Langue: en

Autres versions - même langue

Version: 22 Aug 2010 (fedora - 01/12/10)

Section: 7 (Divers)


schroot - frequently asked questions


This manual page covers various frequently asked questions about configuration and usage of schroot.


Why is schroot overwriting configuration files in the chroot?

By default, schroot copies over the system NSS databases ([oq]passwd[cq], [oq]shadow[cq], [oq]group[cq], [oq]gshadow[cq], [oq]services[cq], [oq]protocols[cq], [oq]networks[cq], and [oq]hosts[cq], etc.) into the chroot. The reason for this is that the chroot environment is not a completely separate system, and it copying them over keeps them synchronised. However, this is not always desirable, particularly if installing a package in the chroot creates system users and groups which are not present on the host, since these will disappear next time the databases are copied over.

The suggested workaround here is to disable the copying by removing or commenting out the databases in the NSSDATABASES file in the script-config file for the chroot. These are /etc/schroot/default/nssdatabases and /etc/schroot/default/config by default.

In the future, we will be working on a better scheme for keeping the host and chroot databases in sync which can merge entries rather than overwriting the entire database, which would preserve chroot-specific changes.

Should I use the plain or directory chroot type?

These two chroot types are basically equivalent, since they are both just directories in the filesystem. plain is very simple and does not perform any setup tasks; the only reason you would want to use it is if you're upgrading from a program such as dchroot(1) or chroot(8) which don't do anything other than running a command or shell in a directory. On the other hand, directory chroots do run setup scripts, which can mount additional filesystems and do other setup tasks.


What are snapshots and unions?

Some chroot types support cloning. This means when you start a session, you get a copy of the chroot which lasts just for the lifetime of the session. This is useful when you want a temporary clean copy of a system for a single task, which is then automatically deleted when you're done with it. For example, the Debian package build d[ae]mons run sbuild(1) to build Debian packages, and this program uses schroot to create a clean build environment for each package. Without snapshotting, the chroot would need to be reset to its initial state at the end of each build to make it ready for the next one, and any debris left over from package removals or earlier builds could interfere with the next build.

The most commonly-used snapshotting method is to use LVM snapshots (chroot type [oq]lvm-snapshot[cq]). In this case the chroot must exist on an LVM logical volume (LV); snapshots of an LV may then be made with lvcreate(8) during chroot session setup. However, these use up a lot of disk space. A newer method is to use Btrfs snapshots which use up much less disk space (chroot type [oq]btrfs-snapshot[cq]), and may be more reliable than LVM snapshots. Btrfs is however still experimental, but it is hoped that it will become the recommended method as it matures.

Unions are an alternative to snapshots. In this situation, instead of creating a copy of the chroot filesystem, we overlay a read-write temporary filesystem on top of the chroot filesystem so that any modifications are stored in the overlay, leaving the original chroot filesystem untouched. The Linux kernel has yet to integrate support for union filesystems such as aufs and unionfs, so LVM snapshots are still the recommended method at present.


Can I run a d[ae]mons in a chroot?

A common problem is trying to run a d[ae]mon in a chroot, and finding that this doesn't work. Typically, the d[ae]mon is killed shortly after it starts up.

When schroot runs, it begins a session, runs the specified command or shell, waits for the command or shell to exit, and then it ends the session. For a normal command or shell, this works just fine. However, d[ae]mons normally start up by running in the background and detaching from the controlling terminal. They do this by forking twice and letting the parent processes exit. Unfortunately, this means schroot detects that the program exited (the d[ae]mon is a orphaned grandchild of this process) and it then ends the session. Part of ending the session is killing all processes running inside the chroot, which means the d[ae]mon is killed as the session ends.

In consequence, it's not possible to run a d[ae]mon directly with schroot. You can however do it if you create a session with --begin-session and then run the d[ae]mon with --run-session. It's your responsibility to end the session with --end-session when the daemon has terminated or you no longer need it.

How do I manually cleaning up a broken session?

Occasionally, it may be necessary to manually clean up sessions. If something changes on your system which causes the setup scripts to fail when ending a session, for example removal of a needed file or directory, it may not be possible for schroot to clean everything up automatically. For each of the session directories listed in the [lq]Session directories[rq] section in schroot(1), any files with the name of the session ID need deleting, and any directories with the name of the session ID need umounting (if there are any filesystems mounted under it), and then also removing.

For example, to remove a session named my-session by hand:

Remove the session configuration file
CR]% CB]rm /var/lib/schroot/session/my-session[CR]
Check for mounted filesystems
CR]% CB]/usr/libexec/schroot/schroot-listmounts -m /var/lib/schroot/mount/my-session[CR]
Unmount any mounted filesystems
Remove /var/lib/schroot/mount/my-session
Repeat for the other directories such as /var/lib/schroot/union/underlay, /var/lib/schroot/union/overlay and /var/lib/schroot/unpack

NOTE: Do not remove any directories without checking if there are any filesystems mounted below them, since filesystems such as /home could still be bind mounted. Doing so could cause irretrievable data loss!


How do I use sessions?

In normal use, running a command might look like this:
CR]% CB]schroot -c squeeze -- command[CR]

which would run the command command in the squeeze chroot. While it's not apparent that a session is being used here, schroot is actually doing the following steps:

Creating a session using the squeeze chroot. This will be automatically given a unique name, such as squeeze-57a69547-e014-4f5d-a98b-f4f35a005307, though you don't usually need to know about this
Setup scripts are run to create the session chroot and configure it for you
The command command is run inside the session chroot
Setup scripts are run to clean up the session chroot
The session is deleted

Now, if you wanted to run more than one command, you could run a shell and run them interactively, or you could put them into shell script and run that instead. But you might want to do something in between, such as running arbitrary commands from a program or script where you don't know which commands to run in advance. You might also want to preseve the chroot state in between commands, where the normal automatic session creation would reset the state in between each command. This is what sessions are for: once created, the session is persistent and won't be automatically removed. With a session, you can run as many commands as you like, but you need to create and delete the session by hand since schroot can't know by itself when you're done with it unlike in the single command case above. This is quite easy:
CR]% CB]schroot --begin-session -c squeeze[CR]

This created a new session based upon the squeeze chroot. The unique name for the session, the session ID, was printed to standard output, so we could also save it as a shell variable at the same time like so:
CR]% CB]SESSION=$(schroot --begin-session -c squeeze)[CR]

Now we have created the session and got the session ID, we can run commands in it using the session ID:
CR]% CB]schroot --run-session -c squeeze-57a69547-e014-4f5d-a98b-f4f35a005307 -- command1[CR]

CR]% CB]schroot --run-session -c "$SESSION" -- command1[CR]

and then as many more commands as we like
CR]% CB]schroot --run-session -c "$SESSION" -- command2[CR]
CR]% CB]schroot --run-session -c "$SESSION" -- command3[CR]
CR]% CB]schroot --run-session -c "$SESSION" -- command4[CR]


When we are done with the session, we can remove it with --end-session:
CR]% CB]schroot --end-session -c squeeze-57a69547-e014-4f5d-a98b-f4f35a005307[CR]

CR]% CB]schroot --end-session -c $SESSION[CR]

Since the automatically generated session names can be long and unwieldy, the --session-name option allows you to provide you own name:

CR]% CB]schroot --begin-session -c squeeze --session-name my-name[CR]


Getting help and getting involved

The mailing list CR]<> is used for both user support and development discussion. The list may be subscribed to from the project website at CR] or the Mailman list interface at CR]

Reporting bugs

On Debian systems, bugs may be reported using the reportbug(1) tool, or alternatively by mailing CR]<> (see CR] for details on how to do that).

Getting the latest sources

schroot is maintained in the git version control system. You can get the latest sources from CR]git://
CR]% CB]git clone git://[CR]

The master branch containes the current development release. Stable releases are found on branches, for example the 1.4 series of releases are on the schroot-1.4 branch.


Roger Leigh. Copyright © 2005-2010 Roger Leigh CR]<>

schroot is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.


dchroot(1), schroot(1), sbuild(1), schroot-setup(5), schroot.conf(5).