Page History

Install and configure the SMF service for rsync and the daemon itself

Before starting with this guide, make sure that rsync software is installed on the machine. It is not installed per default in local zones, but can be done so with pkg: 

pkg install network/rsync

Prepare files for SMF services configuration

To see all attachments described in this page, click on the small paper clip icon at the top left.

Add SMF manifest and method scripts, and initial configuration for rsync itself (as root):

• copy (or wget) the attached file "rsyncd.xml" to: /var/svc/manifest/network/rsyncd.xml
• copy (or wget) the attached file "rsyncd" to: /lib/svc/method/rsyncd

• set the execute permission on the above file for user root:

  Code Block
  chmod u+x /lib/svc/method/rsyncd

  (Thanks to http://robbiecrash.me/?p=22 for pointing out that I missed the above chmod step)

• copy (or wget) the attached file "rsync" to: /etc/default/rsync
  NOTE that the variable RSYNC_ENABLE is already set to true in this file.
• now create a valid /etc/rsyncd.conf file. This must be done or the service will not run.
  For a full list of options see: man rsyncd.conf

For an example see this trivial /etc/rsyncd.conf file:

use chroot = yes
read only = yes
log file = /var/adm/rsyncd_upload.log
log format = - %a - %f
transfer logging = yes

#module to share ISO files
[ISO]
path = /lift/data/ISO
comment = public ISO repository

Import the SMF service manifest

Now you can import the SMF service manifest with the following code:

svccfg -v import /var/svc/manifest/network/rsyncd.xml

Next check to see the status of the rsyncd service – it should be defined but may be not running yet:

svcs -a | grep rsyncd

...you should see something like this:

offline          2:04:55 svc:/network/rsyncd:default

Enable the SMF service

Now you need to enable the service:

svcadm enable network/rsyncd

Test that the service is enabled and the daemon actually runs

Now execute the svcs command again and also execute ps:

svcs -a | grep rsync

The service should be listed as online now:

online         May_16   svc:/network/rsyncd:default

Now execute:

ps aux | grep rsync

you should see something like this

root     13351  0.0  0.0 2688 1340 ?        S 02:04:55  0:00 /usr/bin/rsync --d

Test from a client

Note – if you add new modules to /etc/rsyncd.conf you should not have to restart the rsyncd service, as the file is read each time an rsync client connects.

Congrats, your rsyncd should be working now. Test it out from a client machine using a small file for testing with something like:

rsync servername::ISO/some_file.iso /localpath/here/

or just get a list of the files and/or directories by using verbose mode and piping to less:

rsync -v servername::ISO | less

Consult the log file referenced in your rsyncd.conf file to see info on file transfers. (In this example /var/adm/rsyncd_upload.log).

Thanks to Marcelo Leal for the rsyncd method file (changed here to point to /usr/bin/rsync) and also for parts of the service manifest.

Optional – add automatic ZFS snapshots after rsync uploads

You might want to create ZFS snapshots linked to your incoming rsync upload jobs – for example, after a machine backing up to your OpenIndiana server disconnects its rsyncing session. Tricks like this can be done by defining a post-xfer-exec hook pointing to a script which handles the logic, in this case rsync-zfshot.sh.

Expanding on a trivial example above, here is another one:

use chroot = yes
read only = yes
log file = /var/adm/rsyncd_upload.log
log format = - %a - %f
transfer logging = yes
pid file = /var/run/rsyncd.pid
lock file = /var/run/rsyncd.lock

#module to share ISO files
[ISO]
path = /lift/data/ISO
comment = public ISO repository

#Module available only to a multihomed host which backs up into this path
#and chowns the files to specified UID:GID and sets FS access rights
#(source is a Windows machine with separate user namespace); it is also
#recommended that each backed-up host has a dedicated backup ZFS dataset.
#When a backup attempt is complete, a ZFS snapshot is created, so we can
#walk back to any given complete backup image (i.e. via .zfs/snapshots
#directory with Samba/kCIFS sharing, or via MS Shadow Volumes with kCIFS)
[backup-jamais]
    path = /export/DUMP/windows/JAMAIS
    read only = no
    auth users = backup-jamais
    charset = UTF-8
    ### NOTE: cwRsync clients might require "--iconv=cp1251" on their command
    ###       line to properly upload russian-named files from Windows
    uid = backup-jamais
    gid = backup-jamais
    incoming chmod = Dug=rwX,o-rwx,Fug=rw,o-rwx
    hosts allow = 192.168.1.101/32 123.45.68.101/32
    post-xfer exec = /root/rsync-zfshot.sh -w -v -l -ok

The main trickery for this particular quest is in the rsync-zfshot.sh script attached to this post. It has a number of command-line options to control logging and verbosity, as well as to enable actual snapshotting (dry-runs by default), and perhaps only when the rsync upload was successful (without an rsync error status set). Hopefully, the code and comments (and command-line help) are its best documentation.

• copy (or wget) the attached file "rsync-zfshot.sh" to: /root/rsync-zfshot.sh or any other location referenced in your config file module above


• set the execute permission on the above file for user root:

  Code Block
  chmod u+x /root/rsync-zfshot.sh

• You might be required to also delegate the ZFS snapshooting access rights to the backup user (but test first to see if this is really needed):

  Code Block
  zfs allow -l -d -u backup-jamais snapshot,rename,mount \ rpool/export/DUMP/windows/JAMAIS

  NOTE: the -l -d options set this permission locally on the named dataset and on its descendants, including those created in the future.