nbdkit-ondemand-plugin - Man Page

create filesystems on demand

Synopsis

 nbdkit ondemand dir=EXPORTSDIR [size=]SIZE [wait=true] [share=true]
                 { [type=ext4|xfs|vfat|...] [label=LABEL]
                   | command=COMMAND [VAR=VALUE ...] }

Description

This is a plugin for nbdkit(1) which creates persistent filesystems on demand.  Clients may simply connect to the server, requesting a particular export name, and a new filesystem is created if it does not exist already.  Clients can also disconnect and reconnect with the same export name and the same filesystem will still be available. Filesystems are stored in a directory on the server, so they persist over server restarts.

By default each export is locked while it is in use by a client, preventing two clients from accessing the same filesystem (which would cause corruption).

Similar plugins include nbdkit-file-plugin(1) which can serve a predefined set of exports (clients cannot create more), nbdkit-tmpdisk-plugin(1) which creates a fresh temporary filesystem for each client, and nbdkit-linuxdisk-plugin(1) which exports a single filesystem from a local directory on the server.

Export names

When a new export name is requested by a client, a sparse file of the same name is created in dir=EXPORTSDIR on the server.  The file will be formatted with mkfs(8).  The size of the file is currently fixed by the size=SIZE parameter, but we intend to make this client-configurable in future.  The filesystem type and label may also be specified, otherwise ext4 and no label is used.

Export names must be ≤ NAME_MAX (usually 255) bytes in length and must not contain certain characters including . at the start, and /.  There may be other limitations added in future.  Client requests which do not obey these restrictions are rejected.  As a special case, export name "" is mapped to the file name default.

Security considerations

You should only use this in an environment where you trust all your clients, since clients can use this plugin to consume arbitrary amounts of disk space by creating unlimited exports.  It is therefore best to take steps to limit where clients can connect from using nbdkit-ip-filter(1), firewalls, or TLS client certificates.

The command parameter

Instead of running mkfs you can run an arbitrary command or shell script to create the disk.

The other parameters to the plugin are turned into shell variables passed to the command.  For example type becomes the shell variable $type, etc.  Any parameters you want can be passed to the plugin and will be turned into shell variables (not only type and label) making this a very flexible method to create filesystems and disks of all kinds.

Two special variables are also passed to the shell script fragment:

$disk

The absolute path of the disk file.  This is partially controlled by the client so you should quote it carefully.  This file is not pre-created, the command must create it for example using:

 truncate -s $size "$disk"
$size

The virtual size in bytes.  This is the size parameter, converted to bytes.  Note the final size served to the client is whatever disk size command creates.

Dead client detection

Clients which "go away" without disconnecting (such as clients which reboot) may cause exports to remain locked when they are no longer used.  Practically this can cause problems such as the same client being unable to reconnect after a reboot.  To avoid this you should enable keepalives to detect dead clients.  Use the nbdkit --keepalive option, or for more fine-grained control see the discussion of keepalives in nbdkit-service(1).

Example

Run the server like this:

 mkdir /var/tmp/exports
 nbdkit ondemand dir=/var/tmp/exports 1G

Clients can connect and create 1G ext4 filesystems on demand using commands such as these (note the different export names):

 nbd-client server /dev/nbd0 -N export1
 mount /dev/nbd0 /mnt
 guestfish --format=raw -a nbd://localhost/export2 -m /dev/sda
 qemu-img info nbd:localhost:10809:exportname=export2

On the server you would see two filesystems created:

 $ ls -l /var/tmp/exports
 -rw-rw-r--. 1 rjones rjones 1073741824 Aug 13 21:40 export1
 -rw-rw-r--. 1 rjones rjones 1073741824 Aug 13 21:40 export2

The plugin does not clean these up.  If they are no longer needed then the server admin should delete them (or use a tmp cleaner).

Parameters

command='COMMAND'

Instead of running mkfs(8) to create the initial filesystem, run COMMAND (a shell script fragment which usually must be quoted to protect it from the shell).  See "The command parameter" above.

dir=EXPORTSDIR

The directory where filesystems are saved.  When first using this plugin you should point this to an empty directory.  When clients connect, filesystems are created here.

This parameter is required.

label=LABEL

Select the filesystem label.  The default is not set.

share=true

If set, the export is not locked.  Two or more clients may connect to the same export.  This usually results in filesystem corruption (if using a regular filesystem like ext4), so you have to use some other means to guarantee exclusive access or to ensure sharing is safe.

[size=]SIZE

Specify the virtual size of all of the filesystems.

If using command, this is only a suggested size.  The actual size of the resulting disk will be the size of the disk created by command.

This parameter is required.

size= is a magic config key and may be omitted in most cases. See "Magic parameters" in nbdkit(1).

type=FS

Select the filesystem type.  The default is ext4.  Most non-networked, non-cluster filesystem types supported by the mkfs(8) command can be used here.

wait=true

If set, if two clients try to connect at the same time to the same export then the second client will wait for the first to disconnect. The default behaviour is to reject the second client with the error message:

 filesystem is locked by another client

This setting is sometimes useful if you are making repeated connections and at the network level the first connection does not fully disconnect before the next connection starts.  This can also happen as a side-effect of using guestfish --ro which opens two NBD connections in quick succession.

If share=true then locking is not used and the wait parameter has no effect.

Files

$plugindir/nbdkit-ondemand-plugin.so

The plugin.

Use nbdkit --dump-config to find the location of $plugindir.

Version

nbdkit-ondemand-plugin first appeared in nbdkit 1.22.

See Also

nbdkit(1), nbdkit-plugin(3), nbdkit-file-plugin(1), nbdkit-ip-filter(1), nbdkit-limit-filter(1), nbdkit-linuxdisk-plugin(1), nbdkit-memory-plugin(1), nbdkit-tmpdisk-plugin(1), nbdkit-service(1), nbdkit-tls(1), mkfs(8), mke2fs(8).

Authors

Richard W.M. Jones

License

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

THIS SOFTWARE IS PROVIDED BY RED HAT AND CONTRIBUTORS ''AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL RED HAT OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Referenced By

nbdkit(1), nbdkit-plugin(3), nbdkit-release-notes-1.22(1), nbdkit-release-notes-1.26(1), nbdkit-tmpdisk-plugin(1).

2024-10-14 nbdkit-1.41.8