rbd – manage rados block device (RBD) images


rbd [ -c ceph.conf ] [ -m monaddr ] [–cluster cluster name] [ -p | –pool pool ] [ command … ]


rbd is a utility for manipulating rados block device (RBD) images, used by the Linux rbd driver and the rbd storage driver for Qemu/KVM. RBD images are simple block devices that are striped over objects and stored in a RADOS object store. The size of the objects the image is striped over must be a power of two.


-c ceph.conf, --conf ceph.conf

Use ceph.conf configuration file instead of the default /etc/ceph/ceph.conf to determine monitor addresses during startup.

-m monaddress[:port]

Connect to specified monitor (instead of looking through ceph.conf).

--cluster cluster name

Use different cluster name as compared to default cluster name ceph.

-p pool-name, --pool pool-name

Interact with the given pool. Required by most commands.


Do not output progress information (goes to standard error by default for some commands).


--image-format format-id

Specifies which object layout to use. The default is 2.

  • format 1 - (deprecated) Use the original format for a new rbd image. This format is understood by all versions of librbd and the kernel rbd module, but does not support newer features like cloning.
  • format 2 - Use the second rbd format, which is supported by librbd and kernel since version 3.11 (except for striping). This adds support for cloning and is more easily extensible to allow more features in the future.
-s size-in-M/G/T, --size size-in-M/G/T

Specifies the size of the new rbd image or the new size of the existing rbd image in M/G/T. If no suffix is given, unit M is assumed.

--object-size size-in-B/K/M

Specifies the object size in B/K/M. Object size will be rounded up the nearest power of two; if no suffix is given, unit B is assumed. The default object size is 4M, smallest is 4K and maximum is 32M.

--stripe-unit size-in-B/K/M

Specifies the stripe unit size in B/K/M. If no suffix is given, unit B is assumed. See striping section (below) for more details.

--stripe-count num

Specifies the number of objects to stripe over before looping back to the first object. See striping section (below) for more details.

--snap snap

Specifies the snapshot name for the specific operation.

--id username

Specifies the username (without the client. prefix) to use with the map command.

--keyring filename

Specifies a keyring file containing a secret for the specified user to use with the map command. If not specified, the default keyring locations will be searched.

--keyfile filename

Specifies a file containing the secret key of --id user to use with the map command. This option is overridden by --keyring if the latter is also specified.

--shared lock-tag

Option for lock add that allows multiple clients to lock the same image if they use the same tag. The tag is an arbitrary string. This is useful for situations where an image must be open from more than one client at once, like during live migration of a virtual machine, or for use underneath a clustered filesystem.

--format format

Specifies output formatting (default: plain, json, xml)


Make json or xml formatted output more human-readable.

-o krbd-options, --options krbd-options

Specifies which options to use when mapping or unmapping an image via the rbd kernel driver. krbd-options is a comma-separated list of options (similar to mount(8) mount options). See kernel rbd (krbd) options section below for more details.


Map the image read-only. Equivalent to -o ro.

--image-feature feature-name

Specifies which RBD format 2 feature should be enabled when creating an image. Multiple features can be enabled by repeating this option multiple times. The following features are supported:

  • layering: layering support
  • striping: striping v2 support
  • exclusive-lock: exclusive locking support
  • object-map: object map support (requires exclusive-lock)
  • fast-diff: fast diff calculations (requires object-map)
  • deep-flatten: snapshot flatten support
  • journaling: journaled IO support (requires exclusive-lock)

Specifies that the image will be used concurrently by multiple clients. This will disable features that are dependent upon exclusive ownership of the image.


Specifies that the diff should be limited to the extents of a full object instead of showing intra-object deltas. When the object map feature is enabled on an image, limiting the diff to the object extents will dramatically improve performance since the differences can be computed by examining the in-memory object map instead of querying RADOS for each object within the image.


ls [-l | –long] [pool-name]
Will list all rbd images listed in the rbd_directory object. With -l, also show snapshots, and use longer-format output including size, parent (if clone), format, etc.
du [-p | –pool pool-name] [image-spec | snap-spec]

Will calculate the provisioned and actual disk usage of all images and associated snapshots within the specified pool. It can also be used against individual images and snapshots.

If the RBD fast-diff feature isn’t enabled on images, this operation will require querying the OSDs for every potential object within the image.

info image-spec | snap-spec
Will dump information (such as size and object size) about a specific rbd image. If image is a clone, information about its parent is also displayed. If a snapshot is specified, whether it is protected is shown as well.
create (-s | –size size-in-M/G/T) [–image-format format-id] [–object-size size-in-B/K/M] [–stripe-unit size-in-B/K/M –stripe-count num] [–image-feature feature-name]… [–image-shared] image-spec
Will create a new rbd image. You must also specify the size via –size. The –stripe-unit and –stripe-count arguments are optional, but must be used together.
clone [–object-size size-in-B/K/M] [–stripe-unit size-in-B/K/M –stripe-count num] [–image-feature feature-name] [–image-shared] parent-snap-spec child-image-spec

Will create a clone (copy-on-write child) of the parent snapshot. Object size will be identical to that of the parent image unless specified. Size will be the same as the parent snapshot. The –stripe-unit and –stripe-count arguments are optional, but must be used together.

The parent snapshot must be protected (see rbd snap protect). This requires image format 2.

flatten image-spec

If image is a clone, copy all shared blocks from the parent snapshot and make the child independent of the parent, severing the link between parent snap and child. The parent snapshot can be unprotected and deleted if it has no further dependent clones.

This requires image format 2.

children snap-spec

List the clones of the image at the given snapshot. This checks every pool, and outputs the resulting poolname/imagename.

This requires image format 2.

resize (-s | –size size-in-M/G/T) [–allow-shrink] image-spec
Resizes rbd image. The size parameter also needs to be specified. The –allow-shrink option lets the size be reduced.
rm image-spec
Deletes an rbd image (including all data blocks). If the image has snapshots, this fails and nothing is deleted.
export (image-spec | snap-spec) [dest-path]
Exports image to dest path (use - for stdout).
import [–image-format format-id] [–object-size size-in-B/K/M] [–stripe-unit size-in-B/K/M –stripe-count num] [–image-feature feature-name]… [–image-shared] src-path [image-spec]

Creates a new image and imports its data from path (use - for stdin). The import operation will try to create sparse rbd images if possible. For import from stdin, the sparsification unit is the data block size of the destination image (object size).

The –stripe-unit and –stripe-count arguments are optional, but must be used together.

export-diff [–from-snap snap-name] [–whole-object] (image-spec | snap-spec) dest-path
Exports an incremental diff for an image to dest path (use - for stdout). If an initial snapshot is specified, only changes since that snapshot are included; otherwise, any regions of the image that contain data are included. The end snapshot is specified using the standard –snap option or @snap syntax (see below). The image diff format includes metadata about image size changes, and the start and end snapshots. It efficiently represents discarded or ‘zero’ regions of the image.
merge-diff first-diff-path second-diff-path merged-diff-path
Merge two continuous incremental diffs of an image into one single diff. The first diff’s end snapshot must be equal with the second diff’s start snapshot. The first diff could be - for stdin, and merged diff could be - for stdout, which enables multiple diff files to be merged using something like ‘rbd merge-diff first second - | rbd merge-diff - third result’. Note this command currently only support the source incremental diff with stripe_count == 1
import-diff src-path image-spec
Imports an incremental diff of an image and applies it to the current image. If the diff was generated relative to a start snapshot, we verify that snapshot already exists before continuing. If there was an end snapshot we verify it does not already exist before applying the changes, and create the snapshot when we are done.
diff [–from-snap snap-name] [–whole-object] image-spec | snap-spec
Dump a list of byte extents in the image that have changed since the specified start snapshot, or since the image was created. Each output line includes the starting offset (in bytes), the length of the region (in bytes), and either ‘zero’ or ‘data’ to indicate whether the region is known to be zeros or may contain other data.
cp (src-image-spec | src-snap-spec) dest-image-spec
Copies the content of a src-image into the newly created dest-image. dest-image will have the same size, object size, and image format as src-image.
mv src-image-spec dest-image-spec
Renames an image. Note: rename across pools is not supported.
image-meta list image-spec
Show metadata held on the image. The first column is the key and the second column is the value.
image-meta get image-spec key
Get metadata value with the key.
image-meta set image-spec key value
Set metadata key with the value. They will displayed in image-meta list.
image-meta remove image-spec key
Remove metadata key with the value.
object-map rebuild image-spec | snap-spec
Rebuilds an invalid object map for the specified image. An image snapshot can be specified to rebuild an invalid object map for a snapshot.
snap ls image-spec
Dumps the list of snapshots inside a specific image.
snap create snap-spec
Creates a new snapshot. Requires the snapshot name parameter specified.
snap rollback snap-spec
Rollback image content to snapshot. This will iterate through the entire blocks array and update the data head content to the snapshotted version.
snap rm snap-spec
Removes the specified snapshot.
snap purge image-spec
Removes all snapshots from an image.
snap protect snap-spec

Protect a snapshot from deletion, so that clones can be made of it (see rbd clone). Snapshots must be protected before clones are made; protection implies that there exist dependent cloned children that refer to this snapshot. rbd clone will fail on a nonprotected snapshot.

This requires image format 2.

snap unprotect snap-spec

Unprotect a snapshot from deletion (undo snap protect). If cloned children remain, snap unprotect fails. (Note that clones may exist in different pools than the parent snapshot.)

This requires image format 2.

map [-o | –options krbd-options ] [–read-only] image-spec | snap-spec
Maps the specified image to a block device via the rbd kernel module.
unmap [-o | –options krbd-options ] image-spec | snap-spec | device-path
Unmaps the block device that was mapped via the rbd kernel module.
Show the rbd images that are mapped via the rbd kernel module.
nbd map [–device device-path] [–read-only] image-spec | snap-spec
Maps the specified image to a block device via the rbd-nbd tool.
nbd unmap device-path
Unmaps the block device that was mapped via the rbd-nbd tool.
nbd list
Show the list of used nbd devices via the rbd-nbd tool.
status image-spec
Show the status of the image, including which clients have it open.
feature disable image-spec feature-name
Disables the specified feature on the specified image. Multiple features can be specified.
feature enable image-spec feature-name
Enables the specified feature on the specified image. Multiple features can be specified.
lock list image-spec
Show locks held on the image. The first column is the locker to use with the lock remove command.
lock add [–shared lock-tag] image-spec lock-id
Lock an image. The lock-id is an arbitrary name for the user’s convenience. By default, this is an exclusive lock, meaning it will fail if the image is already locked. The –shared option changes this behavior. Note that locking does not affect any operation other than adding a lock. It does not protect an image from being deleted.
lock remove image-spec lock-id locker
Release a lock on an image. The lock id and locker are as output by lock ls.
bench-write [–io-size size-in-B/K/M/G/T] [–io-threads num-ios-in-flight] [–io-total size-in-B/K/M/G/T] [–io-pattern seq | rand] image-spec
Generate a series of writes to the image and measure the write throughput and latency. If no suffix is given, unit B is assumed for both –io-size and –io-total. Defaults are: –io-size 4096, –io-threads 16, –io-total 1G, –io-pattern seq.

Image and snap specs

image-spec is [pool-name]/image-name
snap-spec is [pool-name]/image-name@snap-name

The default for pool-name is “rbd”. If an image name contains a slash character (‘/’), pool-name is required.

You may specify each name individually, using –pool, –image and –snap options, but this is discouraged in favor of the above spec syntax.


RBD images are striped over many objects, which are then stored by the Ceph distributed object store (RADOS). As a result, read and write requests for the image are distributed across many nodes in the cluster, generally preventing any single node from becoming a bottleneck when individual images get large or busy.

The striping is controlled by three parameters:


The size of objects we stripe over is a power of two. It will be rounded up the nearest power of two. The default object size is 4 MB, smallest is 4K and maximum is 32M.


Each [stripe_unit] contiguous bytes are stored adjacently in the same object, before we move on to the next object.


After we write [stripe_unit] bytes to [stripe_count] objects, we loop back to the initial object and write another stripe, until the object reaches its maximum size. At that point, we move on to the next [stripe_count] objects.

By default, [stripe_unit] is the same as the object size and [stripe_count] is 1. Specifying a different [stripe_unit] requires that the STRIPINGV2 feature be supported (added in Ceph v0.53) and format 2 images be used.

Kernel rbd (krbd) options

Most of these options are useful mainly for debugging and benchmarking. The default values are set in the kernel and may therefore depend on the version of the running kernel.

Per client instance rbd map options:

  • fsid=aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee - FSID that should be assumed by the client.
  • ip=a.b.c.d[:p] - IP and, optionally, port the client should use.
  • share - Enable sharing of client instances with other mappings (default).
  • noshare - Disable sharing of client instances with other mappings.
  • crc - Enable CRC32C checksumming for data writes (default).
  • nocrc - Disable CRC32C checksumming for data writes.
  • cephx_require_signatures - Require cephx message signing (since 3.19, default).
  • nocephx_require_signatures - Don’t require cephx message signing (since 3.19).
  • tcp_nodelay - Disable Nagle’s algorithm on client sockets (since 4.0, default).
  • notcp_nodelay - Enable Nagle’s algorithm on client sockets (since 4.0).
  • cephx_sign_messages - Enable message signing (since 4.4, default).
  • nocephx_sign_messages - Disable message signing (since 4.4).
  • mount_timeout=x - A timeout on various steps in rbd map and rbd unmap sequences (default is 60 seconds). In particular, since 4.2 this can be used to ensure that rbd unmap eventually times out when there is no network connection to a cluster.
  • osdkeepalive=x - OSD keepalive timeout (default is 5 seconds).
  • osd_idle_ttl=x - OSD idle TTL (default is 60 seconds).

Per mapping (block device) rbd map options:

  • rw - Map the image read-write (default).
  • ro - Map the image read-only. Equivalent to –read-only.
  • queue_depth=x - queue depth (since 4.2, default is 128 requests).
  • lock_on_read - Acquire exclusive lock on reads, in addition to writes and discards (since 4.9).

rbd unmap options:

  • force - Force the unmapping of a block device that is open (since 4.9). The driver will wait for running requests to complete and then unmap; requests sent to the driver after initiating the unmap will be failed.


To create a new rbd image that is 100 GB:

rbd create mypool/myimage --size 102400

To use a non-default object size (8 MB):

rbd create mypool/myimage --size 102400 --object-size 8M

To delete an rbd image (be careful!):

rbd rm mypool/myimage

To create a new snapshot:

rbd snap create mypool/myimage@mysnap

To create a copy-on-write clone of a protected snapshot:

rbd clone mypool/myimage@mysnap otherpool/cloneimage

To see which clones of a snapshot exist:

rbd children mypool/myimage@mysnap

To delete a snapshot:

rbd snap rm mypool/myimage@mysnap

To map an image via the kernel with cephx enabled:

rbd map mypool/myimage --id admin --keyfile secretfile

To map an image via the kernel with different cluster name other than default ceph.

rbd map mypool/myimage –cluster cluster name

To unmap an image:

rbd unmap /dev/rbd0

To create an image and a clone from it:

rbd import --image-format 2 image mypool/parent
rbd snap create mypool/parent@snap
rbd snap protect mypool/parent@snap
rbd clone mypool/parent@snap otherpool/child

To create an image with a smaller stripe_unit (to better distribute small writes in some workloads):

rbd create mypool/myimage --size 102400 --stripe-unit 65536B --stripe-count 16

To change an image from one image format to another, export it and then import it as the desired image format:

rbd export mypool/myimage@snap /tmp/img
rbd import --image-format 2 /tmp/img mypool/myimage2

To lock an image for exclusive use:

rbd lock add mypool/myimage mylockid

To release a lock:

rbd lock remove mypool/myimage mylockid client.2485


rbd is part of Ceph, a massively scalable, open-source, distributed storage system. Please refer to the Ceph documentation at http://ceph.com/docs for more information.

See also

ceph(8), rados(8)