It4innovations
Kill all executing tasks on worker if worker crashes
If a worker ends abruptly, its spawned children should be killed. Could be implemented using Linux process groups.
It is also connected to #66, that a each task should be spawned into a process group and when it is canceled we should clean all processes
@Kobzol How do you envision solving this with process groups? I could try to implement that in my spare time..
There was a nifty hack using DEATHSIG prctl in the child process but that won't help if the processes are forked from an interim thread
I have been playing with this for a bit and it seems to be quite tricky. We'll probably do some best effort mechanism using process groups and setsid to kill the group/session leader when a sigterm is received on the worker.
I didn't know about DEATHSIG, I'll take a look, thanks.