System administrators have relied on cron+rsync for years to constantly synchronize files and directories to remote machines. However, this technique has a huge disadvantage for content distribution with near real-time requirements, e.g. podcasts and blogging.

It is not feasible to let authors wait for their content to get synchronized every x hours with regard to the enormous pace of articles and podcasts nowadays.

The inosync daemon leverages the inotify service available in recent linux kernels to monitor and synchronize changes within directories to remote nodes.

inosync [OPTIONS]

  -c FILE     load configuration from FILE
  -d          daemonize (fork to background)
  -p          do not actually call rsync
  -v          print debugging information
  --version   show program's version number and exit
  -h, --help  show this help message and exit

Configuration files are simple python scripts, merely declaring necessary variables. Below is an example configuration to synchronize /var/www except /var/www/localhost to 3 remote locations:

# array of directories that should be watched for changes
wpaths = ["/var/www/"]

# exclude list for rsync
rexcludes = [
      "/localhost",
]

# rpaths has one-to-one correspondence with wpaths for syncing multiple directories
rpaths = ["/var/www/"]

# remote locations in rsync syntax
rnodes = [
      "[email protected]:",
      "[email protected]:",
      "[email protected]:",
]

# extra, raw parameters to rsync
#extra = "--rsh=ssh -a"

# limit remote sync speed (in KB/s, 0 = no limit)
#rspeed = 0

# event mask (only sync on these events)
#emask = [
#     "IN_CLOSE_WRITE",
#     "IN_CREATE",
#     "IN_DELETE",
#     "IN_MOVED_FROM",
#     "IN_MOVED_TO",
#]

# event delay in seconds (this prevents huge
# amounts of syncs, but dicreases the
# realtime side of things)
#edelay = 10

# rsync binary path
#rsync = "/usr/bin/rsync"

There are no known bugs currently, however, due to the design of inosync, there are several shortcomings:

• inosync cannot parse rsync excludes and therefore calls rsync on changes in excluded directories as well. (of course rsync still excludes these directories)
• It is easily possible to flood the daemon with huge amounts of change events, potentially resulting in enormous bandwidth and connection usage.
• Excludes currently apply to each path to be synced and there is no way to specify per-path excludes yet.

To use this script you need the following software installed on your system:

• linux-2.6.13 or later
• Python-2.5 or later
• pyinotify-0.8.7 or later

inosync is similar to lsyncd, but uses a lot less (nasty) magic to parse rsync excludes and shared www directories. Additionally inosync has no limitation on filename size and number of active watchpoints.

A comparision to other approaches like DRBD, incron and FUSE can be found at lsyncds project page, mentioned above.