request to the NSDB server to remove the FsnObject from the NSDB server. An example LDIF for the delete request is shown below. dn: fsnUuid=FSNUUID,dc=fed-fs,dc=com changeType: delete 7.1.3. Create an FSL Creates a new Fileset location at the given location denoted by HOST and PATH for the given FSN. Normally an FSL is identified by the HOST:PATH pair. A UUID is an optional way to identify an FSL if it is recovered to a different HOST:PATH after a backup/restore. The FSL create command will result in the admin server sending an LDAP ADD request to create a new FslObject at the NSDB maintaining the given FSN. The example LDIF is shown below. The PATH is the pathname where the fileset is located on that host. 7.1.3.1. LDAP Request The admin sends an LDAP ADD request to the NSDB server to add the FLS. Lentini, et al. Expires August 24, 2009 [Page 25] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 dn:fslUuid=UUID,fsnUuid=FSNUUID,dc=fed-fs,dc=com changeType: add objectClass: FslObject fsnUuid: FSNUUID nsdbName: NSDB fslUuid: UUID fslHost: HOST fslPath: PATH type: file access protocol type (e.g. nfs4) currency: CURRENCY info: INFO annotation: ANNOTATION 7.1.4. Delete an FSL Deletes the given Fileset location. The admin requests the NSDB location storing the FslObject to delete it from its database. This operation does not result in the fileset location's data being deleted at the fileserver. 7.1.4.1. LDAP Request The admin sends an LDAP DELETE request to the NSDB server to remove the FLS. dn: fslUuid=UUID,fsnUuid=FSNUUID,dc=fed-fs,dc=com changeType: delete 7.1.5. Update an FSL Update the attributes of a given FSL. This command results in a change in the attributes of the FslObject at the NSDB server maintaining this FSL. The attributes that must not change are the fslUuid and the fsnUuid of the fileset this FSL implements. 7.1.5.1. LDAP Request The admin sends an LDAP MODIFY request to the NSDB server to update the FLS. dn: fslUuid=UUID,fsnUuid=FSNUUID,dc=fed-fs,dc=com changeType: modify replace: ATTRIBUTE-TYPE Lentini, et al. Expires August 24, 2009 [Page 26] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 7.2. Fileserver to NSDB Operations 7.2.1. Looking up FSLs for an FSN This operation returns the list of FSLs for the given FSN. The FSN's fsnUuid is used as the search key. The fileserver will convert the list of FSLs to the NFSv4 fs_locations or NFSv4.1 fs_locations_info. The filter may also specify the type of protocol (v4, v4.1), or type of data access (ro, rw).
LDAP Request Search base: fsnUuid=FSNUUID, dc=fed-fs, dc=com Search scope: onelevel Search filter: (objectClass=FslObject) The server can scan through the results and find results whose type corresponds to the type of the client on whose behalf the server is performing the request, extracting the fslHost and fslPath (and possibly additional attributes) and using them to create an fs_locations list or fs_locations_info list that the client can use. Lentini, et al. Expires August 24, 2009 [Page 27] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 8. Security Considerations Both LDAP and NFSv4/NFSv4.1 provide security mechanisms. When used in conjunction with the federated file system protocols described in this document, the use of these mechanisms is RECOMMENDED. Specifically, the use of RPCSEC_GSS [RFC2203] [RFC2743] is RECOMMENDED on all connections between a client and fileserver. For all LDAP connections established by the federated file system protocols, TLS [RFC5246] [RFC4513] is RECOMMENDED. Within a federation, there are two components that an attacker may be able to compromise: a fileserver and an NSDB. If an attacker compromises a fileserver, the attacker can interfere with the client's file system I/O operations (e.g. by returning fictitious data in the response to a read request) or fabricating a referral. The attackers abilities are the same regardless of whether or not the federation protocols are in use. If an attacker compromises an NSDB, the attacker will be able to forge FSL information and thus poison the fileserver's referral information. Therefore an NSDB should be as secure as the fileservers which query it. It should be noted that the federation protocols do not directly provide access to file system data. The federation protocols only provide a mechanism for building a namespace. All data transfers are occur between a client and server just as they would if the federation protocols were not in use. As a result, the federation protocols do not require new user authentication and authorization mechanisms or require a file server to act as a proxy for a client. Lentini, et al. Expires August 24, 2009 [Page 28] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 9. IANA Considerations This document has no actions for IANA. Lentini, et al. Expires August 24, 2009 [Page 29] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 10. Conclusions The federated filesystem protocol manages multiple independently administered fileservers to share namespace and referral information to enable clients to traverse seamlessly across them. Lentini, et al. Expires August 24, 2009 [Page 30] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 11. Glossary Administrator: user with the necessary authority to initiate administrative tasks on one or more servers. Admin entity: A server or agent that administers a collection of fileservers and persistently stores the namespace information. Client: Any client that accesses the fileserver data using a supported filesystem access protocol. Federation: A set of server collections and singleton servers that use a common set of interfaces and protocols in order to provide to their clients a federated namespace accessible through a filesystem access protocol. Fileserver: A server exporting a filesystem via a network filesystem access protocol. Fileset: The abstraction of a set of files and their containing directory tree. A fileset is the fundamental unit of data management in the federation. Note that all files within a fileset are descendants of one directory, and that filesets do not span filesystems. Filesystem: A self-contained unit of export for a fileserver, and the mechanism used to implement filesets. The fileset does not need to be rooted at the root of the filesystem, nor at the export point for the filesystem. A single filesystem MAY implement more than one fileset, if the client protocol and the fileserver permit this. Filesystem access protocol: A network filesystem access protocol such as NFSv2 [RFC1094], NFSv3 [RFC1813], NFSv4 [RFC3530], or CIFS. FSL (Fileset location): The location of the implementation of a fileset at a particular moment in time. A FSL MUST be something that can be translated into a protocol-specific description of a resource that a client can access directly, such as a fs_location (for NFSv4), or share name (for CIFS). Note that not all FSLs need to be explicitly exported as long as they are contained within an exported path on the fileserver. Lentini, et al. Expires August 24, 2009 [Page 31] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 FSN (Fileset name): A platform-independent and globally unique name for a fileset. Two FSLs that implement replicas of the same fileset MUST have the same FSN, and if a fileset is migrated from one location to another, the FSN of that fileset MUST remain the same. Junction: A filesystem object used to link a directory name in the current fileset with an object within another fileset. The server-side "link" from a leaf node in one fileset to the root of another fileset. Namespace: A filename/directory tree that a sufficiently-authorized client can observe. NSDB (Namespace Database Service): A service that maps FSNs to FSLs. The NSDB may also be used to store other information, such as annotations for these mappings and their components. NSDB Node: The name or location of a server that implements part of the NSDB service and is responsible for keeping track of the FSLs (and related info) that implement a given partition of the FSNs. Referral: A server response to a client access that directs the client to evaluate the current object as a reference to an object at a different location (specified by an FSL) in another fileset, and possibly hosted on another fileserver. The client re-attempts the access to the object at the new location. Replica: A replica is a redundant implementation of a fileset. Each replica shares the same FSN, but has a different FSL. Replicas may be used to increase availability or performance. Updates to replicas of the same fileset MUST appear to occur in the same order, and therefore each replica is self-consistent at any moment. We do not assume that updates to each replica occur simultaneously If a replica is offline or unreachable, the other replicas may be updated. Server Collection: A set of fileservers administered as a unit. A server collection may be administered with vendor-specific software. The namespace provided by a server collection could be part of the federated namespace. Lentini, et al. Expires August 24, 2009 [Page 32] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 Singleton Server: A server collection containing only one server; a stand-alone fileserver. Lentini, et al. Expires August 24, 2009 [Page 33] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 12. References 12.1. Normative References [FEDFS-ADMIN] J. Lentini, et al., "Administration Protocol for Federated Filesystems (Work In Progress)", draft-ietf-nfsv4-federated-fs-admin , 2008. [FEDFS-REQTS] J. Lentini, et al., "Requirements for Federated File Systems (Work In Progress)", draft-ietf-nfsv4-federated-fs-reqts , 2008. [NFSv4.1] S. Shepler, et al., "NFS Version 4 Minor Version 1 (Work In Progress)", draft-ietf-nfsv4-minorversion1 , 2008. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC2203] Eisler, M., Chiu, A., and L. Ling, "RPCSEC_GSS Protocol Specification", RFC 2203, September 1997. [RFC2743] Linn, J., "Generic Security Service Application Program Interface Version 2, Update 1", RFC 2743, January 2000. [RFC3530] Shepler, S., Callaghan, B., Robinson, D., Thurlow, R., Beame, C., Eisler, M., and D. Noveck, "Network File System (NFS) version 4 Protocol", RFC 3530, April 2003. [RFC4122] Leach, P., Mealling, M., and R. Salz, "A Universally Unique IDentifier (UUID) URN Namespace", RFC 4122, July 2005. [RFC4510] Zeilenga, K., "Lightweight Directory Access Protocol (LDAP): Technical Specification Road Map", RFC 4510, June 2006. [RFC4513] Harrison, R., "Lightweight Directory Access Protocol (LDAP): Authentication Methods and Security Mechanisms", RFC 4513, June 2006. [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security (TLS) Protocol Version 1.2", RFC 5246, August 2008. Lentini, et al. Expires August 24, 2009 [Page 34] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 12.2. Informational References [RFC1094] Nowicki, B., "NFS: Network File System Protocol specification", RFC 1094, March 1989. [RFC1813] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS Version 3 Protocol Specification", RFC 1813, June 1995. [RFC3254] Alvestrand, H., "Definitions for talking about directories", RFC 3254, April 2002. Lentini, et al. Expires August 24, 2009 [Page 35] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 Appendix A. Acknowledgments We would like to thank Paul Lemahieu of EMC, Robert Thurlow of Sun Microsystems, and Mario Wurzl of EMC for helping to author this document. Lentini, et al. Expires August 24, 2009 [Page 36] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 Authors' Addresses James Lentini NetApp 1601 Trapelo Rd, Suite 16 Waltham, MA 02451 US Phone: +1 781-768-5359 Email: jlentini@netapp.com Craig Everhart NetApp 7301 Kit Creek Rd Research Triangle Park, NC 27709 US Phone: +1 919-476-5320 Email: everhart@netapp.com Daniel Ellard BBN Technologies 10 Moulton Street Cambridge, MA 02138 US Phone: +1 617-873-8000 Email: dellard@bbn.com Renu Tewari IBM Almaden 650 Harry Rd San Jose, CA 95120 US Email: tewarir@us.ibm.com Lentini, et al. Expires August 24, 2009 [Page 37] Internet-Draft NSDB Protocol for Federated Filesystems February 2009 Manoj Naik IBM Almaden 650 Harry Rd San Jose, CA 95120 US Email: manoj@almaden.ibm.com Lentini, et al. Expires August 24, 2009 [Page 38]