feat(jstzd): listen to os signals

[huancheng-trili]

Context

Part of JSTZ-189.

JSTZ-189

Description

Terminate jstzd on SIGTERM and SIGINT.

Note that we still need to find a way to shut down all processes when SIGKILL hits the server process. What happens now is that the subprocesses become orphans after the server is terminated.

Manually testing the PR

• Integration test: added back default_config with SIGTERM and added one test case with SIGINT

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 2 lines in your changes missing coverage. Please review.

Project coverage is 47.07%. Comparing base (c1cfe35) to head (a44e3cd).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/jstzd/src/lib.rs	0.00%	2 Missing ⚠️

Files with missing lines	Coverage Δ
crates/jstzd/src/lib.rs	22.72% <0.00%> (-5.85%)	⬇️

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c1cfe35...a44e3cd. Read the comment docs.

[ryutamago]

lgtm!
For handling the SIGKILL, i'm thikning we could assgin subprocesses to the same process group as the parent?
if we could set the process group ID of the main server process to its own PID, and make sure all the octez tasks inherits it, when the jstzd server process is killed, it should clean up all the sub processses.
I found a crate nix where we can use unix specfic process management feature!
wdyt ? @huancheng-trili @zcabter

[huancheng-trili]

Killing the server process won't take down the entire process group. Child processes do have a new parent but they keep running in the same group. It's tricky because we don't have control over the octez executables and thus we cannot make them listen to anything relevant. (btw with the current setup octez process are already in the same process group as the server process.)

If we want to guarantee that the sandbox is clean, we can use a container with everything running inside and create a wrapper server that handles path mapping and spawns the container, but that's a bit messy. For now we can probably just ignore this issue and deal with it later on.

[ryutamago]

@huancheng-trili

Killing the server process won't take down the entire process group

you are right here.

i've done more digging into it and seems like we could set the PR_SET_PDEATHSIG for Child Processes:
if we modify the the spawning of child processes to include a pre_exec closure that sets the PR_SET_PDEATHSIG to say SIGTERM, This allows that if the parent process is terminated (e.g., via SIGKILL), the child processes receive a SIGTERM and can shut down gracefully. The downside of this approach is that we will have to implment the listener for all of the octez child processes (node, baker, rollup)

[zcabter]

We discussed potential solutions offline, namely requiring a sidecar process. I think we can put SIGKILL support on the backlog for now