temporalio 0.0.0 → 0.0.1

data/bridge/sdk-core/core/src/worker/activities.rs

@@ -0,0 +1,464 @@
+mod activity_heartbeat_manager;
+mod local_activities;
+
+pub(crate) use local_activities::{
+    DispatchOrTimeoutLA, ExecutingLAId, LACompleteAction, LocalActRequest,
+    LocalActivityExecutionResult, LocalActivityManager, LocalActivityResolution,
+    LocalInFlightActInfo, NewLocalAct,
+};
+
+use crate::{
+    abstractions::{MeteredSemaphore, OwnedMeteredSemPermit},
+    pollers::BoxedActPoller,
+    telemetry::metrics::{activity_type, activity_worker_type, workflow_type, MetricsContext},
+    worker::{
+        activities::activity_heartbeat_manager::ActivityHeartbeatError, client::WorkerClient,
+    },
+    PollActivityError, TaskToken,
+};
+use activity_heartbeat_manager::ActivityHeartbeatManager;
+use dashmap::DashMap;
+use governor::{
+    clock::DefaultClock,
+    middleware::NoOpMiddleware,
+    state::{InMemoryState, NotKeyed},
+    Quota, RateLimiter,
+};
+use std::{
+    convert::TryInto,
+    sync::Arc,
+    time::{Duration, Instant},
+};
+use temporal_sdk_core_protos::{
+    coresdk::{
+        activity_result::{self as ar, activity_execution_result as aer},
+        activity_task::{ActivityCancelReason, ActivityTask},
+        ActivityHeartbeat,
+    },
+    temporal::api::{
+        failure::v1::{failure::FailureInfo, CanceledFailureInfo, Failure},
+        workflowservice::v1::PollActivityTaskQueueResponse,
+    },
+};
+use tokio::sync::Notify;
+
+#[derive(Debug, derive_more::Constructor)]
+struct PendingActivityCancel {
+    task_token: TaskToken,
+    reason: ActivityCancelReason,
+}
+
+/// Contains minimal set of details that core needs to store while an activity is running.
+#[derive(Debug)]
+struct InFlightActInfo {
+    pub activity_type: String,
+    pub workflow_type: String,
+    start_time: Instant,
+}
+
+/// Augments [InFlightActInfo] with details specific to remote activities
+struct RemoteInFlightActInfo {
+    pub base: InFlightActInfo,
+    /// Used to calculate aggregation delay between activity heartbeats.
+    pub heartbeat_timeout: Option<prost_types::Duration>,
+    /// Set to true if we have already issued a cancellation activation to lang for this activity
+    pub issued_cancel_to_lang: bool,
+    /// Set to true if we have already learned from the server this activity doesn't exist. EX:
+    /// we have learned from heartbeating and issued a cancel task, in which case we may simply
+    /// discard the reply.
+    pub known_not_found: bool,
+    /// The permit from the max concurrent semaphore
+    _permit: OwnedMeteredSemPermit,
+}
+impl RemoteInFlightActInfo {
+    fn new(
+        activity_type: String,
+        workflow_type: String,
+        heartbeat_timeout: Option<prost_types::Duration>,
+        permit: OwnedMeteredSemPermit,
+    ) -> Self {
+        Self {
+            base: InFlightActInfo {
+                activity_type,
+                workflow_type,
+                start_time: Instant::now(),
+            },
+            heartbeat_timeout,
+            issued_cancel_to_lang: false,
+            known_not_found: false,
+            _permit: permit,
+        }
+    }
+}
+
+struct NonPollActBuffer {
+    tx: async_channel::Sender<PermittedTqResp>,
+    rx: async_channel::Receiver<PermittedTqResp>,
+}
+impl NonPollActBuffer {
+    pub fn new() -> Self {
+        let (tx, rx) = async_channel::unbounded();
+        Self { tx, rx }
+    }
+
+    pub async fn next(&self) -> PermittedTqResp {
+        self.rx.recv().await.expect("Send half cannot be dropped")
+    }
+}
+
+pub(crate) struct WorkerActivityTasks {
+    /// Centralizes management of heartbeat issuing / throttling
+    heartbeat_manager: ActivityHeartbeatManager,
+    /// Activities that have been issued to lang but not yet completed
+    outstanding_activity_tasks: DashMap<TaskToken, RemoteInFlightActInfo>,
+    /// Buffers activity task polling in the event we need to return a cancellation while a poll is
+    /// ongoing.
+    poller: BoxedActPoller,
+    /// Holds activity tasks we have received by non-polling means. EX: In direct response to
+    /// workflow task completion.
+    non_poll_tasks: NonPollActBuffer,
+    /// Ensures we stay at or below this worker's maximum concurrent activity limit
+    activities_semaphore: Arc<MeteredSemaphore>,
+    /// Enables per-worker rate-limiting of activity tasks
+    ratelimiter: Option<RateLimiter<NotKeyed, InMemoryState, DefaultClock, NoOpMiddleware>>,
+    /// Wakes every time an activity is removed from the outstanding map
+    complete_notify: Notify,
+
+    metrics: MetricsContext,
+
+    max_heartbeat_throttle_interval: Duration,
+    default_heartbeat_throttle_interval: Duration,
+}
+
+impl WorkerActivityTasks {
+    pub(crate) fn new(
+        max_activity_tasks: usize,
+        max_worker_act_per_sec: Option<f64>,
+        poller: BoxedActPoller,
+        client: Arc<dyn WorkerClient>,
+        metrics: MetricsContext,
+        max_heartbeat_throttle_interval: Duration,
+        default_heartbeat_throttle_interval: Duration,
+    ) -> Self {
+        Self {
+            heartbeat_manager: ActivityHeartbeatManager::new(client),
+            outstanding_activity_tasks: Default::default(),
+            poller,
+            non_poll_tasks: NonPollActBuffer::new(),
+            activities_semaphore: Arc::new(MeteredSemaphore::new(
+                max_activity_tasks,
+                metrics.with_new_attrs([activity_worker_type()]),
+                MetricsContext::available_task_slots,
+            )),
+            ratelimiter: max_worker_act_per_sec.and_then(|ps| {
+                Quota::with_period(Duration::from_secs_f64(ps.recip())).map(RateLimiter::direct)
+            }),
+            complete_notify: Notify::new(),
+            metrics,
+            max_heartbeat_throttle_interval,
+            default_heartbeat_throttle_interval,
+        }
+    }
+
+    pub(crate) fn notify_shutdown(&self) {
+        self.poller.notify_shutdown();
+    }
+
+    /// Wait for all outstanding activity tasks to finish
+    pub(crate) async fn wait_all_finished(&self) {
+        while !self.outstanding_activity_tasks.is_empty() {
+            self.complete_notify.notified().await
+        }
+    }
+
+    pub(crate) async fn shutdown(self) {
+        self.poller.shutdown_box().await;
+        self.heartbeat_manager.shutdown().await;
+    }
+
+    /// Wait until not at the outstanding activity limit, and then poll for an activity task.
+    ///
+    /// Returns `Ok(None)` if no activity is ready and the overall polling loop should be retried.
+    pub(crate) async fn poll(&self) -> Result<Option<ActivityTask>, PollActivityError> {
+        let poll_with_semaphore = async {
+            // Acquire and subsequently forget a permit for an outstanding activity. When they are
+            // completed, we must add a new permit to the semaphore, since holding the permit the
+            // entire time lang does work would be a challenge.
+            let perm = self
+                .activities_semaphore
+                .acquire_owned()
+                .await
+                .expect("outstanding activity semaphore not closed");
+            if let Some(ref rl) = self.ratelimiter {
+                rl.until_ready().await;
+            }
+            (self.poller.poll().await, perm)
+        };
+
+        tokio::select! {
+            biased;
+
+            cancel_task = self.next_pending_cancel_task() => {
+                cancel_task
+            }
+            task = self.non_poll_tasks.next() => {
+                Ok(Some(self.about_to_issue_task(task)))
+            }
+            (work, permit) = poll_with_semaphore => {
+                match work {
+                    Some(Ok(work)) => {
+                        if work == PollActivityTaskQueueResponse::default() {
+                            // Timeout
+                            self.metrics.act_poll_timeout();
+                            return Ok(None)
+                        }
+                        let work = self.about_to_issue_task(PermittedTqResp {
+                           resp: work, permit
+                        });
+                        Ok(Some(work))
+                    }
+                    None => {
+                        Err(PollActivityError::ShutDown)
+                    }
+                    Some(Err(e)) => Err(e.into())
+                }
+            }
+        }
+    }
+
+    pub(crate) async fn complete(
+        &self,
+        task_token: TaskToken,
+        status: aer::Status,
+        client: &dyn WorkerClient,
+    ) {
+        if let Some((_, act_info)) = self.outstanding_activity_tasks.remove(&task_token) {
+            let act_metrics = self.metrics.with_new_attrs([
+                activity_type(act_info.base.activity_type.clone()),
+                workflow_type(act_info.base.workflow_type.clone()),
+            ]);
+            act_metrics.act_execution_latency(act_info.base.start_time.elapsed());
+            let known_not_found = act_info.known_not_found;
+            drop(act_info); // TODO: Get rid of dashmap. If we hold ref across await, bad stuff.
+            self.heartbeat_manager.evict(task_token.clone()).await;
+            self.complete_notify.notify_waiters();
+
+            // No need to report activities which we already know the server doesn't care about
+            if !known_not_found {
+                let maybe_net_err = match status {
+                    aer::Status::WillCompleteAsync(_) => None,
+                    aer::Status::Completed(ar::Success { result }) => client
+                        .complete_activity_task(task_token.clone(), result.map(Into::into))
+                        .await
+                        .err(),
+                    aer::Status::Failed(ar::Failure { failure }) => {
+                        act_metrics.act_execution_failed();
+                        client
+                            .fail_activity_task(task_token.clone(), failure.map(Into::into))
+                            .await
+                            .err()
+                    }
+                    aer::Status::Cancelled(ar::Cancellation { failure }) => {
+                        let details = if let Some(Failure {
+                            failure_info:
+                                Some(FailureInfo::CanceledFailureInfo(CanceledFailureInfo { details })),
+                            ..
+                        }) = failure
+                        {
+                            details
+                        } else {
+                            warn!(task_token = ? task_token,
+                                "Expected activity cancelled status with CanceledFailureInfo");
+                            None
+                        };
+                        client
+                            .cancel_activity_task(task_token.clone(), details.map(Into::into))
+                            .await
+                            .err()
+                    }
+                };
+
+                if let Some(e) = maybe_net_err {
+                    if e.code() == tonic::Code::NotFound {
+                        warn!(task_token = ?task_token, details = ?e, "Activity not found on \
+                        completion. This may happen if the activity has already been cancelled but \
+                        completed anyway.");
+                    } else {
+                        warn!(error=?e, "Network error while completing activity");
+                    };
+                };
+            };
+        } else {
+            warn!(
+                "Attempted to complete activity task {} but we were not tracking it",
+                &task_token
+            );
+        }
+    }
+
+    /// Attempt to record an activity heartbeat
+    pub(crate) fn record_heartbeat(
+        &self,
+        details: ActivityHeartbeat,
+    ) -> Result<(), ActivityHeartbeatError> {
+        // TODO: Propagate these back as cancels. Silent fails is too nonobvious
+        let heartbeat_timeout: Duration = self
+            .outstanding_activity_tasks
+            .get(&TaskToken(details.task_token.clone()))
+            .ok_or(ActivityHeartbeatError::UnknownActivity)?
+            .heartbeat_timeout
+            .clone()
+            // We treat None as 0 (even though heartbeat_timeout is never set to None by the server)
+            .unwrap_or_default()
+            .try_into()
+            // This technically should never happen since prost duration should be directly mappable
+            // to std::time::Duration.
+            .or(Err(ActivityHeartbeatError::InvalidHeartbeatTimeout))?;
+
+        // There is a bug in the server that translates non-set heartbeat timeouts into 0 duration.
+        // That's why we treat 0 the same way as None, otherwise we wouldn't know which aggregation
+        // delay to use, and using 0 is not a good idea as SDK would hammer the server too hard.
+        let throttle_interval = if heartbeat_timeout.as_millis() == 0 {
+            self.default_heartbeat_throttle_interval
+        } else {
+            heartbeat_timeout.mul_f64(0.8)
+        };
+        let throttle_interval =
+            std::cmp::min(throttle_interval, self.max_heartbeat_throttle_interval);
+        self.heartbeat_manager.record(details, throttle_interval)
+    }
+
+    /// Returns a handle that the workflows management side can use to interact with this manager
+    pub(crate) fn get_handle_for_workflows(&self) -> ActivitiesFromWFTsHandle {
+        ActivitiesFromWFTsHandle {
+            sem: self.activities_semaphore.clone(),
+            tx: self.non_poll_tasks.tx.clone(),
+        }
+    }
+
+    async fn next_pending_cancel_task(&self) -> Result<Option<ActivityTask>, PollActivityError> {
+        let next_pc = self.heartbeat_manager.next_pending_cancel().await;
+        // Issue cancellations for anything we noticed was cancelled during heartbeating
+        if let Some(PendingActivityCancel { task_token, reason }) = next_pc {
+            // It's possible that activity has been completed and we no longer have an
+            // outstanding activity task. This is fine because it means that we no
+            // longer need to cancel this activity, so we'll just ignore such orphaned
+            // cancellations.
+            if let Some(mut details) = self.outstanding_activity_tasks.get_mut(&task_token) {
+                if details.issued_cancel_to_lang {
+                    // Don't double-issue cancellations
+                    return Ok(None);
+                }
+
+                details.issued_cancel_to_lang = true;
+                if reason == ActivityCancelReason::NotFound {
+                    details.known_not_found = true;
+                }
+                Ok(Some(ActivityTask::cancel_from_ids(task_token.0, reason)))
+            } else {
+                debug!(task_token = ?task_token, "Unknown activity task when issuing cancel");
+                // If we can't find the activity here, it's already been completed,
+                // in which case issuing a cancel again is pointless.
+                Ok(None)
+            }
+        } else {
+            // The only situation where the next cancel would return none is if the manager
+            // was dropped, which can only happen on shutdown.
+            Err(PollActivityError::ShutDown)
+        }
+    }
+
+    /// Called when there is a new act task about to be bubbled up out of the manager
+    fn about_to_issue_task(&self, task: PermittedTqResp) -> ActivityTask {
+        if let Some(dur) = task.resp.sched_to_start() {
+            self.metrics.act_sched_to_start_latency(dur);
+        };
+
+        self.outstanding_activity_tasks.insert(
+            task.resp.task_token.clone().into(),
+            RemoteInFlightActInfo::new(
+                task.resp.activity_type.clone().unwrap_or_default().name,
+                task.resp.workflow_type.clone().unwrap_or_default().name,
+                task.resp.heartbeat_timeout.clone(),
+                task.permit,
+            ),
+        );
+
+        ActivityTask::start_from_poll_resp(task.resp)
+    }
+
+    #[cfg(test)]
+    pub(crate) fn remaining_activity_capacity(&self) -> usize {
+        self.activities_semaphore.available_permits()
+    }
+}
+
+/// Provides facilities for the workflow side of things to interact with the activity manager.
+/// Allows for the handling of activities returned by WFT completions.
+pub(crate) struct ActivitiesFromWFTsHandle {
+    sem: Arc<MeteredSemaphore>,
+    tx: async_channel::Sender<PermittedTqResp>,
+}
+
+impl ActivitiesFromWFTsHandle {
+    /// Returns a handle that can be used to reserve an activity slot. EX: When requesting eager
+    /// dispatch of an activity to this worker upon workflow task completion
+    pub(crate) fn reserve_slot(&self) -> Option<OwnedMeteredSemPermit> {
+        self.sem.try_acquire_owned().ok()
+    }
+
+    /// Queue new activity tasks for dispatch received from non-polling sources (ex: eager returns
+    /// from WFT completion)
+    pub(crate) fn add_tasks(&self, tasks: impl IntoIterator<Item = PermittedTqResp>) {
+        for t in tasks.into_iter() {
+            self.tx.try_send(t).expect("Receive half cannot be dropped");
+        }
+    }
+}
+
+pub(crate) struct PermittedTqResp {
+    pub permit: OwnedMeteredSemPermit,
+    pub resp: PollActivityTaskQueueResponse,
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::{
+        test_help::mock_poller_from_resps, worker::client::mocks::mock_manual_workflow_client,
+    };
+
+    #[tokio::test]
+    async fn per_worker_ratelimit() {
+        let poller = mock_poller_from_resps([
+            PollActivityTaskQueueResponse {
+                task_token: vec![1],
+                activity_id: "act1".to_string(),
+                ..Default::default()
+            }
+            .into(),
+            PollActivityTaskQueueResponse {
+                task_token: vec![2],
+                activity_id: "act2".to_string(),
+                ..Default::default()
+            }
+            .into(),
+        ]);
+        let atm = WorkerActivityTasks::new(
+            10,
+            Some(2.0),
+            poller,
+            Arc::new(mock_manual_workflow_client()),
+            MetricsContext::default(),
+            Duration::from_secs(1),
+            Duration::from_secs(1),
+        );
+        let start = Instant::now();
+        atm.poll().await.unwrap().unwrap();
+        atm.poll().await.unwrap().unwrap();
+        // At least half a second will have elapsed since we only allow 2 tasks per second.
+        // With no ratelimit, even on a slow CI server with lots of load, this would typically take
+        // low single digit ms or less.
+        assert!(start.elapsed() > Duration::from_secs_f64(0.5));
+    }
+}

data/bridge/sdk-core/core/src/worker/client/mocks.rs

@@ -0,0 +1,87 @@
+use super::*;
+use futures::Future;
+
+#[cfg(test)]
+/// Create a mock client primed with basic necessary expectations
+pub(crate) fn mock_workflow_client() -> MockWorkerClient {
+    MockWorkerClient::new()
+}
+
+/// Create a mock manual client primed with basic necessary expectations
+pub(crate) fn mock_manual_workflow_client() -> MockManualWorkerClient {
+    MockManualWorkerClient::new()
+}
+
+// Need a version of the mock that can return futures so we can return potentially pending
+// results. This is really annoying b/c of the async trait stuff. Need
+// https://github.com/asomers/mockall/issues/189 to be fixed for it to go away.
+mockall::mock! {
+    pub ManualWorkerClient {}
+    #[allow(unused)]
+    impl WorkerClient for ManualWorkerClient {
+        fn poll_workflow_task<'a, 'b>(&'a self, task_queue: String, is_sticky: bool)
+            -> impl Future<Output = Result<PollWorkflowTaskQueueResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn poll_activity_task<'a, 'b>(&self, task_queue: String, max_tasks_per_sec: Option<f64>)
+            -> impl Future<Output = Result<PollActivityTaskQueueResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn complete_workflow_task<'a, 'b>(
+            &self,
+            request: WorkflowTaskCompletion,
+        ) -> impl Future<Output = Result<RespondWorkflowTaskCompletedResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn complete_activity_task<'a, 'b>(
+            &self,
+            task_token: TaskToken,
+            result: Option<Payloads>,
+        ) -> impl Future<Output = Result<RespondActivityTaskCompletedResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn cancel_activity_task<'a, 'b>(
+            &self,
+            task_token: TaskToken,
+            details: Option<Payloads>,
+        ) -> impl Future<Output = Result<RespondActivityTaskCanceledResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn fail_activity_task<'a, 'b>(
+            &self,
+            task_token: TaskToken,
+            failure: Option<Failure>,
+        ) -> impl Future<Output = Result<RespondActivityTaskFailedResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn fail_workflow_task<'a, 'b>(
+            &self,
+            task_token: TaskToken,
+            cause: WorkflowTaskFailedCause,
+            failure: Option<Failure>,
+        ) -> impl Future<Output = Result<RespondWorkflowTaskFailedResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn record_activity_heartbeat<'a, 'b>(
+           &self,
+           task_token: TaskToken,
+           details: Option<Payloads>,
+        ) -> impl Future<Output = Result<RecordActivityTaskHeartbeatResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn get_workflow_execution_history<'a, 'b>(
+            &self,
+            workflow_id: String,
+            run_id: Option<String>,
+            page_token: Vec<u8>
+        ) -> impl Future<Output = Result<GetWorkflowExecutionHistoryResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+
+        fn respond_legacy_query<'a, 'b>(
+            &self,
+            task_token: TaskToken,
+            query_result: QueryResult,
+        ) -> impl Future<Output = Result<RespondQueryTaskCompletedResponse>> + Send + 'b
+            where 'a: 'b, Self: 'b;
+    }
+}