How to use heroku's ephemeral filesystem
I'm using Python/Django on Heroku (Cedar Stack) and I've got a management command that I need to write that will pull a file out of an S3 bucket and process it. I'm not sure I understand how to use the ephemeral filesystem. Are there only certain directories that are writeable? I found an other article that implied that there were only certain folders that were writable (but, it doesn't seem to apply to the Cedar stack). I found this dev article but it doesn't go into much detail (note: I do understand that it's just temporary. I only need to unzip the file and process the file). Can I just create a folder anywhere under the application's root? And how would I get that? It seems like I could probably just use $HOME. I did a bit of testing by connecting to via
$ heroku run bash
and running:
$ echo #HOME
returns:
/app
and running:
$ mkdir $HOME/tmp
creates a folder in the app's root and gives with the same user and group as the other files and folders.
So... anything I'm missing here? A better way to do it? Is there an OS environment variable for this? I've run "env" and I don't see a better one.
To really understand the ephemeral filesystem, you need to understand what a dyno is. You can read more about how dynos work. In a nutshell, though, a process runs on Heroku in a virtual machine with its own filesystem. That virtual machine can stop for a number of reasons, taking the filesystem along with it.
The underlying filesystem will be destroyed when an app is restarted, reconfigured (e.g. heroku config ...), scaled, etc. For example, if you have two web dynos, write some files to the ephemeral filesystem, and scale to three dynos, those files will be destroyed because your app is running on new dynos.
In general, the ephemeral filesystem works just like any filesystem. Directories you have permission to write to, such as $HOME and /tmp, you can write files to. Any files that require permanence should be written to S3, or a similar durable store. S3 is preferred as Heroku runs on AWS and S3 offers some performance advantages. Any files that can be recreated at will can be stored on the dyno's ephemeral store.
You can create a file under the '/tmp' directory, and that file will be destroyed after the request is complete. I'm doing this on Cedar, and I haven't had any problems.