Notify the service state through sd_notify
#3903
Conversation
Deploying matrix-authentication-service-docs with
|
| Latest commit: |
d57552c
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://32c552c0.matrix-authentication-service-docs.pages.dev |
| Branch Preview URL: | https://quenting-lifecycle-manager.matrix-authentication-service-docs.pages.dev |
| @@ -32,6 +33,7 @@ rand.workspace = true | |||
| rand_chacha = "0.3.1" | |||
| reqwest.workspace = true | |||
| rustls.workspace = true | |||
| sd-notify = "0.4.5" | |||
There was a problem hiding this comment.
does this need to be gated behind a flag, or is it cross-platform enough not to be a problem?
There was a problem hiding this comment.
It does use stuff from std::os::unix, which is 'Unix-only'. We don't claim to support Windows, so I'm happy keeping it like that
| if let Err(e) = sd_notify::notify(false, states) { | ||
| tracing::error!( | ||
| error = &e as &dyn std::error::Error, | ||
| "Failed to notify service manager" |
There was a problem hiding this comment.
is this error expected if someone is not using systemd? If so, should we soften it somewhat so it doesn't sound like a major problem for them?
Maybe it would be good to elaborate in the comment what this is doing, I'm not sure most people would understand what this is for without having seen it before. (at least, I don't think knowledge about this is that common, but especially not amongst non-systemd users)
There was a problem hiding this comment.
If NOTIFY_SOCKET is not set, it returns Ok(()) and is basically a no-op
|
Actually, tokio doesn't like that we create an interval with a zero duration… so it crashed if the watchdog environment variables were not set. Fixed that in d57552c |
This does mainly two things:
The API of the
sd-notifycrate is… weird. I was tempted to implement myself, but the fact that we need to be able to callclock_gettimewithCLOCK_MONOTONICmakes me prefer someone else having thatunsafecode already written.Fixes #3879