- 1 Installation
- 2 Basic Usage
- 3 Running Crashplan on a headless server
- 4 Troubleshooting
- 5 See also
Before accessing CrashPlan's graphical user interface, you should start the
CrashPlan can be configured entirely through its graphical user interface. To start the graphical interface:
To make CrashPlan automatically start upon system startup, enable the systemd unit.
Running Crashplan on a headless server
Running CrashPlan on a headless server is not officially supported. However, it is possible to do so.
The CrashPlan daemon's configuration files (in
/opt/crashplan/conf) are in an obscure XML format, and they are meant to be edited programmatically by the CrashPlan client.
CrashPlan 5 introduced a new client app which unfortunately dropped support for configuring a remote server with a local client, so you'll need to use X11 forwarding.
X11 forwarding over SSH
X11Forwarding is set to
yes in the headless server's
/etc/ssh/sshd_config and from another machine running X11, SSH to the headless machine with
-Y, and from the remote shell run
CrashPlanDesktop. The headless machine's windows will appear on the local X11 server. If you have problems, check
On CrashPlan v4.x and below, the client and daemon communicate on port 4243 by default. Thus, an easy way of configuring the CrashPlan daemon on a headless server is to create an SSH tunnel:
- Start the CrashPlan daemon on the server.
- Create an SSH tunnel. On the client:
ssh -N -L 4243:localhost:4243 headless.example.com.
- Start the CrashPlan client. (Again, the executable is named
Note that the authentication token (located in
/var/lib/crashplan/.ui_info) on the local and remote servers must match. More ideas can be found on these websites:
- The CrashPlan support site details a slightly more complicated method of tunneling traffic from the client (CrashPlan Desktop) to the daemon (CrashPlan Engine) through an SSH tunnel.
- A post by Bryan Ross details how to make CrashPlan Desktop connect directly to CrashPlan Engine. Note that this method can be less secure than tunneling traffic through an SSH tunnel.
Waiting for connection
On some systems it can happen that CrashPlan does not wait until an internet connection is established. If using NetworkManager, you can install AUR[broken link: archived in aur-mirror] which will automatically restart the CrashPlan service once a connection is successfully established.
Waiting for Backup
If the backup is stuck on «Waiting for Backup» even after you engage it manually, it might be that CrashPlan cannot access the tempdir or it is mounted as
noexec. It uses the default Java tmp dir which is normally
/tmp. You can either remove the
noexec mount option (not recommended) or change the tmpdir CrashPlan is using.
To change the tmpdir CrashPlan uses, open
/opt/crashplan/bin/run.conf and insert
SRV_JAVA_OPTS, for example:
SRV_JAVA_OPTS="-Djava.io.tmpdir=/var/tmp/crashplan -Dfile.encoding=UTF-8 …
Make sure to create the new tmpdir and verify CrashPlan's user has access to it.
# mkdir /var/tmp/crashplan
Restore stuck preparing
If a restore gets stuck at «Preparing», it may be due to a permission restriction on
/tmp that causes communication between the restore tool and backup engine to fail. This can be caused by the sysctl variable
fs.protected_fifos restricting the engine (running as
root) from connecting to a named pipe owned by the desktop user contained in
/tmp (similar to tmpfs#Opening symlinks in tmpfs as root fails).
The protection can be disabled to permit the restore to occur with
# sysctl fs.protected_fifos=0
See Sysctl for instructions on making the change permanent.
Desktop GUI Crashes on startup
On systems with Gnome 3 installed, or with libwebkit-gtk installed, there may be an issue where the GUI crashes on launch. This can be fixed by following the instructions here.
Out of Memory
For backup sets containing large numbers of files (more than 100,000 or so), the default maximum heap size of 512M may be too small. If this is filled, the server will silently restart, and will usually get stuck restarting as it continually reaches the memory limit. The only sign of this happening is the creation of many small log files in
/opt/crashplan/bin for each service restart (potentially hundreds of thousands, depending on how long it takes to notice the problem). To increase the heap size limit, adjust the
-Xmx option in
/opt/crashplan/bin/run.conf to a reasonable value for your system.
Real time protection
If you use real time protection for your backup set and have a lot of files to backup, the default system configuration might not be able to allocate all required handles to follow all files in real time. This issue can manifest itself with logs like "inotify_add_watch: No space left on device" in the syslog journal. CrashPlan Support has instructions here describing how to modify inotify max_user_watches to a bigger value to fix the issue. You cannot follow their instructions directly though, you need to create a new file in /etc/sysctl.d as /etc/sysctl.conf is now ignored by systemd. See sysctl#Configuration for more information.
JRE Version Update
If, during upgrade, CrashPlan is attempting to upgrade the self-installed JRE version and the upgrade never gets passed downloading the JRE from CrashPlan (checking in logs/upgrade<unique_number>.log, the last message is a curl/wget for the "latest" JRE tgz), it's possible to stop CrashPlan, download the JRE (from the ugprade log) manually and replace the jre folder in the CrashPlan install with the upgrade version. This should allow CrashPlan to get past being stuck trying to upgrade the JRE.
cd <crashplan/install/dir> ./bin/CrashPlanEngine stop rm -rf jre curl <jre url from crashplan log> tar xzvf <jre.tgz> ./bin/CrashPlanEngine start