the real reason to not use sigkill
the real reason to not use sigkill © Daniel Stori · TurnOff.us · CC BY-NC-SA 4.0
Why it lands
Signal handling as a family drama. A process can catch SIGTERM and use the time to let its threads finish and clean up; SIGKILL cannot be caught or ignored, so nothing gets to say goodbye. The practical rule matches the comic: send SIGTERM, wait, and reach for SIGKILL only when that fails.
Transcript
- This is Bob. Bob is a Linux process.
- HI!
- Like any process, Bob has his threads, with whom he shares context, memories and love.
- And like all processes, inevitably sometime he will be killed. When we gracefully kill a process with a soft SIGTERM...
- ...we give him the chance to talk with his kids about it. So, the kids finish their tasks...
- ...and say goodbye
- to each other. That's a process life!
- On the other hand, when we brutally kill a process with a SIGKILL, we prevent them from finishing their job and say goodbye...
- ...and this is so SAD!
- DAD!!
- So please, DON'T use SIGKILL. Give the kids the chance to leave the kernel in peace. Be nice.
- Dad, where are we going?
- Dad, where are you?
Credits
- Original
- the real reason to not use sigkill — Daniel Stori, TurnOff.us
- Image file
- turnoff.us/image/en/dont-sigkill.png
- License
- CC BY-NC-SA 4.0
The English page shows the original image unmodified. The Japanese page shows an adaptation by wtfe²: the lettering was erased from the art and replaced with a Japanese translation. The adaptation is shared under the same license as the original.