I started using Heroku for one of the projects recently. In past, I have used Openshift, Webfaction (and AWS long time ago). Each of these gave some kind of ssh access. By ssh access, I mean access to the running system.

Heroku is different.

Heroku gives you bash access. At first, it may seem similar to the ssh access, but it is not. Each time you run heroku run bash --app my_app Heroku creates a new instance of the shell. The filesystem is ephemeral i.e. if you have two such bash shells are running, and if you write something to /tmp from one instance, it won't be visible to the other bash instance.

Since there is no “instance” running (at least in a traditional sense) one can't send or receive files from the application “instance”

The only way to “send” files to application instance is to have those files committed to git and push them to Heroku. After all, that is how you update the code on Heroku.

Now, that's a problem.

Create Files

One might ask “why” would I want to send files to Heroku which are not in git

Here is an example:

I have scripts that create some artifacts by reading the CSV file as an input. I may need to run these scripts from time to time with different CSV files each time. The input files are temporary in nature. They may also contain sensitive information which I do not want to commit to git

Normally, I would ftp the files to /tmp of the machine, ssh to the instance and run the script by pointing it to /tmp/myfile.csv

As explained earlier, this does not work for Heroku

So I created a support ticket with Heroku, and here is an option they suggested.

First “login” to Heroku via heroku run bash --app my_app

$ mkdir CSV_files
$ cd CSV_files
$ cat <<'EOF' > myfile.csv
first_name, last_name
John, Doe
Jane, Doe
Mickey, Mouse
Minnie Mouse
$ cd
$ python scripts/script1.py -i CSV_files/myfile.csv

This is not the most optimal way.

I run cat myfile.csv on my local machine, and then copy/paste to the Heroku terminal session, after the cat line

Finally, I type in EOF manually, to finish the input.

This is not the most optimal solution, e.g. for long files, this could be a problem.

While there could be side effects due to how terminals handle “copy/paste”, this seems to be the only option if one doesn't want the files in git

Save Files

Now that I am able to run the script on Heroku, I need to save the output from the script. Most of the times, this is not a problem. I use the “copy/paste from the terminal” But sometimes, the output is so long that it scrolls past and does not fit the terminal window's buffer.

Normally, I would redirect the output to a file (in /tmp) and FTP off the server, but again, it doesn't work with Heroku.

I came across this post on SO.

So there is this service that would accept files from anyone, anywhere, called transfer.sh. You can check the details on their site.

I was a bit hesitant letting someone-I-don't-know, host my files, but it was only the log output, that did not contain any sensitive data.

If you have any sensitive data, you can encrypt it, before uploading

You can also control the duration for which your files are accessible. The default is 14 days.

I understand that neither of these solutions is optimal, but they work in a pinch.

Flesch-Kincaid reading ease score: 93.73

Flesch-Kincaid grade level score: 2.98