/build/source/nativelink-scheduler/src/api_worker_scheduler.rs

Source
// Copyright 2024 The NativeLink Authors. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//    http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

use core::ops::{Deref, DerefMut};
use std::sync::Arc;

use async_lock::Mutex;
use lru::LruCache;
use nativelink_config::schedulers::WorkerAllocationStrategy;
use nativelink_error::{Code, Error, ResultExt, error_if, make_err, make_input_err};
use nativelink_metric::{
    MetricFieldData, MetricKind, MetricPublishKnownKindData, MetricsComponent,
    RootMetricsComponent, group,
};
use nativelink_util::action_messages::{OperationId, WorkerId};
use nativelink_util::operation_state_manager::{UpdateOperationType, WorkerStateManager};
use nativelink_util::platform_properties::PlatformProperties;
use nativelink_util::spawn;
use nativelink_util::task::JoinHandleDropGuard;
use tokio::sync::Notify;
use tokio::sync::mpsc::{self, UnboundedSender};
use tonic::async_trait;
use tracing::{error, warn};

use crate::platform_property_manager::PlatformPropertyManager;
use crate::worker::{ActionInfoWithProps, Worker, WorkerTimestamp, WorkerUpdate};
use crate::worker_scheduler::WorkerScheduler;

#[derive(Debug)]
struct Workers(LruCache<WorkerId, Worker>);

impl Deref for Workers {
    type Target = LruCache<WorkerId, Worker>;

    fn deref(&self) -> &Self::Target {
        &self.0
    }
}

impl DerefMut for Workers {
    fn deref_mut(&mut self) -> &mut Self::Target {
        &mut self.0
    }
}

// Note: This could not be a derive macro because this derive-macro
// does not support LruCache and nameless field structs.
impl MetricsComponent for Workers {
    fn publish(
        &self,
        _kind: MetricKind,
        _field_metadata: MetricFieldData,
    ) -> Result<MetricPublishKnownKindData, nativelink_metric::Error> {
        let _enter = group!("workers").entered();
        for (worker_id, worker) in self.iter() {
            let _enter = group!(worker_id).entered();
            worker.publish(MetricKind::Component, MetricFieldData::default())?;
        }
        Ok(MetricPublishKnownKindData::Component)
    }
}

/// A collection of workers that are available to run tasks.
#[derive(MetricsComponent)]
struct ApiWorkerSchedulerImpl {
    /// A `LruCache` of workers available based on `allocation_strategy`.
    #[metric(group = "workers")]
    workers: Workers,

    /// The worker state manager.
    #[metric(group = "worker_state_manager")]
    worker_state_manager: Arc<dyn WorkerStateManager>,
    /// The allocation strategy for workers.
    allocation_strategy: WorkerAllocationStrategy,
    /// A channel to notify the matching engine that the worker pool has changed.
    worker_change_notify: Arc<Notify>,
    /// A channel to notify that an operation is still alive.
    operation_keep_alive_tx: UnboundedSender<(OperationId, WorkerId)>,
}

impl core::fmt::Debug for ApiWorkerSchedulerImpl {
    fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
        f.debug_struct("ApiWorkerSchedulerImpl")
            .field("workers", &self.workers)
            .field("allocation_strategy", &self.allocation_strategy)
            .field("worker_change_notify", &self.worker_change_notify)
            .field("operation_keep_alive_tx", &self.operation_keep_alive_tx)
            .finish_non_exhaustive()
    }
}

impl ApiWorkerSchedulerImpl {
    /// Refreshes the lifetime of the worker with the given timestamp.
    fn refresh_lifetime(
        &mut self,
        worker_id: &WorkerId,
        timestamp: WorkerTimestamp,
    ) -> Result<(), Error> {
        let worker = self.workers.0.peek_mut(worker_id).ok_or_else(|| {0
            make_input_err!(
                "Worker not found in worker map in refresh_lifetime() {}",
                worker_id
            )
        })?;
        error_if!(
            worker.last_update_timestamp > timestamp,
            "Worker already had a timestamp of {}, but tried to update it with {}",
            worker.last_update_timestamp,
            timestamp
        );
        worker.last_update_timestamp = timestamp;
        for operation_id0 in worker.running_action_infos.keys() {
            if self
                .operation_keep_alive_tx
                .send((operation_id.clone(), worker_id.clone()))
                .is_err()
            {
                error!(
                    ?operation_id,
                    ?worker_id,
                    "OperationKeepAliveTx stream closed"
                );
            }
        }
        Ok(())
    }

    /// Adds a worker to the pool.
    /// Note: This function will not do any task matching.
    fn add_worker(&mut self, worker: Worker) -> Result<(), Error> {
        let worker_id = worker.id.clone();
        self.workers.put(worker_id.clone(), worker);

        // Worker is not cloneable, and we do not want to send the initial connection results until
        // we have added it to the map, or we might get some strange race conditions due to the way
        // the multi-threaded runtime works.
        let worker = self.workers.peek_mut(&worker_id).unwrap();
        let res = worker
            .send_initial_connection_result()
            .err_tip(|| "Failed to send initial connection result to worker");
        if let Err(err0) = &res {
            error!(
                ?worker_id,
                ?err,
                "Worker connection appears to have been closed while adding to pool"
            );
        }
        self.worker_change_notify.notify_one();
        res
    }

    /// Removes worker from pool.
    /// Note: The caller is responsible for any rescheduling of any tasks that might be
    /// running.
    fn remove_worker(&mut self, worker_id: &WorkerId) -> Option<Worker> {
        let result = self.workers.pop(worker_id);
        self.worker_change_notify.notify_one();
        result
    }

    /// Sets if the worker is draining or not.
    async fn set_drain_worker(
        &mut self,
        worker_id: &WorkerId,
        is_draining: bool,
    ) -> Result<(), Error> {
        let worker = self
            .workers
            .get_mut(worker_id)
            .err_tip(|| format!("Worker {worker_id} doesn't exist in the pool"0))?0;
        worker.is_draining = is_draining;
        self.worker_change_notify.notify_one();
        Ok(())
    }

    fn inner_find_worker_for_action(
        &self,
        platform_properties: &PlatformProperties,
    ) -> Option<WorkerId> {
        let mut workers_iter = self.workers.iter();
        let workers_iter = match self.allocation_strategy {
            // Use rfind to get the least recently used that satisfies the properties.
            WorkerAllocationStrategy::LeastRecentlyUsed => workers_iter.rfind(|(_, w)| {34
                w.can_accept_work() && platform_properties33.is_satisfied_by33(&w.platform_properties33)
            }),
            // Use find to get the most recently used that satisfies the properties.
            WorkerAllocationStrategy::MostRecentlyUsed => workers_iter.find(|(_, w)| {
                w.can_accept_work() && platform_properties.is_satisfied_by(&w.platform_properties)
            }),
        };
        workers_iter.map(|(_, w)| w.id28.clone28())
    }

    async fn update_action(
        &mut self,
        worker_id: &WorkerId,
        operation_id: &OperationId,
        update: UpdateOperationType,
    ) -> Result<(), Error> {
        let worker = self.workers.get_mut(worker_id).err_tip(|| {0
            format!("Worker {worker_id} does not exist in SimpleScheduler::update_action")
        })?;

        // Ensure the worker is supposed to be running the operation.
        if !worker.running_action_infos.contains_key(operation_id) {
            let err = make_err!(
                Code::Internal,
                "Operation {operation_id} should not be running on worker {worker_id} in SimpleScheduler::update_action"
            );
            return Result::<(), _>::Err(err.clone())
                .merge(self.immediate_evict_worker(worker_id, err).await);
        }

        let (is_finished, due_to_backpressure) = match &update {
            UpdateOperationType::UpdateWithActionStage(action_stage) => {
                (action_stage.is_finished(), false)
            }
            UpdateOperationType::KeepAlive => (false, false),
            UpdateOperationType::UpdateWithError(err) => {
                (true, err.code == Code::ResourceExhausted)
            }
        };

        // Update the operation in the worker state manager.
        {
            let update_operation_res = self
                .worker_state_manager
                .update_operation(operation_id, worker_id, update)
                .await
                .err_tip(|| "in update_operation on SimpleScheduler::update_action");
            if let Err(err0) = update_operation_res {
                error!(
                    ?operation_id,
                    ?worker_id,
                    ?err,
                    "Failed to update_operation on update_action"
                );
                return Err(err);
            }
        }

        if !is_finished {
            return Ok(());
        }

        // Clear this action from the current worker if finished.
        let complete_action_res = {
            let was_paused = !worker.can_accept_work();

            // Note: We need to run this before dealing with backpressure logic.
            let complete_action_res = worker.complete_action(operation_id).await;

            // Only pause if there's an action still waiting that will unpause.
            if (was_paused || due_to_backpressure) && worker0.has_actions0() {
                worker.is_paused = true;
            }
            complete_action_res
        };

        self.worker_change_notify.notify_one();

        complete_action_res
    }

    /// Notifies the specified worker to run the given action and handles errors by evicting
    /// the worker if the notification fails.
    async fn worker_notify_run_action(
        &mut self,
        worker_id: WorkerId,
        operation_id: OperationId,
        action_info: ActionInfoWithProps,
    ) -> Result<(), Error> {
        if let Some(worker) = self.workers.get_mut(&worker_id) {
            let notify_worker_result = worker
                .notify_update(WorkerUpdate::RunAction((operation_id, action_info.clone())))
                .await;

            if notify_worker_result.is_err() {
                warn!(
                    ?worker_id,
                    ?action_info,
                    ?notify_worker_result,
                    "Worker command failed, removing worker",
                );

                let err = make_err!(
                    Code::Internal,
                    "Worker command failed, removing worker {worker_id} -- {notify_worker_result:?}",
                );

                return Result::<(), _>::Err(err.clone())
                    .merge(self.immediate_evict_worker(&worker_id, err).await);
            }
        } else {
            warn!(
                ?worker_id,
                ?operation_id,
                ?action_info,
                "Worker not found in worker map in worker_notify_run_action"
            );
        }
        Ok(())
    }

    /// Evicts the worker from the pool and puts items back into the queue if anything was being executed on it.
    async fn immediate_evict_worker(
        &mut self,
        worker_id: &WorkerId,
        err: Error,
    ) -> Result<(), Error> {
        let mut result = Ok(());
        if let Some(mut worker) = self.remove_worker(worker_id) {
            // We don't care if we fail to send message to worker, this is only a best attempt.
            drop(worker.notify_update(WorkerUpdate::Disconnect).await);
            for (operation_id4, _) in worker.running_action_infos.drain() {
                result = result.merge(
                    self.worker_state_manager
                        .update_operation(
                            &operation_id,
                            worker_id,
                            UpdateOperationType::UpdateWithError(err.clone()),
                        )
                        .await,
                );
            }
        }
        // Note: Calling this many time is very cheap, it'll only trigger `do_try_match` once.
        // TODO(aaronmondal) This should be moved to inside the Workers struct.
        self.worker_change_notify.notify_one();
        result
    }
}

#[derive(Debug, MetricsComponent)]
pub struct ApiWorkerScheduler {
    #[metric]
    inner: Mutex<ApiWorkerSchedulerImpl>,
    #[metric(group = "platform_property_manager")]
    platform_property_manager: Arc<PlatformPropertyManager>,

    #[metric(
        help = "Timeout of how long to evict workers if no response in this given amount of time in seconds."
    )]
    worker_timeout_s: u64,
    _operation_keep_alive_spawn: JoinHandleDropGuard<()>,
}

impl ApiWorkerScheduler {
    pub fn new(
        worker_state_manager: Arc<dyn WorkerStateManager>,
        platform_property_manager: Arc<PlatformPropertyManager>,
        allocation_strategy: WorkerAllocationStrategy,
        worker_change_notify: Arc<Notify>,
        worker_timeout_s: u64,
    ) -> Arc<Self> {
        let (operation_keep_alive_tx, mut operation_keep_alive_rx) = mpsc::unbounded_channel();
        Arc::new(Self {
            inner: Mutex::new(ApiWorkerSchedulerImpl {
                workers: Workers(LruCache::unbounded()),
                worker_state_manager: worker_state_manager.clone(),
                allocation_strategy,
                worker_change_notify,
                operation_keep_alive_tx,
            }),
            platform_property_manager,
            worker_timeout_s,
            _operation_keep_alive_spawn: spawn!(
                "simple_scheduler_operation_keep_alive",
                async move {
                    const RECV_MANY_LIMIT: usize = 256;
                    let mut messages = Vec::with_capacity(RECV_MANY_LIMIT);
                    loop {
                        messages.clear();
                        operation_keep_alive_rx
                            .recv_many(&mut messages, RECV_MANY_LIMIT)
                            .await;
                        if messages.is_empty() {
                            return; // Looks like our sender has been dropped.
                        }
                        for (operation_id, worker_id) in messages.drain(..) {
                            let update_operation_res = worker_state_manager
                                .update_operation(
                                    &operation_id,
                                    &worker_id,
                                    UpdateOperationType::KeepAlive,
                                )
                                .await;
                            if let Err(err) = update_operation_res {
                                warn!(
                                    ?err,
                                    "Error while running worker_keep_alive_received, maybe job is done?"
                                );
                            }
                        }
                    }
                }
            ),
        })
    }

    pub async fn worker_notify_run_action(
        &self,
        worker_id: WorkerId,
        operation_id: OperationId,
        action_info: ActionInfoWithProps,
    ) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        inner
            .worker_notify_run_action(worker_id, operation_id, action_info)
            .await
    }

    /// Attempts to find a worker that is capable of running this action.
    // TODO(aaronmondal) This algorithm is not very efficient. Simple testing using a tree-like
    // structure showed worse performance on a 10_000 worker * 7 properties * 1000 queued tasks
    // simulation of worst cases in a single threaded environment.
    pub async fn find_worker_for_action(
        &self,
        platform_properties: &PlatformProperties,
    ) -> Option<WorkerId> {
        let inner = self.inner.lock().await;
        inner.inner_find_worker_for_action(platform_properties)
    }

    /// Checks to see if the worker exists in the worker pool. Should only be used in unit tests.
    #[must_use]
    pub async fn contains_worker_for_test(&self, worker_id: &WorkerId) -> bool {
        let inner = self.inner.lock().await;
        inner.workers.contains(worker_id)
    }

    /// A unit test function used to send the keep alive message to the worker from the server.
    pub async fn send_keep_alive_to_worker_for_test(
        &self,
        worker_id: &WorkerId,
    ) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        let worker = inner.workers.get_mut(worker_id).ok_or_else(|| {0
            make_input_err!("WorkerId '{}' does not exist in workers map", worker_id)
        })?;
        worker.keep_alive()
    }
}

#[async_trait]
impl WorkerScheduler for ApiWorkerScheduler {
    fn get_platform_property_manager(&self) -> &PlatformPropertyManager {
        self.platform_property_manager.as_ref()
    }

    async fn add_worker(&self, worker: Worker) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        let worker_id = worker.id.clone();
        let result = inner
            .add_worker(worker)
            .err_tip(|| "Error while adding worker, removing from pool");
        if let Err(err0) = result {
            return Result::<(), _>::Err(err.clone())
                .merge(inner.immediate_evict_worker(&worker_id, err).await);
        }
        Ok(())
    }

    async fn update_action(
        &self,
        worker_id: &WorkerId,
        operation_id: &OperationId,
        update: UpdateOperationType,
    ) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        inner.update_action(worker_id, operation_id, update).await
    }

    async fn worker_keep_alive_received(
        &self,
        worker_id: &WorkerId,
        timestamp: WorkerTimestamp,
    ) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        inner
            .refresh_lifetime(worker_id, timestamp)
            .err_tip(|| "Error refreshing lifetime in worker_keep_alive_received()")
    }

    async fn remove_worker(&self, worker_id: &WorkerId) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        inner
            .immediate_evict_worker(
                worker_id,
                make_err!(Code::Internal, "Received request to remove worker"),
            )
            .await
    }

    async fn remove_timedout_workers(&self, now_timestamp: WorkerTimestamp) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;

        let mut result = Ok(());
        // Items should be sorted based on last_update_timestamp, so we don't need to iterate the entire
        // map most of the time.
        let worker_ids_to_remove: Vec<WorkerId> = inner
            .workers
            .iter()
            .rev()
            .map_while5(|(worker_id, worker)| {
                if worker.last_update_timestamp <= now_timestamp - self.worker_timeout_s {
                    Some(worker_id.clone())
                } else {
                    None
                }
            })
            .collect();
        for worker_id2 in &worker_ids_to_remove {
            warn!(?worker_id, "Worker timed out, removing from pool");
            result = result.merge(
                inner
                    .immediate_evict_worker(
                        worker_id,
                        make_err!(
                            Code::Internal,
                            "Worker {worker_id} timed out, removing from pool"
                        ),
                    )
                    .await,
            );
        }

        result
    }

    async fn set_drain_worker(&self, worker_id: &WorkerId, is_draining: bool) -> Result<(), Error> {
        let mut inner = self.inner.lock().await;
        inner.set_drain_worker(worker_id, is_draining).await
    }
}

impl RootMetricsComponent for ApiWorkerScheduler {}

Coverage Report

Created: 2025-05-30 16:37

Line	Count	Source
1		// Copyright 2024 The NativeLink Authors. All rights reserved.
2		//
3		// Licensed under the Apache License, Version 2.0 (the "License");
4		// you may not use this file except in compliance with the License.
5		// You may obtain a copy of the License at
6		//
7		// http://www.apache.org/licenses/LICENSE-2.0
8		//
9		// Unless required by applicable law or agreed to in writing, software
10		// distributed under the License is distributed on an "AS IS" BASIS,
11		// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12		// See the License for the specific language governing permissions and
13		// limitations under the License.
14
15		use core::ops::{Deref, DerefMut};
16		use std::sync::Arc;
17
18		use async_lock::Mutex;
19		use lru::LruCache;
20		use nativelink_config::schedulers::WorkerAllocationStrategy;
21		use nativelink_error::{Code, Error, ResultExt, error_if, make_err, make_input_err};
22		use nativelink_metric::{
23		MetricFieldData, MetricKind, MetricPublishKnownKindData, MetricsComponent,
24		RootMetricsComponent, group,
25		};
26		use nativelink_util::action_messages::{OperationId, WorkerId};
27		use nativelink_util::operation_state_manager::{UpdateOperationType, WorkerStateManager};
28		use nativelink_util::platform_properties::PlatformProperties;
29		use nativelink_util::spawn;
30		use nativelink_util::task::JoinHandleDropGuard;
31		use tokio::sync::Notify;
32		use tokio::sync::mpsc::{self, UnboundedSender};
33		use tonic::async_trait;
34		use tracing::{error, warn};
35
36		use crate::platform_property_manager::PlatformPropertyManager;
37		use crate::worker::{ActionInfoWithProps, Worker, WorkerTimestamp, WorkerUpdate};
38		use crate::worker_scheduler::WorkerScheduler;
39
40		#[derive(Debug)]
41		struct Workers(LruCache<WorkerId, Worker>);
42
43		impl Deref for Workers {
44		type Target = LruCache<WorkerId, Worker>;
45
46	57	fn deref(&self) -> &Self::Target {
47	57	&self.0
48	57	}
49		}
50
51		impl DerefMut for Workers {
52	109	fn deref_mut(&mut self) -> &mut Self::Target {
53	109	&mut self.0
54	109	}
55		}
56
57		// Note: This could not be a derive macro because this derive-macro
58		// does not support LruCache and nameless field structs.
59		impl MetricsComponent for Workers {
60	0	fn publish(
61	0	&self,
62	0	_kind: MetricKind,
63	0	_field_metadata: MetricFieldData,
64	0	) -> Result<MetricPublishKnownKindData, nativelink_metric::Error> {
65	0	let _enter = group!("workers").entered();
66	0	for (worker_id, worker) in self.iter() {
67	0	let _enter = group!(worker_id).entered();
68	0	worker.publish(MetricKind::Component, MetricFieldData::default())?;
69		}
70	0	Ok(MetricPublishKnownKindData::Component)
71	0	}
72		}
73
74		/// A collection of workers that are available to run tasks.
75		#[derive(MetricsComponent)]
76		struct ApiWorkerSchedulerImpl {
77		/// A `LruCache` of workers available based on `allocation_strategy`.
78		#[metric(group = "workers")]
79		workers: Workers,
80
81		/// The worker state manager.
82		#[metric(group = "worker_state_manager")]
83		worker_state_manager: Arc<dyn WorkerStateManager>,
84		/// The allocation strategy for workers.
85		allocation_strategy: WorkerAllocationStrategy,
86		/// A channel to notify the matching engine that the worker pool has changed.
87		worker_change_notify: Arc<Notify>,
88		/// A channel to notify that an operation is still alive.
89		operation_keep_alive_tx: UnboundedSender<(OperationId, WorkerId)>,
90		}
91
92		impl core::fmt::Debug for ApiWorkerSchedulerImpl {
93	0	fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
94	0	f.debug_struct("ApiWorkerSchedulerImpl")
95	0	.field("workers", &self.workers)
96	0	.field("allocation_strategy", &self.allocation_strategy)
97	0	.field("worker_change_notify", &self.worker_change_notify)
98	0	.field("operation_keep_alive_tx", &self.operation_keep_alive_tx)
99	0	.finish_non_exhaustive()
100	0	}
101		}
102
103		impl ApiWorkerSchedulerImpl {
104		/// Refreshes the lifetime of the worker with the given timestamp.
105	2	fn refresh_lifetime(
106	2	&mut self,
107	2	worker_id: &WorkerId,
108	2	timestamp: WorkerTimestamp,
109	2	) -> Result<(), Error> {
110	2	let worker = self.workers.0.peek_mut(worker_id).ok_or_else(\|\| {0
111	0	make_input_err!(
112		"Worker not found in worker map in refresh_lifetime() {}",
113		worker_id
114		)
115	0	})?;
116	0	error_if!(
117	2	worker.last_update_timestamp > timestamp, Branch (117:13): [True: 0, False: 2] Branch (117:13): [Folded - Ignored]
118		"Worker already had a timestamp of {}, but tried to update it with {}",
119		worker.last_update_timestamp,
120		timestamp
121		);
122	2	worker.last_update_timestamp = timestamp;
123	2	for operation_id0 in worker.running_action_infos.keys() {
124	0	if self Branch (124:16): [True: 0, False: 0] Branch (124:16): [Folded - Ignored]
125	0	.operation_keep_alive_tx
126	0	.send((operation_id.clone(), worker_id.clone()))
127	0	.is_err()
128		{
129	0	error!(
130		?operation_id,
131		?worker_id,
132	0	"OperationKeepAliveTx stream closed"
133		);
134	0	}
135		}
136	2	Ok(())
137	2	}
138
139		/// Adds a worker to the pool.
140		/// Note: This function will not do any task matching.
141	31	fn add_worker(&mut self, worker: Worker) -> Result<(), Error> {
142	31	let worker_id = worker.id.clone();
143	31	self.workers.put(worker_id.clone(), worker);
144
145		// Worker is not cloneable, and we do not want to send the initial connection results until
146		// we have added it to the map, or we might get some strange race conditions due to the way
147		// the multi-threaded runtime works.
148	31	let worker = self.workers.peek_mut(&worker_id).unwrap();
149	31	let res = worker
150	31	.send_initial_connection_result()
151	31	.err_tip(\|\| "Failed to send initial connection result to worker");
152	31	if let Err( err0 ) = &res { Branch (152:16): [True: 0, False: 31] Branch (152:16): [Folded - Ignored]
153	0	error!(
154		?worker_id,
155		?err,
156	0	"Worker connection appears to have been closed while adding to pool"
157		);
158	31	}
159	31	self.worker_change_notify.notify_one();
160	31	res
161	31	}
162
163		/// Removes worker from pool.
164		/// Note: The caller is responsible for any rescheduling of any tasks that might be
165		/// running.
166	6	fn remove_worker(&mut self, worker_id: &WorkerId) -> Option<Worker> {
167	6	let result = self.workers.pop(worker_id);
168	6	self.worker_change_notify.notify_one();
169	6	result
170	6	}
171
172		/// Sets if the worker is draining or not.
173	2	async fn set_drain_worker(
174	2	&mut self,
175	2	worker_id: &WorkerId,
176	2	is_draining: bool,
177	2	) -> Result<(), Error> {
178	2	let worker = self
179	2	.workers
180	2	.get_mut(worker_id)
181	2	.err_tip(\|\| format!( "Worker {worker_id} doesn't exist in the pool"0 )) ?0 ;
182	2	worker.is_draining = is_draining;
183	2	self.worker_change_notify.notify_one();
184	2	Ok(())
185	2	}
186
187	45	fn inner_find_worker_for_action(
188	45	&self,
189	45	platform_properties: &PlatformProperties,
190	45	) -> Option<WorkerId> {
191	45	let mut workers_iter = self.workers.iter();
192	45	let workers_iter = match self.allocation_strategy {
193		// Use rfind to get the least recently used that satisfies the properties.
194	45	WorkerAllocationStrategy::LeastRecentlyUsed => workers_iter.rfind(\|(_, w)\| {34
195	34	w.can_accept_work() && platform_properties33 . is_satisfied_by33 ( &w.platform_properties33 ) Branch (195:17): [True: 33, False: 1] Branch (195:17): [Folded - Ignored]
196	34	}),
197		// Use find to get the most recently used that satisfies the properties.
198	0	WorkerAllocationStrategy::MostRecentlyUsed => workers_iter.find(\|(_, w)\| {
199	0	w.can_accept_work() && platform_properties.is_satisfied_by(&w.platform_properties) Branch (199:17): [True: 0, False: 0] Branch (199:17): [Folded - Ignored]
200	0	}),
201		};
202	45	workers_iter.map(\|(_, w)\| w.id28 . clone28 ())
203	45	}
204
205	10	async fn update_action(
206	10	&mut self,
207	10	worker_id: &WorkerId,
208	10	operation_id: &OperationId,
209	10	update: UpdateOperationType,
210	10	) -> Result<(), Error> {
211	10	let worker = self.workers.get_mut(worker_id).err_tip(\|\| {0
212	0	format!("Worker {worker_id} does not exist in SimpleScheduler::update_action")
213	0	})?;
214
215		// Ensure the worker is supposed to be running the operation.
216	10	if !worker.running_action_infos.contains_key(operation_id) { Branch (216:12): [True: 1, False: 9] Branch (216:12): [Folded - Ignored]
217	1	let err = make_err!(
218	1	Code::Internal,
219		"Operation {operation_id} should not be running on worker {worker_id} in SimpleScheduler::update_action"
220		);
221	1	return Result::<(), _>::Err(err.clone())
222	1	.merge(self.immediate_evict_worker(worker_id, err).await);
223	9	}
224
225	9	let (is_finished, due_to_backpressure) = match &update {
226	6	UpdateOperationType::UpdateWithActionStage(action_stage) => {
227	6	(action_stage.is_finished(), false)
228		}
229	0	UpdateOperationType::KeepAlive => (false, false),
230	3	UpdateOperationType::UpdateWithError(err) => {
231	3	(true, err.code == Code::ResourceExhausted)
232		}
233		};
234
235		// Update the operation in the worker state manager.
236		{
237	9	let update_operation_res = self
238	9	.worker_state_manager
239	9	.update_operation(operation_id, worker_id, update)
240	9	.await
241	9	.err_tip(\|\| "in update_operation on SimpleScheduler::update_action");
242	9	if let Err( err0 ) = update_operation_res { Branch (242:20): [True: 0, False: 9] Branch (242:20): [Folded - Ignored]
243	0	error!(
244		?operation_id,
245		?worker_id,
246		?err,
247	0	"Failed to update_operation on update_action"
248		);
249	0	return Err(err);
250	9	}
251		}
252
253	9	if !is_finished { Branch (253:12): [True: 0, False: 9] Branch (253:12): [Folded - Ignored]
254	0	return Ok(());
255	9	}
256
257		// Clear this action from the current worker if finished.
258	9	let complete_action_res = {
259	9	let was_paused = !worker.can_accept_work();
260
261		// Note: We need to run this before dealing with backpressure logic.
262	9	let complete_action_res = worker.complete_action(operation_id).await;
263
264		// Only pause if there's an action still waiting that will unpause.
265	9	if (was_paused \|\| due_to_backpressure) && worker0 . has_actions0 () { Branch (265:17): [True: 0, False: 9] Branch (265:31): [True: 0, False: 9] Branch (265:55): [True: 0, False: 0] Branch (265:17): [Folded - Ignored] Branch (265:31): [Folded - Ignored] Branch (265:55): [Folded - Ignored]
266	0	worker.is_paused = true;
267	9	}
268	9	complete_action_res
269		};
270
271	9	self.worker_change_notify.notify_one();
272
273	9	complete_action_res
274	10	}
275
276		/// Notifies the specified worker to run the given action and handles errors by evicting
277		/// the worker if the notification fails.
278	28	async fn worker_notify_run_action(
279	28	&mut self,
280	28	worker_id: WorkerId,
281	28	operation_id: OperationId,
282	28	action_info: ActionInfoWithProps,
283	28	) -> Result<(), Error> {
284	28	if let Some(worker) = self.workers.get_mut(&worker_id) { Branch (284:16): [True: 0, False: 0] Branch (284:16): [Folded - Ignored] Branch (284:16): [True: 27, False: 0] Branch (284:16): [True: 1, False: 0]
285	28	let notify_worker_result = worker
286	28	.notify_update(WorkerUpdate::RunAction((operation_id, action_info.clone())))
287	28	.await;
288
289	28	if notify_worker_result.is_err() { Branch (289:16): [True: 0, False: 0] Branch (289:16): [Folded - Ignored] Branch (289:16): [True: 1, False: 26] Branch (289:16): [True: 0, False: 1]
290	1	warn!(
291		?worker_id,
292		?action_info,
293		?notify_worker_result,
294	1	"Worker command failed, removing worker",
295		);
296
297	1	let err = make_err!(
298	1	Code::Internal,
299		"Worker command failed, removing worker {worker_id} -- {notify_worker_result:?}",
300		);
301
302	1	return Result::<(), _>::Err(err.clone())
303	1	.merge(self.immediate_evict_worker(&worker_id, err).await);
304	27	}
305		} else {
306	0	warn!(
307		?worker_id,
308		?operation_id,
309		?action_info,
310	0	"Worker not found in worker map in worker_notify_run_action"
311		);
312		}
313	27	Ok(())
314	28	}
315
316		/// Evicts the worker from the pool and puts items back into the queue if anything was being executed on it.
317	6	async fn immediate_evict_worker(
318	6	&mut self,
319	6	worker_id: &WorkerId,
320	6	err: Error,
321	6	) -> Result<(), Error> {
322	6	let mut result = Ok(());
323	6	if let Some(mut worker) = self.remove_worker(worker_id) { Branch (323:16): [True: 5, False: 0] Branch (323:16): [Folded - Ignored] Branch (323:16): [True: 1, False: 0] Branch (323:16): [True: 0, False: 0]
324		// We don't care if we fail to send message to worker, this is only a best attempt.
325	6	drop(worker.notify_update(WorkerUpdate::Disconnect).await);
326	6	for ( operation_id4 , _) in worker.running_action_infos.drain() {
327	4	result = result.merge(
328	4	self.worker_state_manager
329	4	.update_operation(
330	4	&operation_id,
331	4	worker_id,
332	4	UpdateOperationType::UpdateWithError(err.clone()),
333	4	)
334	4	.await,
335		);
336		}
337	0	}
338		// Note: Calling this many time is very cheap, it'll only trigger `do_try_match` once.
339		// TODO(aaronmondal) This should be moved to inside the Workers struct.
340	6	self.worker_change_notify.notify_one();
341	6	result
342	6	}
343		}
344
345		#[derive(Debug, MetricsComponent)]
346		pub struct ApiWorkerScheduler {
347		#[metric]
348		inner: Mutex<ApiWorkerSchedulerImpl>,
349		#[metric(group = "platform_property_manager")]
350		platform_property_manager: Arc<PlatformPropertyManager>,
351
352		#[metric(
353		help = "Timeout of how long to evict workers if no response in this given amount of time in seconds."
354		)]
355		worker_timeout_s: u64,
356		_operation_keep_alive_spawn: JoinHandleDropGuard<()>,
357		}
358
359		impl ApiWorkerScheduler {
360	28	pub fn new(
361	28	worker_state_manager: Arc<dyn WorkerStateManager>,
362	28	platform_property_manager: Arc<PlatformPropertyManager>,
363	28	allocation_strategy: WorkerAllocationStrategy,
364	28	worker_change_notify: Arc<Notify>,
365	28	worker_timeout_s: u64,
366	28	) -> Arc<Self> {
367	28	let (operation_keep_alive_tx, mut operation_keep_alive_rx) = mpsc::unbounded_channel();
368	28	Arc::new(Self {
369	28	inner: Mutex::new(ApiWorkerSchedulerImpl {
370	28	workers: Workers(LruCache::unbounded()),
371	28	worker_state_manager: worker_state_manager.clone(),
372	28	allocation_strategy,
373	28	worker_change_notify,
374	28	operation_keep_alive_tx,
375	28	}),
376	28	platform_property_manager,
377	28	worker_timeout_s,
378	28	_operation_keep_alive_spawn: spawn!(
379		"simple_scheduler_operation_keep_alive",
380	23	async move {
381		const RECV_MANY_LIMIT: usize = 256;
382	23	let mut messages = Vec::with_capacity(RECV_MANY_LIMIT);
383		loop {
384	23	messages.clear();
385	23	operation_keep_alive_rx
386	23	.recv_many(&mut messages, RECV_MANY_LIMIT)
387	23	.await;
388	0	if messages.is_empty() { Branch (388:28): [True: 0, False: 0] Branch (388:28): [Folded - Ignored]
389	0	return; // Looks like our sender has been dropped.
390	0	}
391	0	for (operation_id, worker_id) in messages.drain(..) {
392	0	let update_operation_res = worker_state_manager
393	0	.update_operation(
394	0	&operation_id,
395	0	&worker_id,
396	0	UpdateOperationType::KeepAlive,
397	0	)
398	0	.await;
399	0	if let Err(err) = update_operation_res { Branch (399:36): [True: 0, False: 0] Branch (399:36): [Folded - Ignored]
400	0	warn!(
401		?err,
402	0	"Error while running worker_keep_alive_received, maybe job is done?"
403		);
404	0	}
405		}
406		}
407	0	}
408		),
409		})
410	28	}
411
412	28	pub async fn worker_notify_run_action(
413	28	&self,
414	28	worker_id: WorkerId,
415	28	operation_id: OperationId,
416	28	action_info: ActionInfoWithProps,
417	28	) -> Result<(), Error> {
418	28	let mut inner = self.inner.lock().await;
419	28	inner
420	28	.worker_notify_run_action(worker_id, operation_id, action_info)
421	28	.await
422	28	}
423
424		/// Attempts to find a worker that is capable of running this action.
425		// TODO(aaronmondal) This algorithm is not very efficient. Simple testing using a tree-like
426		// structure showed worse performance on a 10_000 worker * 7 properties * 1000 queued tasks
427		// simulation of worst cases in a single threaded environment.
428	45	pub async fn find_worker_for_action(
429	45	&self,
430	45	platform_properties: &PlatformProperties,
431	45	) -> Option<WorkerId> {
432	45	let inner = self.inner.lock().await;
433	45	inner.inner_find_worker_for_action(platform_properties)
434	45	}
435
436		/// Checks to see if the worker exists in the worker pool. Should only be used in unit tests.
437		#[must_use]
438	7	pub async fn contains_worker_for_test(&self, worker_id: &WorkerId) -> bool {
439	7	let inner = self.inner.lock().await;
440	7	inner.workers.contains(worker_id)
441	7	}
442
443		/// A unit test function used to send the keep alive message to the worker from the server.
444	0	pub async fn send_keep_alive_to_worker_for_test(
445	0	&self,
446	0	worker_id: &WorkerId,
447	1	) -> Result<(), Error> {
448	1	let mut inner = self.inner.lock().await;
449	1	let worker = inner.workers.get_mut(worker_id).ok_or_else(\|\| {0
450	0	make_input_err!("WorkerId '{}' does not exist in workers map", worker_id)
451	0	})?;
452	1	worker.keep_alive()
453	1	}
454		}
455
456		#[async_trait]
457		impl WorkerScheduler for ApiWorkerScheduler {
458	1	fn get_platform_property_manager(&self) -> &PlatformPropertyManager {
459	1	self.platform_property_manager.as_ref()
460	1	}
461
462	62	async fn add_worker(&self, worker: Worker) -> Result<(), Error> {
463	31	let mut inner = self.inner.lock().await;
464	31	let worker_id = worker.id.clone();
465	31	let result = inner
466	31	.add_worker(worker)
467	31	.err_tip(\|\| "Error while adding worker, removing from pool");
468	31	if let Err( err0 ) = result { Branch (468:16): [True: 0, False: 31] Branch (468:16): [Folded - Ignored]
469	0	return Result::<(), _>::Err(err.clone())
470	0	.merge(inner.immediate_evict_worker(&worker_id, err).await);
471	31	}
472	31	Ok(())
473	62	}
474
475		async fn update_action(
476		&self,
477		worker_id: &WorkerId,
478		operation_id: &OperationId,
479		update: UpdateOperationType,
480	20	) -> Result<(), Error> {
481	10	let mut inner = self.inner.lock().await;
482	10	inner.update_action(worker_id, operation_id, update).await
483	20	}
484
485		async fn worker_keep_alive_received(
486		&self,
487		worker_id: &WorkerId,
488		timestamp: WorkerTimestamp,
489	4	) -> Result<(), Error> {
490	2	let mut inner = self.inner.lock().await;
491	2	inner
492	2	.refresh_lifetime(worker_id, timestamp)
493	2	.err_tip(\|\| "Error refreshing lifetime in worker_keep_alive_received()")
494	4	}
495
496	4	async fn remove_worker(&self, worker_id: &WorkerId) -> Result<(), Error> {
497	2	let mut inner = self.inner.lock().await;
498	2	inner
499	2	.immediate_evict_worker(
500	2	worker_id,
501	2	make_err!(Code::Internal, "Received request to remove worker"),
502	2	)
503	2	.await
504	4	}
505
506	10	async fn remove_timedout_workers(&self, now_timestamp: WorkerTimestamp) -> Result<(), Error> {
507	5	let mut inner = self.inner.lock().await;
508
509	5	let mut result = Ok(());
510		// Items should be sorted based on last_update_timestamp, so we don't need to iterate the entire
511		// map most of the time.
512	5	let worker_ids_to_remove: Vec<WorkerId> = inner
513	5	.workers
514	5	.iter()
515	5	.rev()
516	6	. map_while5 (\|(worker_id, worker)\| {
517	6	if worker.last_update_timestamp <= now_timestamp - self.worker_timeout_s { Branch (517:20): [True: 2, False: 4] Branch (517:20): [Folded - Ignored]
518	2	Some(worker_id.clone())
519		} else {
520	4	None
521		}
522	6	})
523	5	.collect();
524	7	for worker_id2 in &worker_ids_to_remove {
525	2	warn!(?worker_id, "Worker timed out, removing from pool");
526	2	result = result.merge(
527	2	inner
528	2	.immediate_evict_worker(
529	2	worker_id,
530	2	make_err!(
531	2	Code::Internal,
532	2	"Worker {worker_id} timed out, removing from pool"
533	2	),
534	2	)
535	2	.await,
536		);
537		}
538
539	5	result
540	10	}
541
542	4	async fn set_drain_worker(&self, worker_id: &WorkerId, is_draining: bool) -> Result<(), Error> {
543	2	let mut inner = self.inner.lock().await;
544	2	inner.set_drain_worker(worker_id, is_draining).await
545	4	}
546		}
547
548		impl RootMetricsComponent for ApiWorkerScheduler {}