1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572
// Copyright Materialize, Inc. and contributors. All rights reserved.
//
// Use of this software is governed by the Business Source License
// included in the LICENSE file.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0.
//! A source that reads from an a persist shard.
use std::any::Any;
use std::rc::Rc;
use std::sync::Arc;
use std::task::{Context, Poll};
use std::time::Instant;
use differential_dataflow::Hashable;
use futures::stream::StreamExt;
use futures::Stream as FuturesStream;
use timely::dataflow::channels::pact::Exchange;
use timely::dataflow::operators::generic::builder_rc::OperatorBuilder;
use timely::dataflow::operators::OkErr;
use timely::dataflow::{Scope, Stream};
use timely::progress::Antichain;
use tokio::sync::{mpsc, Mutex};
use tracing::trace;
use mz_expr::MfpPlan;
use mz_ore::cast::CastFrom;
use mz_persist::location::ExternalError;
use mz_persist_client::cache::PersistClientCache;
use mz_persist_client::fetch::{LeasedBatchPart, SerdeLeasedBatchPart};
use mz_repr::{Diff, GlobalId, Row, Timestamp};
use mz_timely_util::builder_async::{Event, OperatorBuilder as AsyncOperatorBuilder};
use crate::controller::CollectionMetadata;
use crate::types::errors::DataflowError;
use crate::types::sources::SourceData;
/// Creates a new source that reads from a persist shard, distributing the work
/// of reading data to all timely workers.
///
/// All times emitted will have been [advanced by] the given `as_of` frontier.
/// All updates at times greater or equal to `until` will be suppressed.
/// The `map_filter_project` argument, if supplied, may be partially applied,
/// and any un-applied part of the argument will be left behind in the argument.
///
/// [advanced by]: differential_dataflow::lattice::Lattice::advance_by
pub fn persist_source<G, YFn>(
scope: &G,
source_id: GlobalId,
persist_clients: Arc<Mutex<PersistClientCache>>,
metadata: CollectionMetadata,
as_of: Option<Antichain<Timestamp>>,
until: Antichain<Timestamp>,
map_filter_project: Option<&mut MfpPlan>,
yield_fn: YFn,
) -> (
Stream<G, (Row, Timestamp, Diff)>,
Stream<G, (DataflowError, Timestamp, Diff)>,
Rc<dyn Any>,
)
where
G: Scope<Timestamp = mz_repr::Timestamp>,
YFn: Fn(Instant, usize) -> bool + 'static,
{
let (stream, token) = persist_source_core(
scope,
source_id,
persist_clients,
metadata,
as_of,
until,
map_filter_project,
yield_fn,
);
let (ok_stream, err_stream) = stream.ok_err(|(d, t, r)| match d {
Ok(row) => Ok((row, t, r)),
Err(err) => Err((err, t, r)),
});
(ok_stream, err_stream, token)
}
/// The stream of batches from persist cannot be dropped at the discretion of
/// the program unaided without potentially panicking (check `LeasedBatchPart`).
/// To prevent panics, ensure that all of the stream's values are consumed,
/// irrespective of the source getting dropped.
struct DropSafeLeaseStream {
// `pin` is an `Option` only so we can move it into a task on drop.
pinned_stream: Option<
std::pin::Pin<
Box<
dyn FuturesStream<
Item = Result<
(Vec<LeasedBatchPart<Timestamp>>, Antichain<Timestamp>),
ExternalError,
>,
> + Send,
>,
>,
>,
}
impl DropSafeLeaseStream {
fn new(
pinned_stream: std::pin::Pin<
Box<
dyn FuturesStream<
Item = Result<
(Vec<LeasedBatchPart<Timestamp>>, Antichain<Timestamp>),
ExternalError,
>,
> + Send,
>,
>,
) -> DropSafeLeaseStream {
DropSafeLeaseStream {
pinned_stream: Some(pinned_stream),
}
}
fn poll_next(
&mut self,
cx: &mut Context<'_>,
) -> Poll<Option<Result<(Vec<LeasedBatchPart<Timestamp>>, Antichain<Timestamp>), ExternalError>>>
{
self.pinned_stream
.as_mut()
.expect("while being polled, pin is in place")
.as_mut()
.poll_next(cx)
}
}
impl Drop for DropSafeLeaseStream {
fn drop(&mut self) {
let mut pin = self.pinned_stream.take().expect("pin only taken on drop");
mz_ore::task::spawn(|| "LeaseStreamDrainer", async move {
// Drain the remainder of the items from the stream; this ensures
// that nothing gets dropped while in flight.
//
// Note that at the end of this task, we expect the internal stream
// to have spawned another task that receives any leases returned to
// the subscribe.
while let Some(item) = pin.as_mut().next().await {
match item {
Ok((parts, _)) => {
for part in parts {
// We don't expect this task to run for long, so we
// don't bother sending leases back to `subscribe`;
// this delays compaction slightly, so could instead
// be optimized to meaningfully return batches
// instead of dropping their leases on the floor.
let _ = part.into_exchangeable_part();
}
}
Err::<_, ExternalError>(_) => {}
}
}
});
}
}
/// Creates a new source that reads from a persist shard, distributing the work
/// of reading data to all timely workers.
///
/// All times emitted will have been [advanced by] the given `as_of` frontier.
///
/// [advanced by]: differential_dataflow::lattice::Lattice::advance_by
pub fn persist_source_core<G, YFn>(
scope: &G,
source_id: GlobalId,
persist_clients: Arc<Mutex<PersistClientCache>>,
metadata: CollectionMetadata,
as_of: Option<Antichain<Timestamp>>,
until: Antichain<Timestamp>,
mut map_filter_project: Option<&mut MfpPlan>,
yield_fn: YFn,
) -> (
Stream<G, (Result<Row, DataflowError>, Timestamp, Diff)>,
Rc<dyn Any>,
)
where
G: Scope<Timestamp = mz_repr::Timestamp>,
YFn: Fn(Instant, usize) -> bool + 'static,
{
// WARNING! If emulating any of this code, you should read the doc string on
// [`LeasedBatchPart`] and [`Subscribe`] or will likely run into intentional
// panics.
//
// This source is split as such:
// 1. Sets up `async_stream`, which only yields data (parts) on one chosen
// worker. Generating also generates SeqNo leases on the chosen worker,
// ensuring `part`s do not get GCed while in flight.
// 2. Part distribution: A timely source operator which continuously reads
// from that stream, and distributes the data among workers.
// 3. Part fetcher: A timely operator which downloads the part's contents
// from S3, and outputs them to a timely stream. Additionally, the
// operator returns the `LeasedBatchPart` to the original worker, so it
// can release the SeqNo lease.
// 4. Consumed part collector: A timely operator running only on the
// original worker that collects workers' `LeasedBatchPart`s. Internally,
// this drops the part's SeqNo lease, allowing GC to occur.
let worker_index = scope.index();
let peers = scope.peers();
let chosen_worker = usize::cast_from(source_id.hashed()) % peers;
// Extract the MFP if it exists; leave behind an identity MFP in that case.
let mut map_filter_project = map_filter_project.as_mut().map(|mfp| mfp.take());
// All of these need to be cloned out here because they're moved into the
// `try_stream!` generator.
let persist_clients_stream = Arc::<Mutex<PersistClientCache>>::clone(&persist_clients);
let persist_location_stream = metadata.persist_location.clone();
let data_shard = metadata.data_shard.clone();
let as_of_stream = as_of;
// Connects the consumed part collector operator with the part-issuing
// Subscribe.
let (consumed_part_tx, mut consumed_part_rx): (
mpsc::UnboundedSender<SerdeLeasedBatchPart>,
mpsc::UnboundedReceiver<SerdeLeasedBatchPart>,
) = mpsc::unbounded_channel();
// We want our async task to notice if we have dropped the token, as we may have reported
// the dataflow complete and advanced the read frontier of the persist collection, making
// it unsafe to attempt to open a subscription at `as_of`, leading to a crash. This token
// reveals whether we are still live to the async stream.
let handle_creation_token = Arc::new(());
let weak_handle_token = Arc::downgrade(&handle_creation_token);
let until_clone = until.clone();
// This is a generator that sets up an async `Stream` that can be continuously polled to get the
// values that are `yield`-ed from it's body.
let async_stream = async_stream::try_stream!({
// Only one worker is responsible for distributing parts
if worker_index != chosen_worker {
trace!(
"We are not the chosen worker ({}), exiting...",
chosen_worker
);
return;
}
let read = persist_clients_stream
.lock()
.await
.open(persist_location_stream)
.await;
// This is a moment where we may have dropped our source if our token
// has been dropped, but if we still hold it we should be good to go.
if weak_handle_token.upgrade().is_none() {
return;
}
let read = read
.expect("could not open persist client")
.open_reader::<SourceData, (), mz_repr::Timestamp, mz_repr::Diff>(data_shard)
.await;
// This is a moment where we may have dropped our source if our token
// has been dropped, but if we still hold it we should be good to go.
if weak_handle_token.upgrade().is_none() {
return;
}
let read = read.expect("could not open persist shard");
let as_of_stream = as_of_stream.unwrap_or_else(|| read.since().clone());
// Eagerly yield the initial as_of. This makes sure that the output
// frontier of the `persist_source` closely tracks the `upper` frontier
// of the persist shard. It might be that the snapshot for `as_of` is
// not initially available yet, but this makes sure we already downgrade
// to it.
//
// Downstream consumers might rely on close frontier tracking for making
// progress. For example, the `persist_sink` needs to know the
// up-to-date uppper of the output shard to make progress because it
// will only write out new data once it knows that earlier writes went
// through, including the initial downgrade of the shard upper to the
// `as_of`.
yield (Vec::new(), as_of_stream.clone());
let subscription = read.subscribe(as_of_stream.clone()).await;
// This is a moment where we may have let persist compact if our token
// has been dropped, but if we still hold it we should be good to go.
if weak_handle_token.upgrade().is_none() {
return;
}
let mut subscription = subscription.unwrap_or_else(|e| {
panic!(
"{source_id}: {} cannot serve requested as_of {:?}: {:?}",
data_shard, as_of_stream, e
)
});
let mut done = false;
while !done {
while let Ok(leased_part) = consumed_part_rx.try_recv() {
subscription.return_leased_part(leased_part.into());
}
let (parts, progress) = subscription.next().await;
// If `until.less_equal(progress)`, it means that all subsequent batches will
// contain only times greater or equal to `until`, which means they can be dropped
// in their entirety. The current batch must be emitted, but we can stop afterwards.
if timely::PartialOrder::less_equal(&until_clone, &progress) {
done = true;
}
yield (parts, progress);
}
// Rather than simply end, we spawn a task that can continue to return leases.
// This keeps the `ReadHandle` alive until we have confirmed each lease as read.
mz_ore::task::spawn(|| "LeaseReturner", async move {
while let Some(leased_part) = consumed_part_rx.recv().await {
subscription.return_leased_part(leased_part.into());
}
});
});
let mut pinned_stream = DropSafeLeaseStream::new(Box::pin(async_stream));
let (inner, token) = crate::source::util::source(
scope,
format!("persist_source {}: part distribution", source_id),
move |info| {
let waker_activator = Arc::new(scope.sync_activator_for(&info.address[..]));
let waker = futures::task::waker(waker_activator);
let mut current_ts = timely::progress::Timestamp::minimum();
move |cap_set, output| {
let mut context = Context::from_waker(&waker);
while let Poll::Ready(item) = pinned_stream.poll_next(&mut context) {
match item {
Some(Ok((parts, progress))) => {
let session_cap = cap_set.delayed(¤t_ts);
let mut session = output.session(&session_cap);
for part in parts {
// Give the part to a random worker.
let worker_idx = usize::cast_from(Instant::now().hashed()) % peers;
session.give((worker_idx, part.into_exchangeable_part()));
}
cap_set.downgrade(progress.iter());
match progress.into_option() {
Some(ts) => {
current_ts = ts;
}
None => {
cap_set.downgrade(&[]);
return;
}
}
}
Some(Err::<_, ExternalError>(e)) => {
panic!("unexpected error from persist {e}")
}
// We never expect any further output from
// `pinned_stream`, so propagate that information
// downstream.
None => {
cap_set.downgrade(&[]);
return;
}
}
}
}
},
);
let mut fetcher_builder = AsyncOperatorBuilder::new(
format!("persist_source {}: part fetcher", source_id),
scope.clone(),
);
let mut fetcher_input = fetcher_builder.new_input(
&inner,
Exchange::new(|&(i, _): &(usize, _)| u64::cast_from(i)),
);
let (mut update_output, update_output_stream) = fetcher_builder.new_output();
let (mut consumed_part_output, consumed_part_output_stream) = fetcher_builder.new_output();
// Re-used state for processing and building rows.
let mut datum_vec = mz_repr::DatumVec::new();
let mut row_builder = Row::default();
fetcher_builder.build(move |_capabilities| async move {
let fetcher = persist_clients
.lock()
.await
.open(metadata.persist_location.clone())
.await
.expect("could not open persist client")
.open_reader::<SourceData, (), mz_repr::Timestamp, mz_repr::Diff>(data_shard)
.await
.expect("could not open persist shard")
.batch_fetcher()
.await;
let mut buffer = Vec::new();
while let Some(event) = fetcher_input.next().await {
// Re-acquire the output handle on each invocation and drop it
// when we're done.
let mut output_handle = update_output.activate();
let mut consumed_part_output_handle = consumed_part_output.activate();
if let Event::Data(cap, data) = event {
// `LeasedBatchPart`es cannot be dropped at this point w/o
// panicking, so swap them to an owned version.
data.swap(&mut buffer);
let mut update_session = output_handle.session(&cap);
let mut consumed_part_session = consumed_part_output_handle.session(&cap);
for (_idx, part) in buffer.drain(..) {
let (consumed_part, fetched_part) =
fetcher.fetch_leased_part(part.into()).await;
let fetched_part = fetched_part
.expect("shard_id generated for sources must match across all workers");
// SUBTLE: This operator yields back to timely whenever an await returns a
// Pending result from the overall async/await state machine `poll`. Since
// this is fetching from remote storage, it will yield and thus we can reset
// our yield counters here.
let mut decode_start = Instant::now();
// Apply as much logic to `updates` as we can, before we emit anything.
let (updates_size_hint_min, updates_size_hint_max) = fetched_part.size_hint();
let mut updates =
Vec::with_capacity(updates_size_hint_max.unwrap_or(updates_size_hint_min));
for ((key, val), time, diff) in fetched_part {
if !until.less_equal(&time) {
match (key, val) {
(Ok(SourceData(Ok(row))), Ok(())) => {
if let Some(mfp) = &mut map_filter_project {
let arena = mz_repr::RowArena::new();
let mut datums_local = datum_vec.borrow_with(&row);
for result in mfp.evaluate(
&mut datums_local,
&arena,
time,
diff,
|time| !until.less_equal(time),
&mut row_builder,
) {
match result {
Ok((row, time, diff)) => {
// Additional `until` filtering due to temporal filters.
if !until.less_equal(&time) {
updates.push((Ok(row), time, diff));
}
}
Err((err, time, diff)) => {
// Additional `until` filtering due to temporal filters.
if !until.less_equal(&time) {
updates.push((Err(err), time, diff));
}
}
}
}
} else {
updates.push((Ok(row), time, diff));
}
}
(Ok(SourceData(Err(err))), Ok(())) => {
updates.push((Err(err), time, diff));
}
// TODO(petrosagg): error handling
(Err(_), Ok(_)) | (Ok(_), Err(_)) | (Err(_), Err(_)) => {
panic!("decoding failed")
}
}
}
if yield_fn(decode_start, updates.len()) {
// A large part of the point of yielding is to let later operators
// reduce down the data, so emit what we have. Note that this means
// we don't get to consolidate everything, but that's part of the
// tradeoff in tuning yield_fn.
differential_dataflow::consolidation::consolidate_updates(&mut updates);
update_session.give_vec(&mut updates);
force_yield().await;
decode_start = Instant::now();
}
}
differential_dataflow::consolidation::consolidate_updates(&mut updates);
update_session.give_vec(&mut updates);
consumed_part_session.give(consumed_part.into_exchangeable_part());
}
}
}
});
// This operator is meant to only run on the chosen worker. All workers will
// exchange their fetched ("consumed") parts back to the leasor.
let mut consumed_part_builder = OperatorBuilder::new(
format!("persist_source {}: consumed part collector", source_id),
scope.clone(),
);
// Exchange all "consumed" parts back to the chosen worker/leasor.
let mut consumed_part_input = consumed_part_builder.new_input(
&consumed_part_output_stream,
Exchange::new(move |_| u64::cast_from(chosen_worker)),
);
let last_token = Rc::new(token);
let token = Rc::clone(&last_token);
consumed_part_builder.build(|_initial_capabilities| {
let mut buffer = Vec::new();
move |_frontiers| {
// The chosen worker is the leasor because it issues batches.
if worker_index != chosen_worker {
trace!(
"We are not the batch leasor for {:?}, exiting...",
source_id
);
return;
}
while let Some((_cap, data)) = consumed_part_input.next() {
data.swap(&mut buffer);
for part in buffer.drain(..) {
if let Err(mpsc::error::SendError(_part)) = consumed_part_tx.send(part) {
// Subscribe loop dropped, which drops its ReadHandle,
// which in turn drops all leases, so doing anything
// else here is both moot and impossible.
//
// The parts we tried to send will just continue being
// `SerdeLeasedBatchPart`'es.
}
}
}
}
});
let token = Rc::new((token, handle_creation_token));
(update_output_stream, token)
}
// The build_async operator yields to timely whenever the Future handed to it
// (in practice an async/await state machine) returns from `poll` with a
// `Pending`. Force a yield by constructing a future that returns Pending the
// first time it's polled and `Ready` the second.
//
// This allows us to yield without having to do anything special to stash
// in-progress work. (Basically, the async/await state machine does it for us.)
async fn force_yield() {
let mut polled = false;
let () = futures::future::poll_fn(move |cx| match polled {
true => Poll::Ready(()),
false => {
polled = true;
cx.waker().wake_by_ref();
Poll::Pending
}
})
.await;
}