Implement an optional method using sigstop/sigcont instead of sigkill

[gto2023]

make, ninja and therefore package managers like Gentoo's portage allow to run compilation jobs in parallel. Sometimes, if a few jobs simultaneously require very large amounts of memory (as gcc with lto or qtwebengine regularily do) , the swapper starts trashing pages, prolonging the installation of such a package from hours to days and equally impeding user interaction .

Killing an offending task would be pointless in such a situation, whereas manually stopping each of such tasks and allowing to continue e.g. only one of them at a time until the memory pressure subsides does work very well.

[havedill]

Similar to this request -

I'd like to send SIGKILL instead of SIGTERM. When we send sigterm, systemd seems to treat it like it is exiting gracefully, and runs our services stop scripts (ExecStop) which tells our monitoring "we expect this service to be going down".

If i could send a SIGKILL, systemd will label the process as failed and my alerting would work to notify us that the process went down unexpectedly

[rfjakob]

I'd like to send SIGKILL instead of SIGTERM

This already works. You just set the SIGKILL limit to the same as SIGTERM. For example:

$ earlyoom -m 10,10 -s 10,10
earlyoom v1.8-3-g9f20cdc
mem total: 23883 MiB, user mem total: 21653 MiB, swap total: 8191 MiB
sending SIGTERM when mem avail <= 10.00% and swap free <= 10.00%,
        SIGKILL when mem avail <= 10.00% and swap free <= 10.00%

[rfjakob]

manually stopping each of such tasks

Problem here is, how many processes do you suspend? Because suspending a process will not reduce memory usage.