5 .. versionadded:: Jewel
7 Ceph Object Gateway namespaces can now be exported over file-based
8 access protocols such as NFSv3 and NFSv4, alongside traditional HTTP access
9 protocols (S3 and Swift).
11 In particular, the Ceph Object Gateway can now be configured to
12 provide file-based access when embedded in the NFS-Ganesha NFS server.
17 The librgw.so shared library (Unix) provides a loadable interface to
18 Ceph Object Gateway services, and instantiates a full Ceph Object Gateway
19 instance on initialization.
21 In turn, librgw.so exports rgw_file, a stateful API for file-oriented
22 access to RGW buckets and objects. The API is general, but its design
23 is strongly influenced by the File System Abstraction Layer (FSAL) API
24 of NFS-Ganesha, for which it has been primarily designed.
26 A set of Python bindings is also provided.
31 The implementation conforms to Amazon Web Services (AWS) hierarchical
32 namespace conventions which map UNIX-style path names onto S3 buckets
35 The top level of the attached namespace consists of S3 buckets,
36 represented as NFS directories. Files and directories subordinate to
37 buckets are each represented as objects, following S3 prefix and
38 delimiter conventions, with '/' being the only supported path
41 For example, if an NFS client has mounted an RGW namespace at "/nfs",
42 then a file "/nfs/mybucket/www/index.html" in the NFS namespace
43 corresponds to an RGW object "www/index.html" in a bucket/container
46 Although it is generally invisible to clients, the NFS namespace is
47 assembled through concatenation of the corresponding paths implied by
48 the objects in the namespace. Leaf objects, whether files or
49 directories, will always be materialized in an RGW object of the
50 corresponding key name, "<name>" if a file, "<name>/" if a directory.
51 Non-leaf directories (e.g., "www" above) might only be implied by
52 their appearance in the names of one or more leaf objects. Directories
53 created within NFS or directly operated on by an NFS client (e.g., via
54 an attribute-setting operation such as chown or chmod) always have a
55 leaf object representation used to store materialized attributes such
56 as Unix ownership and permissions.
61 The RGW NFS interface supports most operations on files and
62 directories, with the following restrictions:
64 - Links, including symlinks, are not supported
65 - NFS ACLs are not supported
67 + Unix user and group ownership and permissions *are* supported
69 - Directories may not be moved/renamed
71 + files may be moved between directories
73 - Only full, sequential *write* i/o is supported
75 + i.e., write operations are constrained to be **uploads**
76 + many typical i/o operations such as editing files in place will necessarily fail as they perform non-sequential stores
77 + some file utilities *apparently* writing sequentially (e.g., some versions of GNU tar) may fail due to infrequent non-sequential stores
78 + When mounting via NFS, sequential application i/o can generally be constrained to be written sequentially to the NFS server via a synchronous mount option (e.g. -osync in Linux)
79 + NFS clients which cannot mount synchronously (e.g., MS Windows) will not be able to upload files
84 The RGW NFS interface provides a hybrid security model with the
85 following characteristics:
87 - NFS protocol security is provided by the NFS-Ganesha server, as negotiated by the NFS server and clients
89 + e.g., clients can by trusted (AUTH_SYS), or required to present Kerberos user credentials (RPCSEC_GSS)
90 + RPCSEC_GSS wire security can be integrity only (krb5i) or integrity and privacy (encryption, krb5p)
91 + various NFS-specific security and permission rules are available
93 * e.g., root-squashing
95 - a set of RGW/S3 security credentials (unknown to NFS) is associated with each RGW NFS mount (i.e., NFS-Ganesha EXPORT)
97 + all RGW object operations performed via the NFS server will be performed by the RGW user associated with the credentials stored in the export being accessed (currently only RGW and RGW LDAP credentials are supported)
99 * additional RGW authentication types such as Keystone are not currently supported
101 Configuring an NFS-Ganesha Instance
102 ===================================
104 Each NFS RGW instance is an NFS-Ganesha server instance *embeddding*
105 a full Ceph RGW instance.
107 Therefore, the RGW NFS configuration includes Ceph and Ceph Object
108 Gateway-specific configuration in a local ceph.conf, as well as
109 NFS-Ganesha-specific configuration in the NFS-Ganesha config file,
115 Required ceph.conf configuration for RGW NFS includes:
117 * valid [client.radosgw.{instance-name}] section
118 * valid values for minimal instance configuration, in particular, an installed and correct ``keyring``
120 Other config variables are optional, front-end-specific and front-end
121 selection variables (e.g., ``rgw data`` and ``rgw frontends``) are
122 optional and in some cases ignored.
124 A small number of config variables (e.g., ``rgw_namespace_expire_secs``)
125 are unique to RGW NFS.
130 A strictly minimal ganesha.conf for use with RGW NFS includes one
131 EXPORT block with embedded FSAL block of type RGW::
135 Export_ID={numeric-id};
141 Transport_Protocols = TCP;
143 # optional, permit unsquashed access by client "root" user
144 #Squash = No_Root_Squash;
148 User_Id = {s3-user-id};
149 Access_Key_Id ="{s3-access-key}";
150 Secret_Access_Key = "{s3-secret}";
154 ``Export_ID`` must have an integer value, e.g., "77"
156 ``Path`` (for RGW) should be "/"
158 ``Pseudo`` defines an NFSv4 pseudo root name (NFSv4 only)
160 ``SecType = sys;`` allows clients to attach without Kerberos
163 ``Squash = No_Root_Squash;`` enables the client root user to override
164 permissions (Unix convention). When root-squashing is enabled,
165 operations attempted by the root user are performed as if by the local
166 "nobody" (and "nogroup") user on the NFS-Ganesha server
168 The RGW FSAL additionally supports RGW-specific configuration
169 variables in the RGW config section::
172 cluster = "{cluster name, default 'ceph'}";
173 name = "client.rgw.{instance-name}";
174 ceph_conf = "/opt/ceph-rgw/etc/ceph/ceph.conf";
175 init_args = "-d --debug-rgw=16";
178 ``cluster`` sets a Ceph cluster name (must match the cluster being exported)
180 ``name`` sets an RGW instance name (must match the cluster being exported)
182 ``ceph_conf`` gives a path to a non-default ceph.conf file to use
185 Other useful NFS-Ganesha configuration:
186 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
188 Any EXPORT block which should support NFSv3 should include version 3
189 in the NFS_Protocols setting. Additionally, NFSv3 is the last major
190 version to support the UDP transport. To enable UDP, include it in the
191 Transport_Protocols setting. For example::
196 Transport_Protocols = UDP,TCP;
200 One important family of options pertains to interaction with the Linux
201 idmapping service, which is used to normalize user and group names
202 across systems. Details of idmapper integration are not provided here.
204 With Linux NFS clients, NFS-Ganesha can be configured
205 to accept client-supplied numeric user and group identifiers with
206 NFSv4, which by default stringifies these--this may be useful in small
207 setups and for experimentation::
210 Allow_Numeric_Owners = true;
211 Only_Numeric_Owners = true;
217 NFS-Ganesha configuration problems are usually debugged by running the
218 server with debugging options, controlled by the LOG config section.
220 NFS-Ganesha log messages are grouped into various components, logging
221 can be enabled separately for each component. Valid values for
222 component logging include::
224 *FATAL* critical errors only
225 *WARN* unusual condition
226 *DEBUG* mildly verbose trace output
227 *FULL_DEBUG* verbose trace output
242 CACHE_INODE_LRU = FATAL;
244 HASHTABLE_CACHE = FATAL;
265 # optional: redirect log output
268 # destination = "/tmp/ganesha-rgw.log";
273 Running Multiple NFS Gateways
274 =============================
276 Each NFS-Ganesha instance acts as a full gateway endpoint, with the
277 limitation that currently an NFS-Ganesha instance cannot be configured
278 to export HTTP services. As with ordinary gateway instances, any
279 number of NFS-Ganesha instances can be started, exporting the same or
280 different resources from the cluster. This enables the clustering of
281 NFS-Ganesha instances. However, this does not imply high availability.
283 When regular gateway instances and NFS-Ganesha instances overlap the
284 same data resources, they will be accessible from both the standard S3
285 API and through the NFS-Ganesha instance as exported. You can
286 co-locate the NFS-Ganesha instance with a Ceph Object Gateway instance
292 Exporting an NFS namespace and other RGW namespaces (e.g., S3 or Swift
293 via the Civetweb HTTP front-end) from the same program instance is
294 currently not supported.
296 When adding objects and buckets outside of NFS, those objects will
297 appear in the NFS namespace in the time set by
298 ``rgw_nfs_namespace_expire_secs``, which defaults to 300 seconds (5 minutes).
299 Override the default value for ``rgw_nfs_namespace_expire_secs`` in the
300 Ceph configuration file to change the refresh rate.
302 If exporting Swift containers that do not conform to valid S3 bucket
303 naming requirements, set ``rgw_relaxed_s3_bucket_names`` to true in the
304 [client.radosgw] section of the Ceph configuration file. For example,
305 if a Swift container name contains underscores, it is not a valid S3
306 bucket name and will be rejected unless ``rgw_relaxed_s3_bucket_names``
309 Configuring NFSv4 clients
310 =========================
312 To access the namespace, mount the configured NFS-Ganesha export(s)
313 into desired locations in the local POSIX namespace. As noted, this
314 implementation has a few unique restrictions:
316 - NFS 4.1 and higher protocol flavors are preferred
318 + NFSv4 OPEN and CLOSE operations are used to track upload transactions
320 - To upload data successfully, clients must preserve write ordering
322 + on Linux and many Unix NFS clients, use the -osync mount option
324 Conventions for mounting NFS resources are platform-specific. The
325 following conventions work on Linux and some Unix platforms:
327 From the command line::
329 mount -t nfs -o nfsvers=4,noauto,soft,proto=tcp <ganesha-host-name>:/ <mount-point>
333 <ganesha-host-name>:/ <mount-point> nfs noauto,soft,nfsvers=4.1,sync,proto=tcp 0 0
335 Specify the NFS-Ganesha host name and the path to the mount point on
338 Configuring NFSv3 Clients
339 =========================
341 Linux clients can be configured to mount with NFSv3 by supplying
342 ``nfsvers=3`` and ``noacl`` as mount options. To use UDP as the
343 transport, add ``proto=udp`` to the mount options. However, TCP is the
344 preferred transport::
346 <ganesha-host-name>:/ <mount-point> nfs noauto,noacl,soft,nfsvers=3,sync,proto=tcp 0 0
348 Configure the NFS Ganesha EXPORT block Protocols setting with version
349 3 and the Transports setting with UDP if the mount will use version 3 with UDP.
354 Since NFSv3 does not communicate client OPEN and CLOSE operations to
355 file servers, RGW NFS cannot use these operations to mark the
356 beginning and ending of file upload transactions. Instead, RGW NFS
357 starts a new upload when the first write is sent to a file at offset
358 0, and finalizes the upload when no new writes to the file have been
359 seen for a period of time, by default, 10 seconds. To change this
360 timeout, set an alternate value for ``rgw_nfs_write_completion_interval_s``
361 in the RGW section(s) of the Ceph configuration file.
366 .. [#] http://docs.aws.amazon.com/AmazonS3/latest/dev/ListingKeysHierarchy.html