metrics: stabilize num_alive_tasks

[rcoh]

Motivation

Stabilize num_active_tasks num_alive_tasks so it can be used outside of --cfg tokio_unstable

Solution

1. stabilize num_alive_tasks
2. Rename internal fields to match num_alive_tasks
3. Split rt_metrics into rt_metrics and rt_unstable_metrics.

Refs: #6546

[mox692]

Since #6114 has been merged, we need to resolve conflict here.
Also, perhaps we should stabilize num_active_tasks instead of active_tasks_count?

[mox692]

Looks good to me.

[Darksonn]

If this is being stabilized, then the documentation should no longer say that it's unstable.

[rcoh]

Ah that comment was attempting to refer to spawned_task_count

…

On Sat, Jun 15, 2024, 3:06 PM Alice Ryhl ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In tokio/src/runtime/metrics/runtime.rs <#6619 (comment)>: > @@ -47,6 +47,29 @@ impl RuntimeMetrics { self.handle.inner.num_workers() } + /// Returns the current number of active tasks in the runtime. + /// + /// This value increases and decreases over time as tasks are spawned and as they are completed or cancelled. + /// + /// To see the total number of spawned tasks, see `spawned_tasks_count`. Note that this API currently requires using `--cfg tokio_unstable`. If this is being stabilized, then the documentation should no longer say that it's unstable. — Reply to this email directly, view it on GitHub <#6619 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADYKZZO5E3NIQNX465VC7LZHSGEFAVCNFSM6AAAAABI5FHB4SVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDCMRQG44TGNZYHE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[Darksonn]

Okay. It's a bit confusing like this. You don't need to say that it's unstable in the docs for other methods.

[Darksonn]

I'm sorry for responding slowly here.

I'm wondering whether the name should be something else. The word "active" is the opposite of "idle", but the count includes idle tasks that have not yet exited. How about a name like alive_tasks_count?

[rcoh]

runnable? That has some issues too. I wonder if this would be helped with a TaskMetrics struct that could give some docs to explain the different task states & group the metrics.

…

On Sun, Jun 30, 2024, 4:46 AM Alice Ryhl ***@***.***> wrote: I'm sorry for responding slowly here. I'm wondering whether the name should be something else. The word "active" is the opposite of "idle", but the count includes idle tasks that have not yet exited. How about a name like alive_tasks_count? — Reply to this email directly, view it on GitHub <#6619 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADYKZ5ICVVOKM5II2OKRUTZJ7OXTAVCNFSM6AAAAABI5FHB4SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOJYGUYTKNJRGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[Darksonn]

I ended up renaming it in #6667.

[Darksonn]

I don't think runnable works either. That also implies that idle tasks shouldn't be counted, but they are.

[rcoh]

Sorry for the delay on this. @Darksonn I've rebased. I noticed there was a mix between alive_tasks_count and num_alive_tasks — would you like me to unify them or leave as is?

[Darksonn]

That makes sense. The publicly used name is num_alive_tasks, so let's rename the internal things called alive_tasks_count to num_alive_tasks.

[Darksonn]

Thanks.