Script to monitor folder for new files?

Question

How can I immediately detect when new files added to a folder within a bash script? I would like the script to process files as soon as they are created in the folder. Are there any methods aside from scheduling a cron job that checks for new files each minute or so?

Solution 1

You should consider using inotifywait, as an example:

inotifywait -m /path -e create -e moved_to |
    while read dir action file; do
        echo "The file '$file' appeared in directory '$dir' via '$action'"
        # do something with the file
    done

In Ubuntu inotifywait is provided by the inotify-tools package. As of version 3.13 (current in Ubuntu 12.04) inotifywait will include the filename without the -f option. Older versions may need to be coerced. What is important to note is that the -e option to inotifywait is the best way to do event filtering. Also, your read command can assign the positional output into multiple variables that you can choose to use or ignore. There is no need to use grep/sed/awk to preprocess the output.

Solution 2

In case you came here for a simple, fast, handy solution reading the title, you could use watch

watch -n 0.1 ls <your_folder>

Monitors your folder and lists you everything in it every 0.1 seconds

Drawback

Non-scriptable (For scripting options, have a look at other answers)

Not real time, so if a file was created and deleted in less than 0.1 second, then this would not work, watch only supports minimum of 0.1 seconds.

Solution 3

I just cooked up this, and see no huge problems with it, other than a tiny chance of missing files in between checks.

while true
do
       touch  ./lastwatch
       sleep 10
       find /YOUR/WATCH/PATH -cnewer ./lastwatch -exec SOMECOMMAND {} \;
done

If your file processing doesn't take too long, you should not miss any new file. You could also background the activities... It's not bullet proof, but it serves some purposes without external tools like inotify.

Solution 4

I prefer incron, as its easier to manage. Essentially it's a service that leverages inotify and you can setup configurations to take action based on file change operations.

Ex:

<directory> <file change mask> <command or action>  options
/var/www/html IN_CREATE /root/scripts/backup.sh

You can see a full example here: http://www.cyberciti.biz/faq/linux-inotify-examples-to-replicate-directories/

Solution 5

entr

Using entr is the new way to do this (it's cross platform). Note entr doesn't use polling giving it a huge advantage over many of the alternatives.

Uses kqueue(2) or inotify(7) to avoid polling. entr was written to make rapid feedback and automated testing natural and completely ordinary.

On BSD it uses pledge(2)

You can install it with

apt-get install entr
dnf install entr
brew install entr

You can track a directory for new additions using

while $(true); do
  # echo ./my_watch_dir | entr -dnr echo "Running trigger..."
  echo ./my_watch_dir | entr -dnr ##MY COMMAND##
done;

Options explained (from the docs),

• -d Track the directories of regular files provided as input and exit if a new file is added. This option also enables directories to be specified explicitly. Files with names beginning with ‘.’ are ignored.
• -n Run in non-interactive mode. In this mode entr does not attempt to read from the TTY or change its properties.
• -r Reload a persistent child process. As with the standard mode of operation, a utility which terminates is not executed again until a file system or keyboard event is processed. SIGTERM is used to terminate the utility before it is restarted. A process group is created to prevent shell scripts from masking signals. entr waits for the utility to exit to ensure that resources such as sockets have been closed. Control of the TTY is not transferred the child process.

---

Note entr isn't polling there. It's after there is an operation on an inode it hasn't seen. Note this isn't a failsafe mechanism because the inode could be recycled to a different file name entirely. Ie,

mkdir foo;
cd foo;
touch bar;
touch baz;
ls -1i;
echo "delete bar; add quz"
rm bar;
touch quz;
ls -1i;

See the inotifywait answer for a better, and more powerful method of doing this. Albeit with a much worse interface.

View more solutions

Comments

[ztank1013]

Are you going to remove files from the folder once they are processed?

[Gilles 'SO- stop being evil']

possible duplicate of How to run a command when a directory's contents are updated?

[George Birbilis]

actually the answers at that other question mentioned as possible duplicate are less diversified and not that helpful in case of cross-platform need. In fact for cross-platform I'd vote for fswatch (emcrisostomo.github.io/fswatch), also mentioned at a comment in another answer below

[ihatetoregister]

Great! The inotifywait was just what I wanted.

[daisy]

Why is fflush() necessary here? Since a newline is already created I think

[enzotib]

@warl0ck: I think awk does not print each line as it can, but it waits for a bigger buffer (containing more than one line) before it decides to output something.

[Lugoues]

Just want to update this. You do not need awk to achieve this. you can filter the events with '-e create' and get only the filename by doing '-f %f' or the full path using '-f %w%f'. So the first line of the above script becomes: inotifywait -m /path -f %w%f -e create |

[mbaljeetsingh]

@Lugoues and now when you try to use -f you get The '--filename' option no longer exists. The option it enabled in earlier versions of inotifywait is now turned on by default. So, you only have to do inotifywait -m /path -e create | I'm going to try and edit this answer.

[Ankit]

This solution works for mounted remote filesystems. inotify-tools developer(s) is working on fuse (or was in mid 2014).

[Admin]

Now there is also a portable tool for it called fswatch. I did not write it, but it's open source and I use it.

[Michael Sacchi]

Good catch. I improved it a bit to support spaces in filenames.

[Michael Sacchi]

Absolutely. That's the way to go. Not really sure why I went down that road, I use -exec routinely.

[Farhan]

its not realtime. realtime is always best

[alightholder]

You shouldn't ever use ls for scripting. Use find or simple globbing instead: mywiki.wooledge.org/ParsingLs

[agrublev]

For this in need sudo apt-get install inotify-tools

[Devs love ZenUML]

Best solution if inotify is not available. I would add -type f to filter out files only. Otherwise the folder will also be returned.

[SDsolar]

Yep - the -f filename option is great. So then the only remaining question is how to get this to start upon reboot. I am going to use this with my solar plant to os.system("ssh me@mysystem ' ( touch /home/me/alarms/low24 ) '") so then the creation of this file will cause the master computer to use espeak and announce the low voltage. It already sends me an email but since my system already speaks the time at the top of the hour it has all the rest. askubuntu.com/questions/977613/…

[MKatleast3]

Even thou this is not a real-time, sleeping just a second is enough for me. Most importantly, this solution can watch newly created folders and files. This is great especially for monitoring Hadoop steps log which create new folders for every new step.

[Wender]

can you explain me the "while" line? Where did "path", "action" and "file" come from?

[Tim]

@Wender inotfiywait outputs 3 pieces of information on a single line when triggered. The 'read' bash builtin reads the input line and assigns each of the three pieces of information to a variable. Thus the first piece is assigned to the variable path, the second to action, and the third to file. Having assigned values to those variables, they are then available to be used later (like on the echo line). More information: tldp.org/LDP/Bash-Beginners-Guide/html/sect_08_02.html

[Joabe Lucena]

That was exactly what I was trying to remember! Thanks a lot!!

[Maksym Ganenko]

Please replace name path within read path action file with filepath because it may overwrite $PATH environment variable and so all your non-builtin commands within loop will report command not found. It happened within my fresh Arch Linux zsh terminal and I spent a lot of time googling and experimenting until I found the reason.

[bob]

incron never worked for me. based on SE questions about it, it seems like many people have unresolved issues with it.

[kodmanyagha]

this is very cool. I like it

[Bob]

For CentOS: yum --enablerepo=epel -y install inotify-tools

[ArcanisGK507]

this not work if the director is end Point of FTP point mount.

[George Birbilis]

For Solaris do note the comment "The behaviour depends on the monitor. Solaris File Events Notification will tell you that a directory has changed when a file has been created, but won't give you any more details. It's up to the receiver to scan the directory contents and see what file has been created." at github.com/emcrisostomo/fswatch/issues/228

[George Birbilis]

Also, note the license is GPL v3.0, so you should avoid statically linking to libfswatch (but dynamically link to its shared library instead) or prefer using the fswatch utility instead if you don't want to be forced to share the source code of your app too. Regarding having to share fswatch's source code, guess sharing a URL to its source code repository should be enough (that one has the most up to date version of the code anyway).

[MaXi32]

i use inotify for my bash script and most of the script have awk, sed. and reading again this answer carefully, I felt regret not reading about how to implement that 'built-in' variables path, action, file instead of using grep, etc.. thanks a lot for this answer

[user3821306]

I needed a bunch of other options inotifywait -me modify,create,delete /home/Mansoor/ITB/uploads. This works on Windows WSL2, ChromeOS, RHEL even Android! stackoverflow.com/questions/50168397/…

[Thanos M]

maybe this is a dumb question, but how can it achieve that without using polling

[Dawid Wolski]

Thanks. I've updated the link.

[clay]

how to run it as background daemon?

[MrE]

entr is great to restart a process when files changes, but it's not able to give you the name of the file that changed or was added to a folder: it gives the first file in the folder that was modified, which is not useful