main: handle termination signals #152
Conversation
So that we can log some of the initial errors instead of printing them.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #152 +/- ##
======================================
Coverage 0.00% 0.00%
======================================
Files 1 1
Lines 105 126 +21
======================================
- Misses 105 126 +21 ☔ View full report in Codecov by Sentry. |
|
Want to note that since you already use |
|
@tnull Good point, thank you! |
orbitalturtle
force-pushed
the
handle-sigterm
branch
from
dd9e22a to
f94892c
Compare
orbitalturtle
requested review from
mrfelton and
dunxen
and removed request for
mrfelton
| let mut sigterm_stream = tokio::signal::unix::signal(SignalKind::terminate()) | ||
| .map_err(|e| error!("Error initializing sigterm signal: {e}."))?; | ||
| let mut sigint_stream = tokio::signal::unix::signal(SignalKind::interrupt()) | ||
| .map_err(|e| error!("Error initializing sigint signal: {e}."))?; | ||
|
|
||
| tokio::spawn(async move { | ||
| tokio::select! { | ||
| _ = sigint_stream.recv() => { | ||
| info!("Received CTRL-C, shutting down.."); | ||
| shutdown.trigger(); | ||
| } | ||
| _ = sigterm_stream.recv() => { | ||
| info!("Received SIGTERM, shutting down.."); | ||
| shutdown.trigger(); | ||
| } | ||
| } | ||
| }); |
There was a problem hiding this comment.
Maybe we can add an issue to get integration tests for this up? I can look into it then.
There was a problem hiding this comment.
I had actually already gotten pretty far into writing a basic integration test for this, before realizing it was probably too big for this PR heh. I'll either put it up as a draft or finish it off. But I have some further integration test ideas I'll post as issues today, if you happen to be interested.
This PR fixes #150 with the following:
signalfeature to process SIGTERM and SIGINT signals, tying them to thetriggeredLifecycleSignals we had in place already.serve_with_shutdowntonic so the server shuts down more gracefully. Maybe it'd be worth looking to see if we can do more for a graceful shutdown later on. With this change, the server seems to work normally. But I'm realizing we don't have any quick integration tests of the server in place. This doesn't necessarily need to go in this PR... but if I have time I'll add something quick.