DOKK / manpages / debian 12 / libnbd-bin / nbdcopy.1.en
nbdcopy(1) LIBNBD nbdcopy(1)

nbdcopy - copy to and from an NBD server

 nbdcopy [--allocated] [-C N|--connections=N]
         [--destination-is-zero|--target-is-zero] [--flush]
         [--no-extents] [-p|--progress|--progress=FD]
         [--queue-size=N] [--request-size=N] [-R N|--requests=N]
         [-S N|--sparse=N] [--synchronous] [-T N|--threads=N]
         [-v|--verbose]
         SOURCE DESTINATION

 SOURCE, DESTINATION := - | FILE | DEVICE | NBD-URI | [ CMD ARGS ... ]
 DESTINATION += null:

 nbdcopy --help

 nbdcopy -V|--version

This copies everything from the NBD server at "example.com" to a local file called local.img.

This streams the first part of the disk on the NBD server at "example.com" into the file(1) command. Note here that "-" means to stream to stdout (and therefore into the pipe to the file command).

nbdinfo(1) is another way to detect the content on an NBD server.

Copy the full local hard disk "/dev/sdX" to the NBD server listening on the Unix domain socket /tmp/unixsock. Because of the -p option this will print a progress bar.

Copy a full disk from one NBD server to another.

nbdcopy -- [ qemu-nbd -r -f qcow2 https://example.com/disk.qcow2 ] -

Run qemu-nbd(8) as a subprocess to open URL "https://example.com/disk.qcow2", which is then streamed to stdout ("-"), converted to blocks of raw format data. Note "--" to prevent qemu-nbd flags from being interpreted as nbdcopy flags.

Concatenate two raw-format disk images into the qcow2 file output.qcow2. The output file has to be precreated.

Read the contents of the NBD server as fast as possible and throw it away (write it to a virtual null device). This is useful for benchmarking NBD servers and/or nbdcopy.

nbdcopy copies to and from an NBD server. It can upload a local file to an NBD server, or download the contents of an NBD server to a local file, device or stdin/stdout. It can also copy between NBD servers.

The local file can be a file, a block device (eg. "/dev/cdrom"), or "-" which means stream in from stdin or stream out to stdout.

The NBD server can be specified using an NBD URI (like "nbd://localhost"). The NBD server can be local or remote, and encryption can be used if libnbd was built with encryption support. Alternately you can use square brackets around a qemu-nbd(8) or nbdkit(1) command to run the NBD server as a subprocess of nbdcopy.

The destination may be the special word "null:" to throw away the output.

For more complex copying operations including converting between disk formats use "qemu-img convert", see qemu-img(1).

Display brief command line help and exit.
Normally nbdcopy tries to create sparse output (with holes) if the destination supports that. It does this in two ways: either using extent information from the source to copy holes (see --no-extents), or by detecting runs of zeroes (see -S). If you use --allocated then nbdcopy creates a fully allocated, non-sparse output on the destination.
Set the maximum number of NBD connections ("multi-conn"). By default nbdcopy will try to use multi-conn with up to 4 connections if the NBD server supports it. If copying between NBD servers then nbdcopy cannot use multi-conn if either of the servers does not support it.
Assume the destination is already zeroed. This allows nbdcopy to skip copying blocks of zeroes from the source to the destination. This is not safe unless the destination device is already zeroed. (--target-is-zero is provided for compatibility with qemu-img(1).)
Flush writes to ensure that everything is written to persistent storage before nbdcopy exits.
Normally nbdcopy uses extent metadata to skip over parts of the source disk which contain holes. If you use this flag, nbdcopy ignores extent information and reads everything, which is usually slower. You might use this flag in two situations: the source NBD server has incorrect metadata information; or the source has very slow extent querying so it's faster to simply read all of the data.
Display a progress bar.
Write a progress bar to the file descriptor "FD" (a number) in a format which is easily parsable by other programs. nbdcopy will periodically write the string "N/100\n" (where N is an integer between 0 and 100) to the file descriptor.

To get nbdcopy to write the progress bar to a file you can use the following shell commands:

 exec 3>/tmp/progress
 nbdcopy --progress=3 ...
 exec 3>&-
    
Set the maximum number of bytes to queue for in flight requests. The default value is 16 MiB, allowing up to 64 256k requests per NBD connection. If you use larger --request-size you may want to increase this value.
Set the maximum request size in bytes. The maximum value is 32 MiB, specified by the NBD protocol.
Set the maximum number of requests in flight per NBD connection.
Detect all zero blocks of size N (bytes) and make them sparse on the output. You can also turn off sparse detection using -S 0. The default is 4096 bytes, or the destination preferred block size, whichever is larger.
Force synchronous copying using the libnbd(3) synchronous ("high level") API. This is slow but may be necessary for some broken NBD servers which cannot handle multiple requests in flight. This mode is also used when streaming to and from stdio, pipes and sockets.
Use up to N threads for copying. By default this is set to the number of processor cores available.

Note --threads=0 means autodetect and --threads=1 means use a single thread.

Verbose output. This enables debug in libnbd (see nbd_set_debug(3)) as well as printing other useful information.
Display the package name and version and exit.

The three options --connections, --threads and --requests are related and control the amount of parallelism available. The defaults should ensure a reasonable amount of parallelism if possible and you don’t need to adjust them, but this section tries to describe what is going on.

Firstly if either side of the copy is streaming to or from stdio, a pipe, or a socket, or if you use the --synchronous option, then nbdcopy works in synchronous mode with no parallelism, and nothing else in this section applies.

The --connections=N option controls NBD multi-conn (see "Multi-conn" in libnbd(3)), opening up to N connections to the NBD server (or to both NBD servers if copying between NBD servers). This defaults to 4. The NBD servers must support and advertise multi-conn. For nbdkit(1) availability of multi-conn can depend on the plugin. You can use nbdinfo(1) to find out if a particular NBD server is advertising multi-conn. If the NBD server doesn’t advertise multi-conn then only one connection will be opened regardless of the --connections flag.

When copying between two NBD servers, the number of connections is limited to the minimum multi-conn supported on both sides. For the purposes of this calculation, you can consider local files and block devices as supporting infinite multi-conn.

When you run an NBD server as a subprocess (using the "[ ... ]" syntax) multi-conn cannot be used.

The --threads=N option allows nbdcopy to start up to N threads (defaulting to the number of cores). However nbdcopy cannot use more threads than the number of NBD connections.

The --requests=N option controls the maximum number of requests in flight on each NBD connection. This enables the NBD server to process requests in parallel even when multi-conn isn’t available or when using a single thread. The default is chosen to allow a reasonable amount of parallelism without using too much memory.

Because of this parallelism, nbdcopy does not read or write blocks in order. If for some reason you require that blocks are copied in strict order then you must use --synchronous.

Instead of connecting to an already running server using an NBD URI, you can run an NBD server as a subprocess using:

 nbdcopy -- [ CMD ARGS ... ] ...

This requires the server to support systemd socket activation, which both qemu-nbd(8) and nbdkit(1) support (see also nbd_connect_systemd_socket_activation(3)).

"[" and "]" must be separate command line parameters. You will usually need to use "--" to stop nbdcopy from misinterpreting NBD server flags as nbdcopy flags. Both the source and destination may be subprocesses. nbdcopy cleans up the subprocess on exit.

Some examples follow.

nbdcopy -- [ qemu-nbd -f qcow2 disk.qcow2 ] - | hexdump -C

In this example, qemu-nbd(8) is run as a subprocess. The subprocess opens disk.qcow2 and exposes it as NBD to nbdcopy. nbdcopy streams this to stdout ("-") into the pipe which is read by hexdump(1). (See also nbddump(1))

nbdcopy -- [ qemu-nbd -f qcow2 disk.qcow2 ] [ nbdkit memory 1G ]

Two subprocesses are created, qemu-nbd(8) as the source and nbdkit(1) as the destination. The qcow2 file is converted to raw and stored temporarily in the RAM disk (nbdkit-memory-plugin(1)).

When nbdcopy exits both servers are killed and the RAM disk goes away, so this command has no overall effect, but is useful for testing.

libnbd(3), nbddump(1), nbdfuse(1), nbdinfo(1), nbdsh(1), nbdkit(1), qemu-img(1).

Richard W.M. Jones

Copyright (C) 2020-2021 Red Hat Inc.

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with this library; if not, write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

2023-01-04 libnbd-1.14.2