Occasionally outside your control you need to run an application in docker that might write to disk. If that program doesn't clean up for some reason eventually your docker container might become full.
A lot of applications will catch write errors and attempt to handle them. This can be a disadavantage when trying to handle disk space failures as the application will still pass health checks but incorrectly serve requests.
spacekill solves this by overriding the write call with a version that checks error codes for ENOSPC and EDQUOT (free space, and disk quota full errors). If it detects this, it termiates the program.
gcc -Wall -fPIC -shared -o spacekill.so spacekill.c
Start your program with an LD_PRELOAD set to spacekill.so
LD_PRELOAD=spacekill.so python example.py
An example docker file is included.
91 - spacekill termianted the application because a write failed
I'm not a C dev. I have no idea what horrible things this will cause. I have not tested this in production. I have barely tested this myself. This doesn't check if the file descriptor is pointing to a real file. This was made in literally in a couple of hours after a long drive.