1 <?xml version='1.0'?> <!--*-nxml-*-->
2 <!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN"
3 "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">
6 This file is part of systemd.
8 Copyright 2010 Lennart Poettering
10 systemd is free software; you can redistribute it and/or modify it
11 under the terms of the GNU Lesser General Public License as published by
12 the Free Software Foundation; either version 2.1 of the License, or
13 (at your option) any later version.
15 systemd is distributed in the hope that it will be useful, but
16 WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 Lesser General Public License for more details.
20 You should have received a copy of the GNU Lesser General Public License
21 along with systemd; If not, see <http://www.gnu.org/licenses/>.
24 <refentry id="systemd-nspawn">
27 <title>systemd-nspawn</title>
28 <productname>systemd</productname>
32 <contrib>Developer</contrib>
33 <firstname>Lennart</firstname>
34 <surname>Poettering</surname>
35 <email>lennart@poettering.net</email>
41 <refentrytitle>systemd-nspawn</refentrytitle>
42 <manvolnum>1</manvolnum>
46 <refname>systemd-nspawn</refname>
47 <refpurpose>Spawn a namespace container for debugging, testing and building</refpurpose>
52 <command>systemd-nspawn</command>
53 <arg choice="opt" rep="repeat">OPTIONS</arg>
54 <arg choice="opt"><replaceable>COMMAND</replaceable>
55 <arg choice="opt" rep="repeat">ARGS</arg>
59 <command>systemd-nspawn</command>
60 <arg choice="plain">-b</arg>
61 <arg choice="opt" rep="repeat">OPTIONS</arg>
62 <arg choice="opt" rep="repeat">ARGS</arg>
67 <title>Description</title>
69 <para><command>systemd-nspawn</command> may be used to
70 run a command or OS in a light-weight namespace
71 container. In many ways it is similar to
72 <citerefentry><refentrytitle>chroot</refentrytitle><manvolnum>1</manvolnum></citerefentry>,
73 but more powerful since it fully virtualizes the file
74 system hierarchy, as well as the process tree, the
75 various IPC subsystems and the host and domain
78 <para><command>systemd-nspawn</command> limits access
79 to various kernel interfaces in the container to
80 read-only, such as <filename>/sys</filename>,
81 <filename>/proc/sys</filename> or
82 <filename>/sys/fs/selinux</filename>. Network
83 interfaces and the system clock may not be changed
84 from within the container. Device nodes may not be
85 created. The host system cannot be rebooted and kernel
86 modules may not be loaded from within the
89 <para>Note that even though these security precautions
90 are taken <command>systemd-nspawn</command> is not
91 suitable for secure container setups. Many of the
92 security features may be circumvented and are hence
93 primarily useful to avoid accidental changes to the
94 host system from the container. The intended use of
95 this program is debugging and testing as well as
96 building of packages, distributions and software
97 involved with boot and systems management.</para>
100 <citerefentry><refentrytitle>chroot</refentrytitle><manvolnum>1</manvolnum></citerefentry> <command>systemd-nspawn</command>
101 may be used to boot full Linux-based operating systems
102 in a container.</para>
104 <para>Use a tool like
105 <citerefentry><refentrytitle>yum</refentrytitle><manvolnum>8</manvolnum></citerefentry>,
106 <citerefentry><refentrytitle>debootstrap</refentrytitle><manvolnum>8</manvolnum></citerefentry>,
108 <citerefentry><refentrytitle>pacman</refentrytitle><manvolnum>8</manvolnum></citerefentry>
109 to set up an OS directory tree suitable as file system
110 hierarchy for <command>systemd-nspawn</command>
113 <para>Note that <command>systemd-nspawn</command> will
114 mount file systems private to the container to
115 <filename>/dev</filename>,
116 <filename>/run</filename> and similar. These will
117 not be visible outside of the container, and their
118 contents will be lost when the container exits.</para>
120 <para>Note that running two
121 <command>systemd-nspawn</command> containers from the
122 same directory tree will not make processes in them
123 see each other. The PID namespace separation of the
124 two containers is complete and the containers will
125 share very few runtime objects except for the
126 underlying file system. Use
127 <citerefentry><refentrytitle>machinectl</refentrytitle><manvolnum>1</manvolnum></citerefentry>'s
128 <command>login</command> command to request an
129 additional login prompt in a running container.</para>
131 <para><command>systemd-nspawn</command> implements the
133 url="http://www.freedesktop.org/wiki/Software/systemd/ContainerInterface">Container
134 Interface</ulink> specification.</para>
136 <para>As a safety check
137 <command>systemd-nspawn</command> will verify the
138 existence of <filename>/etc/os-release</filename> in
139 the container tree before starting the container (see
140 <citerefentry><refentrytitle>os-release</refentrytitle><manvolnum>5</manvolnum></citerefentry>). It
141 might be necessary to add this file to the container
142 tree manually if the OS of the container is too old to
143 contain this file out-of-the-box.</para>
147 <title>Options</title>
149 <para>If option <option>-b</option> is specified, the
150 arguments are used as arguments for the init
151 binary. Otherwise, <replaceable>COMMAND</replaceable>
152 specifies the program to launch in the container, and
153 the remaining arguments are used as arguments for this
154 program. If <option>-b</option> is not used and no
155 arguments are specifed, a shell is launched in the
158 <para>The following options are understood:</para>
162 <term><option>-h</option></term>
163 <term><option>--help</option></term>
165 <listitem><para>Prints a short help
166 text and exits.</para></listitem>
170 <term><option>--version</option></term>
172 <listitem><para>Prints a version string
173 and exits.</para></listitem>
177 <term><option>-q</option></term>
178 <term><option>--quiet</option></term>
180 <listitem><para>Turns off any status
181 output by the tool itself. When this
182 switch is used, then the only output
183 by nspawn will be the console output
184 of the container OS itself.</para></listitem>
188 <term><option>-D</option></term>
189 <term><option>--directory=</option></term>
191 <listitem><para>Directory to use as
192 file system root for the namespace
193 container. If omitted, the current
195 used.</para></listitem>
199 <term><option>-b</option></term>
200 <term><option>--boot</option></term>
202 <listitem><para>Automatically search
203 for an init binary and invoke it
204 instead of a shell or a user supplied
205 program. If this option is used,
206 arguments specified on the command
207 line are used as arguments for the
208 init binary. This option may not be
210 <option>--share-system</option>.
215 <term><option>-u</option></term>
216 <term><option>--user=</option></term>
218 <listitem><para>Run the command
219 under specified user, create home
220 directory and cd into it. As rest
221 of systemd-nspawn, this is not
222 the security feature and limits
223 against accidental changes only.
228 <term><option>-M</option></term>
229 <term><option>--machine=</option></term>
231 <listitem><para>Sets the machine name
232 for this container. This name may be
233 used to identify this container on the
234 host, and is used to initialize the
235 container's hostname (which the
236 container can choose to override,
237 however). If not specified, the last
238 component of the root directory of the
239 container is used.</para></listitem>
243 <term><option>--uuid=</option></term>
245 <listitem><para>Set the specified UUID
246 for the container. The init system
248 <filename>/etc/machine-id</filename>
249 from this if this file is not set yet.
254 <term><option>--slice=</option></term>
256 <listitem><para>Make the container
257 part of the specified slice, instead
259 <filename>machine.slice</filename>.</para>
264 <term><option>--private-network</option></term>
266 <listitem><para>Disconnect networking
267 of the container from the host. This
268 makes all network interfaces
269 unavailable in the container, with the
270 exception of the loopback device and
272 <option>--network-interface=</option>
274 <option>--network-veth</option>. If
275 this option is specified the
276 CAP_NET_ADMIN capability will be added
277 to the set of capabilities the
278 container retains. The latter may be
280 <option>--drop-capability=</option>.</para></listitem>
284 <term><option>--network-interface=</option></term>
286 <listitem><para>Assign the specified
287 network interface to the
288 container. This will move the
289 specified interface from the calling
290 namespace and place it in the
291 container. When the container
292 terminates it is moved back to the
293 host namespace. Note that
294 <option>--network-interface=</option>
296 <option>--private-network</option>. This
297 option may be used more than once to
298 add multiple network interfaces to the
299 container.</para></listitem>
303 <term><option>--network-veth</option></term>
305 <listitem><para>Create a virtual
306 ethernet link between host and
307 container. The host side of the
308 ethernet link will be available as
309 network interface named after the
310 container's name (as specified with
311 <option>--machine=</option>), prefixed
312 with <literal>ve-</literal>. The
313 container side of the the ethernet
315 <literal>host0</literal>. Note that
316 <option>--network-veth</option>
318 <option>--private-network</option>.</para></listitem>
322 <term><option>-Z</option></term>
323 <term><option>--selinux-context=</option></term>
325 <listitem><para>Sets the SELinux
326 security context to be used to label
327 processes in the container.</para>
332 <term><option>-L</option></term>
333 <term><option>--selinux-apifs-context=</option></term>
335 <listitem><para>Sets the SELinux security
336 context to be used to label files in
337 the virtual API file systems in the
343 <term><option>--capability=</option></term>
345 <listitem><para>List one or more
346 additional capabilities to grant the
347 container. Takes a comma-separated
348 list of capability names, see
349 <citerefentry><refentrytitle>capabilities</refentrytitle><manvolnum>7</manvolnum></citerefentry>
350 for more information. Note that the
351 following capabilities will be granted
352 in any way: CAP_CHOWN,
353 CAP_DAC_OVERRIDE, CAP_DAC_READ_SEARCH,
354 CAP_FOWNER, CAP_FSETID, CAP_IPC_OWNER,
357 CAP_NET_BIND_SERVICE,
358 CAP_NET_BROADCAST, CAP_NET_RAW,
359 CAP_SETGID, CAP_SETFCAP, CAP_SETPCAP,
360 CAP_SETUID, CAP_SYS_ADMIN,
361 CAP_SYS_CHROOT, CAP_SYS_NICE,
362 CAP_SYS_PTRACE, CAP_SYS_TTY_CONFIG,
363 CAP_SYS_RESOURCE, CAP_SYS_BOOT,
365 CAP_AUDIT_CONTROL. Also CAP_NET_ADMIN
367 <option>--private-network</option> is
368 specified. If the special value
369 <literal>all</literal> is passed all
371 retained.</para></listitem>
375 <term><option>--drop-capability=</option></term>
377 <listitem><para>Specify one or more
378 additional capabilities to drop for
379 the container. This allows running the
380 container with fewer capabilities than
381 the default (see above).</para></listitem>
385 <term><option>--link-journal=</option></term>
387 <listitem><para>Control whether the
388 container's journal shall be made
389 visible to the host system. If enabled,
390 allows viewing the container's journal
391 files from the host (but not vice
393 <literal>no</literal>,
394 <literal>host</literal>,
395 <literal>guest</literal>,
396 <literal>auto</literal>. If
397 <literal>no</literal>, the journal is
398 not linked. If <literal>host</literal>,
399 the journal files are stored on the
400 host file system (beneath
401 <filename>/var/log/journal/<replaceable>machine-id</replaceable></filename>)
402 and the subdirectory is bind-mounted
403 into the container at the same
404 location. If <literal>guest</literal>,
405 the journal files are stored on the
406 guest file system (beneath
407 <filename>/var/log/journal/<replaceable>machine-id</replaceable></filename>)
408 and the subdirectory is symlinked into the host
409 at the same location. If
410 <literal>auto</literal> (the default),
411 and the right subdirectory of
412 <filename>/var/log/journal</filename>
413 exists, it will be bind mounted
414 into the container. If the
415 subdirectory does not exist, no
416 linking is performed. Effectively,
417 booting a container once with
418 <literal>guest</literal> or
419 <literal>host</literal> will link the
420 journal persistently if further on
421 the default of <literal>auto</literal>
422 is used.</para></listitem>
426 <term><option>-j</option></term>
428 <listitem><para>Equivalent to
429 <option>--link-journal=guest</option>.</para></listitem>
433 <term><option>--read-only</option></term>
435 <listitem><para>Mount the root file
436 system read-only for the
437 container.</para></listitem>
441 <term><option>--bind=</option></term>
442 <term><option>--bind-ro=</option></term>
444 <listitem><para>Bind mount a file or
445 directory from the host into the
446 container. Either takes a path
447 argument -- in which case the
448 specified path will be mounted from
449 the host to the same path in the
450 container --, or a colon-separated
451 pair of paths -- in which case the
452 first specified path is the source in
453 the host, and the second path is the
454 destination in the container. The
455 <option>--bind-ro=</option> option
456 creates read-only bind
457 mount.</para></listitem>
461 <term><option>--setenv=</option></term>
463 <listitem><para>Specifies an
464 environment variable assignment to
465 pass to the init process in the
466 container, in the format
467 <literal>NAME=VALUE</literal>. This
468 may be used to override the default
469 variables or to set additional
470 variables. This parameter may be used
471 more than once.</para></listitem>
475 <term><option>--share-system</option></term>
477 <listitem><para>Allows the container
478 to share certain system facilities
479 with the host. More specifically, this
480 turns off PID namespacing, UTS
481 namespacing and IPC namespacing, and
482 thus allows the guest to see and
483 interact more easily with processes
484 outside of the container. Note that
485 using this option makes it impossible
486 to start up a full Operating System in
487 the container, as an init system
488 cannot operate in this mode. It is
489 only useful to run specific programs
490 or applications this way, without
491 involving an init system in the
492 container. This option implies
493 <option>--register=no</option>. This
494 option may not be combined with
495 <option>--boot</option>.</para></listitem>
499 <term><option>--register=</option></term>
501 <listitem><para>Controls whether the
502 container is registered with
503 <citerefentry><refentrytitle>systemd-machined</refentrytitle><manvolnum>8</manvolnum></citerefentry>. Takes
504 a boolean argument, defaults to
505 <literal>yes</literal>. This option
506 should be enabled when the container
507 runs a full Operating System (more
508 specifically: an init system), and is
509 useful to ensure that the container is
511 <citerefentry><refentrytitle>machinectl</refentrytitle><manvolnum>1</manvolnum></citerefentry>
512 and shown by tools such as
513 <citerefentry><refentrytitle>ps</refentrytitle><manvolnum>1</manvolnum></citerefentry>. If
514 the container does not run an init
515 system it is recommended to set this
516 option to <literal>no</literal>. Note
517 that <option>--share-system</option>
519 <option>--register=no</option>.
524 <term><option>--keep-unit</option></term>
526 <listitem><para>Instead of creating a
527 transient scope unit to run the
528 container in, simply register the
529 service or scope unit
530 <command>systemd-nspawn</command> has
532 <citerefentry><refentrytitle>systemd-machined</refentrytitle><manvolnum>8</manvolnum></citerefentry>. This
534 <option>--register=no</option> is
535 used. This switch should be used if
536 <command>systemd-nspawn</command> is
537 invoked from within an a service unit,
538 and the service unit's sole purpose
540 <command>systemd-nspawn</command>
541 container. This option is not
542 available if run from a user
543 session.</para></listitem>
551 <title>Example 1</title>
553 <programlisting># yum -y --releasever=19 --nogpg --installroot=/srv/mycontainer --disablerepo='*' --enablerepo=fedora install systemd passwd yum fedora-release vim-minimal
554 # systemd-nspawn -bD /srv/mycontainer</programlisting>
556 <para>This installs a minimal Fedora distribution into
557 the directory <filename noindex='true'>/srv/mycontainer/</filename> and
558 then boots an OS in a namespace container in
563 <title>Example 2</title>
565 <programlisting># debootstrap --arch=amd64 unstable ~/debian-tree/
566 # systemd-nspawn -D ~/debian-tree/</programlisting>
568 <para>This installs a minimal Debian unstable
569 distribution into the directory
570 <filename>~/debian-tree/</filename> and then spawns a
571 shell in a namespace container in it.</para>
575 <title>Example 3</title>
577 <programlisting># pacstrap -c -d ~/arch-tree/ base
578 # systemd-nspawn -bD ~/arch-tree/</programlisting>
580 <para>This installs a mimimal Arch Linux distribution into
581 the directory <filename>~/arch-tree/</filename> and then
582 boots an OS in a namespace container in it.</para>
586 <title>Example 4</title>
588 <programlisting># mv ~/arch-tree /var/lib/container/arch
589 # systemctl enable systemd-nspawn@arch.service
590 # systemctl start systemd-nspawn@arch.service</programlisting>
592 <para>This makes the Arch Linux container part of the
593 <filename>multi-user.target</filename> on the host.
598 <title>Example 5</title>
600 <programlisting># btrfs subvolume snapshot / /.tmp
601 # systemd-nspawn --private-network -D /.tmp -b</programlisting>
603 <para>This runs a copy of the host system in a
604 btrfs snapshot.</para>
608 <title>Example 6</title>
610 <programlisting># chcon system_u:object_r:svirt_sandbox_file_t:s0:c0,c1 -R /srv/container
611 # systemd-nspawn -L system_u:object_r:svirt_sandbox_file_t:s0:c0,c1 -Z system_u:system_r:svirt_lxc_net_t:s0:c0,c1 -D /srv/container /bin/sh</programlisting>
613 <para>This runs a container with SELinux sandbox security contexts.</para>
617 <title>Exit status</title>
619 <para>The exit code of the program executed in the
620 container is returned.</para>
624 <title>See Also</title>
626 <citerefentry><refentrytitle>systemd</refentrytitle><manvolnum>1</manvolnum></citerefentry>,
627 <citerefentry><refentrytitle>chroot</refentrytitle><manvolnum>1</manvolnum></citerefentry>,
628 <citerefentry><refentrytitle>yum</refentrytitle><manvolnum>8</manvolnum></citerefentry>,
629 <citerefentry><refentrytitle>debootstrap</refentrytitle><manvolnum>8</manvolnum></citerefentry>,
630 <citerefentry><refentrytitle>pacman</refentrytitle><manvolnum>8</manvolnum></citerefentry>,
631 <citerefentry><refentrytitle>systemd.slice</refentrytitle><manvolnum>5</manvolnum></citerefentry>,
632 <citerefentry><refentrytitle>machinectl</refentrytitle><manvolnum>1</manvolnum></citerefentry>