PR #512 tests that zone persistence works.
PR #550 implements a first draft version of zone persistence.
However the test added in #512 doesn't work. After updating it to save state now instead of mark it dirty to save after a delay it still didn't work. While investigating why I realized that the pkill issued by the test to stop Cascade doesn't cause Cascade to write its state to disk.
Looking in main.rs Cascade waits for CTRL-C but not for SIGTERM.
Replacing:
let res = match tokio::signal::ctrl_c().await {
Ok(_) => ExitCode::SUCCESS,
Err(error) => {
error!("Listening for CTRL-C (SIGINT) failed: {error}");
ExitCode::FAILURE
}
};With:
info!("Running");
sigterm::wait().await;
info!("Shutting down");Resolves the issue (using the sigterm crate), and also makes it clear when Cascade has started and when it is stopping, the latter wasn't recorded anywhere currently in the logs.
I understood that some thought had gone into ensuring state was persisted so I'm assuming I am missing something, and am not sure that we want a dependency on another crate, both of which are why this is being reported as an issue rather than as a PR.
PR #512 tests that zone persistence works.
PR #550 implements a first draft version of zone persistence.
However the test added in #512 doesn't work. After updating it to save state now instead of mark it dirty to save after a delay it still didn't work. While investigating why I realized that the
pkillissued by the test to stop Cascade doesn't cause Cascade to write its state to disk.Looking in
main.rsCascade waits for CTRL-C but not for SIGTERM.Replacing:
With:
Resolves the issue (using the
sigtermcrate), and also makes it clear when Cascade has started and when it is stopping, the latter wasn't recorded anywhere currently in the logs.I understood that some thought had gone into ensuring state was persisted so I'm assuming I am missing something, and am not sure that we want a dependency on another crate, both of which are why this is being reported as an issue rather than as a PR.