Line data Source code
1 : //! An ImageLayer represents an image or a snapshot of a key-range at
2 : //! one particular LSN.
3 : //!
4 : //! It contains an image of all key-value pairs in its key-range. Any key
5 : //! that falls into the image layer's range but does not exist in the layer,
6 : //! does not exist.
7 : //!
8 : //! An image layer is stored in a file on disk. The file is stored in
9 : //! timelines/<timeline_id> directory. Currently, there are no
10 : //! subdirectories, and each image layer file is named like this:
11 : //!
12 : //! ```text
13 : //! <key start>-<key end>__<LSN>
14 : //! ```
15 : //!
16 : //! For example:
17 : //!
18 : //! ```text
19 : //! 000000067F000032BE0000400000000070B6-000000067F000032BE0000400000000080B6__00000000346BC568
20 : //! ```
21 : //!
22 : //! Every image layer file consists of three parts: "summary",
23 : //! "index", and "values". The summary is a fixed size header at the
24 : //! beginning of the file, and it contains basic information about the
25 : //! layer, and offsets to the other parts. The "index" is a B-tree,
26 : //! mapping from Key to an offset in the "values" part. The
27 : //! actual page images are stored in the "values" part.
28 : use std::collections::{HashMap, VecDeque};
29 : use std::fs::File;
30 : use std::io::SeekFrom;
31 : use std::ops::Range;
32 : use std::os::unix::prelude::FileExt;
33 : use std::str::FromStr;
34 : use std::sync::Arc;
35 :
36 : use anyhow::{Context, Result, bail, ensure};
37 : use bytes::Bytes;
38 : use camino::{Utf8Path, Utf8PathBuf};
39 : use hex;
40 : use itertools::Itertools;
41 : use pageserver_api::config::MaxVectoredReadBytes;
42 : use pageserver_api::key::{DBDIR_KEY, KEY_SIZE, Key};
43 : use pageserver_api::keyspace::KeySpace;
44 : use pageserver_api::shard::{ShardIdentity, TenantShardId};
45 : use pageserver_api::value::Value;
46 : use rand::Rng;
47 : use rand::distributions::Alphanumeric;
48 : use serde::{Deserialize, Serialize};
49 : use tokio::sync::OnceCell;
50 : use tokio_stream::StreamExt;
51 : use tokio_util::sync::CancellationToken;
52 : use tracing::*;
53 : use utils::bin_ser::BeSer;
54 : use utils::id::{TenantId, TimelineId};
55 : use utils::lsn::Lsn;
56 :
57 : use super::layer_name::ImageLayerName;
58 : use super::{
59 : AsLayerDesc, LayerName, OnDiskValue, OnDiskValueIo, PersistentLayerDesc, ResidentLayer,
60 : ValuesReconstructState,
61 : };
62 : use crate::config::PageServerConf;
63 : use crate::context::{PageContentKind, RequestContext, RequestContextBuilder};
64 : use crate::page_cache::{self, FileId, PAGE_SZ};
65 : use crate::tenant::blob_io::BlobWriter;
66 : use crate::tenant::block_io::{BlockBuf, FileBlockReader};
67 : use crate::tenant::disk_btree::{
68 : DiskBtreeBuilder, DiskBtreeIterator, DiskBtreeReader, VisitDirection,
69 : };
70 : use crate::tenant::timeline::GetVectoredError;
71 : use crate::tenant::vectored_blob_io::{
72 : BlobFlag, BufView, StreamingVectoredReadPlanner, VectoredBlobReader, VectoredRead,
73 : VectoredReadPlanner,
74 : };
75 : use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
76 : use crate::virtual_file::{self, IoBufferMut, MaybeFatalIo, VirtualFile};
77 : use crate::{IMAGE_FILE_MAGIC, STORAGE_FORMAT_VERSION, TEMP_FILE_SUFFIX};
78 :
79 : ///
80 : /// Header stored in the beginning of the file
81 : ///
82 : /// After this comes the 'values' part, starting on block 1. After that,
83 : /// the 'index' starts at the block indicated by 'index_start_blk'
84 : ///
85 0 : #[derive(Debug, Serialize, Deserialize, PartialEq, Eq)]
86 : pub struct Summary {
87 : /// Magic value to identify this as a neon image file. Always IMAGE_FILE_MAGIC.
88 : pub magic: u16,
89 : pub format_version: u16,
90 :
91 : pub tenant_id: TenantId,
92 : pub timeline_id: TimelineId,
93 : pub key_range: Range<Key>,
94 : pub lsn: Lsn,
95 :
96 : /// Block number where the 'index' part of the file begins.
97 : pub index_start_blk: u32,
98 : /// Block within the 'index', where the B-tree root page is stored
99 : pub index_root_blk: u32,
100 : // the 'values' part starts after the summary header, on block 1.
101 : }
102 :
103 : impl From<&ImageLayer> for Summary {
104 0 : fn from(layer: &ImageLayer) -> Self {
105 0 : Self::expected(
106 0 : layer.desc.tenant_shard_id.tenant_id,
107 0 : layer.desc.timeline_id,
108 0 : layer.desc.key_range.clone(),
109 0 : layer.lsn,
110 0 : )
111 0 : }
112 : }
113 :
114 : impl Summary {
115 900 : pub(super) fn expected(
116 900 : tenant_id: TenantId,
117 900 : timeline_id: TimelineId,
118 900 : key_range: Range<Key>,
119 900 : lsn: Lsn,
120 900 : ) -> Self {
121 900 : Self {
122 900 : magic: IMAGE_FILE_MAGIC,
123 900 : format_version: STORAGE_FORMAT_VERSION,
124 900 : tenant_id,
125 900 : timeline_id,
126 900 : key_range,
127 900 : lsn,
128 900 :
129 900 : index_start_blk: 0,
130 900 : index_root_blk: 0,
131 900 : }
132 900 : }
133 : }
134 :
135 : /// This is used only from `pagectl`. Within pageserver, all layers are
136 : /// [`crate::tenant::storage_layer::Layer`], which can hold an [`ImageLayerInner`].
137 : pub struct ImageLayer {
138 : path: Utf8PathBuf,
139 : pub desc: PersistentLayerDesc,
140 : // This entry contains an image of all pages as of this LSN, should be the same as desc.lsn
141 : pub lsn: Lsn,
142 : inner: OnceCell<ImageLayerInner>,
143 : }
144 :
145 : impl std::fmt::Debug for ImageLayer {
146 0 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
147 : use super::RangeDisplayDebug;
148 :
149 0 : f.debug_struct("ImageLayer")
150 0 : .field("key_range", &RangeDisplayDebug(&self.desc.key_range))
151 0 : .field("file_size", &self.desc.file_size)
152 0 : .field("lsn", &self.lsn)
153 0 : .field("inner", &self.inner)
154 0 : .finish()
155 0 : }
156 : }
157 :
158 : /// ImageLayer is the in-memory data structure associated with an on-disk image
159 : /// file.
160 : pub struct ImageLayerInner {
161 : // values copied from summary
162 : index_start_blk: u32,
163 : index_root_blk: u32,
164 :
165 : key_range: Range<Key>,
166 : lsn: Lsn,
167 :
168 : file: Arc<VirtualFile>,
169 : file_id: FileId,
170 :
171 : max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
172 : }
173 :
174 : impl ImageLayerInner {
175 0 : pub(crate) fn layer_dbg_info(&self) -> String {
176 0 : format!(
177 0 : "image {}..{} {}",
178 0 : self.key_range().start,
179 0 : self.key_range().end,
180 0 : self.lsn()
181 0 : )
182 0 : }
183 : }
184 :
185 : impl std::fmt::Debug for ImageLayerInner {
186 0 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
187 0 : f.debug_struct("ImageLayerInner")
188 0 : .field("index_start_blk", &self.index_start_blk)
189 0 : .field("index_root_blk", &self.index_root_blk)
190 0 : .finish()
191 0 : }
192 : }
193 :
194 : impl ImageLayerInner {
195 0 : pub(super) async fn dump(&self, ctx: &RequestContext) -> anyhow::Result<()> {
196 0 : let block_reader = FileBlockReader::new(&self.file, self.file_id);
197 0 : let tree_reader = DiskBtreeReader::<_, KEY_SIZE>::new(
198 0 : self.index_start_blk,
199 0 : self.index_root_blk,
200 0 : block_reader,
201 0 : );
202 0 :
203 0 : tree_reader.dump(ctx).await?;
204 :
205 0 : tree_reader
206 0 : .visit(
207 0 : &[0u8; KEY_SIZE],
208 0 : VisitDirection::Forwards,
209 0 : |key, value| {
210 0 : println!("key: {} offset {}", hex::encode(key), value);
211 0 : true
212 0 : },
213 0 : ctx,
214 0 : )
215 0 : .await?;
216 :
217 0 : Ok(())
218 0 : }
219 : }
220 :
221 : /// Boilerplate to implement the Layer trait, always use layer_desc for persistent layers.
222 : impl std::fmt::Display for ImageLayer {
223 0 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
224 0 : write!(f, "{}", self.layer_desc().short_id())
225 0 : }
226 : }
227 :
228 : impl AsLayerDesc for ImageLayer {
229 0 : fn layer_desc(&self) -> &PersistentLayerDesc {
230 0 : &self.desc
231 0 : }
232 : }
233 :
234 : impl ImageLayer {
235 0 : pub async fn dump(&self, verbose: bool, ctx: &RequestContext) -> Result<()> {
236 0 : self.desc.dump();
237 0 :
238 0 : if !verbose {
239 0 : return Ok(());
240 0 : }
241 :
242 0 : let inner = self.load(ctx).await?;
243 :
244 0 : inner.dump(ctx).await?;
245 :
246 0 : Ok(())
247 0 : }
248 :
249 3744 : fn temp_path_for(
250 3744 : conf: &PageServerConf,
251 3744 : timeline_id: TimelineId,
252 3744 : tenant_shard_id: TenantShardId,
253 3744 : fname: &ImageLayerName,
254 3744 : ) -> Utf8PathBuf {
255 3744 : let rand_string: String = rand::thread_rng()
256 3744 : .sample_iter(&Alphanumeric)
257 3744 : .take(8)
258 3744 : .map(char::from)
259 3744 : .collect();
260 3744 :
261 3744 : conf.timeline_path(&tenant_shard_id, &timeline_id)
262 3744 : .join(format!("{fname}.{rand_string}.{TEMP_FILE_SUFFIX}"))
263 3744 : }
264 :
265 : ///
266 : /// Open the underlying file and read the metadata into memory, if it's
267 : /// not loaded already.
268 : ///
269 0 : async fn load(&self, ctx: &RequestContext) -> Result<&ImageLayerInner> {
270 0 : self.inner
271 0 : .get_or_try_init(|| self.load_inner(ctx))
272 0 : .await
273 0 : .with_context(|| format!("Failed to load image layer {}", self.path()))
274 0 : }
275 :
276 0 : async fn load_inner(&self, ctx: &RequestContext) -> Result<ImageLayerInner> {
277 0 : let path = self.path();
278 :
279 0 : let loaded =
280 0 : ImageLayerInner::load(&path, self.desc.image_layer_lsn(), None, None, ctx).await?;
281 :
282 : // not production code
283 0 : let actual_layer_name = LayerName::from_str(path.file_name().unwrap()).unwrap();
284 0 : let expected_layer_name = self.layer_desc().layer_name();
285 0 :
286 0 : if actual_layer_name != expected_layer_name {
287 0 : println!("warning: filename does not match what is expected from in-file summary");
288 0 : println!("actual: {:?}", actual_layer_name.to_string());
289 0 : println!("expected: {:?}", expected_layer_name.to_string());
290 0 : }
291 :
292 0 : Ok(loaded)
293 0 : }
294 :
295 : /// Create an ImageLayer struct representing an existing file on disk.
296 : ///
297 : /// This variant is only used for debugging purposes, by the 'pagectl' binary.
298 0 : pub fn new_for_path(path: &Utf8Path, file: File) -> Result<ImageLayer> {
299 0 : let mut summary_buf = vec![0; PAGE_SZ];
300 0 : file.read_exact_at(&mut summary_buf, 0)?;
301 0 : let summary = Summary::des_prefix(&summary_buf)?;
302 0 : let metadata = file
303 0 : .metadata()
304 0 : .context("get file metadata to determine size")?;
305 :
306 : // This function is never used for constructing layers in a running pageserver,
307 : // so it does not need an accurate TenantShardId.
308 0 : let tenant_shard_id = TenantShardId::unsharded(summary.tenant_id);
309 0 :
310 0 : Ok(ImageLayer {
311 0 : path: path.to_path_buf(),
312 0 : desc: PersistentLayerDesc::new_img(
313 0 : tenant_shard_id,
314 0 : summary.timeline_id,
315 0 : summary.key_range,
316 0 : summary.lsn,
317 0 : metadata.len(),
318 0 : ), // Now we assume image layer ALWAYS covers the full range. This may change in the future.
319 0 : lsn: summary.lsn,
320 0 : inner: OnceCell::new(),
321 0 : })
322 0 : }
323 :
324 0 : fn path(&self) -> Utf8PathBuf {
325 0 : self.path.clone()
326 0 : }
327 : }
328 :
329 : #[derive(thiserror::Error, Debug)]
330 : pub enum RewriteSummaryError {
331 : #[error("magic mismatch")]
332 : MagicMismatch,
333 : #[error(transparent)]
334 : Other(#[from] anyhow::Error),
335 : }
336 :
337 : impl From<std::io::Error> for RewriteSummaryError {
338 0 : fn from(e: std::io::Error) -> Self {
339 0 : Self::Other(anyhow::anyhow!(e))
340 0 : }
341 : }
342 :
343 : impl ImageLayer {
344 0 : pub async fn rewrite_summary<F>(
345 0 : path: &Utf8Path,
346 0 : rewrite: F,
347 0 : ctx: &RequestContext,
348 0 : ) -> Result<(), RewriteSummaryError>
349 0 : where
350 0 : F: Fn(Summary) -> Summary,
351 0 : {
352 0 : let mut file = VirtualFile::open_with_options(
353 0 : path,
354 0 : virtual_file::OpenOptions::new().read(true).write(true),
355 0 : ctx,
356 0 : )
357 0 : .await
358 0 : .with_context(|| format!("Failed to open file '{}'", path))?;
359 0 : let file_id = page_cache::next_file_id();
360 0 : let block_reader = FileBlockReader::new(&file, file_id);
361 0 : let summary_blk = block_reader.read_blk(0, ctx).await?;
362 0 : let actual_summary = Summary::des_prefix(summary_blk.as_ref()).context("deserialize")?;
363 0 : if actual_summary.magic != IMAGE_FILE_MAGIC {
364 0 : return Err(RewriteSummaryError::MagicMismatch);
365 0 : }
366 0 :
367 0 : let new_summary = rewrite(actual_summary);
368 0 :
369 0 : let mut buf = Vec::with_capacity(PAGE_SZ);
370 0 : // TODO: could use smallvec here but it's a pain with Slice<T>
371 0 : Summary::ser_into(&new_summary, &mut buf).context("serialize")?;
372 0 : file.seek(SeekFrom::Start(0)).await?;
373 0 : let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
374 0 : res?;
375 0 : Ok(())
376 0 : }
377 : }
378 :
379 : impl ImageLayerInner {
380 840 : pub(crate) fn key_range(&self) -> &Range<Key> {
381 840 : &self.key_range
382 840 : }
383 :
384 840 : pub(crate) fn lsn(&self) -> Lsn {
385 840 : self.lsn
386 840 : }
387 :
388 900 : pub(super) async fn load(
389 900 : path: &Utf8Path,
390 900 : lsn: Lsn,
391 900 : summary: Option<Summary>,
392 900 : max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
393 900 : ctx: &RequestContext,
394 900 : ) -> anyhow::Result<Self> {
395 900 : let file = Arc::new(
396 900 : VirtualFile::open_v2(path, ctx)
397 900 : .await
398 900 : .context("open layer file")?,
399 : );
400 900 : let file_id = page_cache::next_file_id();
401 900 : let block_reader = FileBlockReader::new(&file, file_id);
402 900 : let summary_blk = block_reader
403 900 : .read_blk(0, ctx)
404 900 : .await
405 900 : .context("read first block")?;
406 :
407 : // length is the only way how this could fail, so it's not actually likely at all unless
408 : // read_blk returns wrong sized block.
409 : //
410 : // TODO: confirm and make this into assertion
411 900 : let actual_summary =
412 900 : Summary::des_prefix(summary_blk.as_ref()).context("deserialize first block")?;
413 :
414 900 : if let Some(mut expected_summary) = summary {
415 : // production code path
416 900 : expected_summary.index_start_blk = actual_summary.index_start_blk;
417 900 : expected_summary.index_root_blk = actual_summary.index_root_blk;
418 900 : // mask out the timeline_id, but still require the layers to be from the same tenant
419 900 : expected_summary.timeline_id = actual_summary.timeline_id;
420 900 :
421 900 : if actual_summary != expected_summary {
422 0 : bail!(
423 0 : "in-file summary does not match expected summary. actual = {:?} expected = {:?}",
424 0 : actual_summary,
425 0 : expected_summary
426 0 : );
427 900 : }
428 0 : }
429 :
430 900 : Ok(ImageLayerInner {
431 900 : index_start_blk: actual_summary.index_start_blk,
432 900 : index_root_blk: actual_summary.index_root_blk,
433 900 : lsn,
434 900 : file,
435 900 : file_id,
436 900 : max_vectored_read_bytes,
437 900 : key_range: actual_summary.key_range,
438 900 : })
439 900 : }
440 :
441 : // Look up the keys in the provided keyspace and update
442 : // the reconstruct state with whatever is found.
443 181512 : pub(super) async fn get_values_reconstruct_data(
444 181512 : &self,
445 181512 : this: ResidentLayer,
446 181512 : keyspace: KeySpace,
447 181512 : reconstruct_state: &mut ValuesReconstructState,
448 181512 : ctx: &RequestContext,
449 181512 : ) -> Result<(), GetVectoredError> {
450 181512 : let reads = self
451 181512 : .plan_reads(keyspace, None, ctx)
452 181512 : .await
453 181512 : .map_err(GetVectoredError::Other)?;
454 :
455 181512 : self.do_reads_and_update_state(this, reads, reconstruct_state, ctx)
456 181512 : .await;
457 :
458 181512 : reconstruct_state.on_image_layer_visited(&self.key_range);
459 181512 :
460 181512 : Ok(())
461 181512 : }
462 :
463 : /// Traverse the layer's index to build read operations on the overlap of the input keyspace
464 : /// and the keys in this layer.
465 : ///
466 : /// If shard_identity is provided, it will be used to filter keys down to those stored on
467 : /// this shard.
468 181560 : async fn plan_reads(
469 181560 : &self,
470 181560 : keyspace: KeySpace,
471 181560 : shard_identity: Option<&ShardIdentity>,
472 181560 : ctx: &RequestContext,
473 181560 : ) -> anyhow::Result<Vec<VectoredRead>> {
474 181560 : let mut planner = VectoredReadPlanner::new(
475 181560 : self.max_vectored_read_bytes
476 181560 : .expect("Layer is loaded with max vectored bytes config")
477 181560 : .0
478 181560 : .into(),
479 181560 : );
480 181560 :
481 181560 : let block_reader = FileBlockReader::new(&self.file, self.file_id);
482 181560 : let tree_reader =
483 181560 : DiskBtreeReader::new(self.index_start_blk, self.index_root_blk, block_reader);
484 181560 :
485 181560 : let ctx = RequestContextBuilder::from(ctx)
486 181560 : .page_content_kind(PageContentKind::ImageLayerBtreeNode)
487 181560 : .attached_child();
488 :
489 269148 : for range in keyspace.ranges.iter() {
490 269148 : let mut range_end_handled = false;
491 269148 : let mut search_key: [u8; KEY_SIZE] = [0u8; KEY_SIZE];
492 269148 : range.start.write_to_byte_slice(&mut search_key);
493 269148 :
494 269148 : let index_stream = tree_reader.clone().into_stream(&search_key, &ctx);
495 269148 : let mut index_stream = std::pin::pin!(index_stream);
496 :
497 1209540 : while let Some(index_entry) = index_stream.next().await {
498 1204944 : let (raw_key, offset) = index_entry?;
499 :
500 1204944 : let key = Key::from_slice(&raw_key[..KEY_SIZE]);
501 1204944 : assert!(key >= range.start);
502 :
503 1204944 : let flag = if let Some(shard_identity) = shard_identity {
504 393216 : if shard_identity.is_key_disposable(&key) {
505 294912 : BlobFlag::Ignore
506 : } else {
507 98304 : BlobFlag::None
508 : }
509 : } else {
510 811728 : BlobFlag::None
511 : };
512 :
513 1204944 : if key >= range.end {
514 264552 : planner.handle_range_end(offset);
515 264552 : range_end_handled = true;
516 264552 : break;
517 940392 : } else {
518 940392 : planner.handle(key, self.lsn, offset, flag);
519 940392 : }
520 : }
521 :
522 269148 : if !range_end_handled {
523 4596 : let payload_end = self.index_start_blk as u64 * PAGE_SZ as u64;
524 4596 : planner.handle_range_end(payload_end);
525 264552 : }
526 : }
527 :
528 181560 : Ok(planner.finish())
529 181560 : }
530 :
531 : /// Given a key range, select the parts of that range that should be retained by the ShardIdentity,
532 : /// then execute vectored GET operations, passing the results of all read keys into the writer.
533 48 : pub(super) async fn filter(
534 48 : &self,
535 48 : shard_identity: &ShardIdentity,
536 48 : writer: &mut ImageLayerWriter,
537 48 : ctx: &RequestContext,
538 48 : ) -> anyhow::Result<usize> {
539 : // Fragment the range into the regions owned by this ShardIdentity
540 48 : let plan = self
541 48 : .plan_reads(
542 48 : KeySpace {
543 48 : // If asked for the total key space, plan_reads will give us all the keys in the layer
544 48 : ranges: vec![Key::MIN..Key::MAX],
545 48 : },
546 48 : Some(shard_identity),
547 48 : ctx,
548 48 : )
549 48 : .await?;
550 :
551 48 : let vectored_blob_reader = VectoredBlobReader::new(&self.file);
552 48 : let mut key_count = 0;
553 48 : for read in plan.into_iter() {
554 48 : let buf_size = read.size();
555 48 :
556 48 : let buf = IoBufferMut::with_capacity(buf_size);
557 48 : let blobs_buf = vectored_blob_reader.read_blobs(&read, buf, ctx).await?;
558 :
559 48 : let view = BufView::new_slice(&blobs_buf.buf);
560 :
561 98304 : for meta in blobs_buf.blobs.iter() {
562 : // Just read the raw header+data and pass it through to the target layer, without
563 : // decoding and recompressing it.
564 98304 : let raw = meta.raw_with_header(&view);
565 98304 : key_count += 1;
566 98304 : writer
567 98304 : .put_image_raw(meta.meta.key, raw.into_bytes(), ctx)
568 98304 : .await
569 98304 : .context(format!("Storing key {}", meta.meta.key))?;
570 : }
571 : }
572 :
573 48 : Ok(key_count)
574 48 : }
575 :
576 181512 : async fn do_reads_and_update_state(
577 181512 : &self,
578 181512 : this: ResidentLayer,
579 181512 : reads: Vec<VectoredRead>,
580 181512 : reconstruct_state: &mut ValuesReconstructState,
581 181512 : ctx: &RequestContext,
582 181512 : ) {
583 181512 : let max_vectored_read_bytes = self
584 181512 : .max_vectored_read_bytes
585 181512 : .expect("Layer is loaded with max vectored bytes config")
586 181512 : .0
587 181512 : .into();
588 :
589 207264 : for read in reads.into_iter() {
590 207264 : let mut ios: HashMap<(Key, Lsn), OnDiskValueIo> = Default::default();
591 547176 : for (_, blob_meta) in read.blobs_at.as_slice() {
592 547176 : let io = reconstruct_state.update_key(&blob_meta.key, blob_meta.lsn, true);
593 547176 : ios.insert((blob_meta.key, blob_meta.lsn), io);
594 547176 : }
595 :
596 207264 : let buf_size = read.size();
597 207264 :
598 207264 : if buf_size > max_vectored_read_bytes {
599 : // If the read is oversized, it should only contain one key.
600 0 : let offenders = read
601 0 : .blobs_at
602 0 : .as_slice()
603 0 : .iter()
604 0 : .filter_map(|(_, blob_meta)| {
605 0 : if blob_meta.key.is_rel_dir_key()
606 0 : || blob_meta.key == DBDIR_KEY
607 0 : || blob_meta.key.is_aux_file_key()
608 : {
609 : // The size of values for these keys is unbounded and can
610 : // grow very large in pathological cases.
611 0 : None
612 : } else {
613 0 : Some(format!("{}@{}", blob_meta.key, blob_meta.lsn))
614 : }
615 0 : })
616 0 : .join(", ");
617 0 :
618 0 : if !offenders.is_empty() {
619 0 : tracing::warn!(
620 0 : "Oversized vectored read ({} > {}) for keys {}",
621 : buf_size,
622 : max_vectored_read_bytes,
623 : offenders
624 : );
625 0 : }
626 207264 : }
627 :
628 207264 : let read_extend_residency = this.clone();
629 207264 : let read_from = self.file.clone();
630 207264 : let read_ctx = ctx.attached_child();
631 207264 : reconstruct_state
632 207264 : .spawn_io(async move {
633 207264 : let buf = IoBufferMut::with_capacity(buf_size);
634 207264 : let vectored_blob_reader = VectoredBlobReader::new(&read_from);
635 207264 : let res = vectored_blob_reader.read_blobs(&read, buf, &read_ctx).await;
636 :
637 207264 : match res {
638 207264 : Ok(blobs_buf) => {
639 207264 : let view = BufView::new_slice(&blobs_buf.buf);
640 547176 : for meta in blobs_buf.blobs.iter() {
641 547176 : let io: OnDiskValueIo =
642 547176 : ios.remove(&(meta.meta.key, meta.meta.lsn)).unwrap();
643 547176 : let img_buf = meta.read(&view).await;
644 :
645 547176 : let img_buf = match img_buf {
646 547176 : Ok(img_buf) => img_buf,
647 0 : Err(e) => {
648 0 : io.complete(Err(e));
649 0 : continue;
650 : }
651 : };
652 :
653 547176 : io.complete(Ok(OnDiskValue::RawImage(img_buf.into_bytes())));
654 : }
655 :
656 207264 : assert!(ios.is_empty());
657 : }
658 0 : Err(err) => {
659 0 : for (_, io) in ios {
660 0 : io.complete(Err(std::io::Error::new(
661 0 : err.kind(),
662 0 : "vec read failed",
663 0 : )));
664 0 : }
665 : }
666 : }
667 :
668 : // keep layer resident until this IO is done; this spawned IO future generally outlives the
669 : // call to `self` / the `Arc<DownloadedLayer>` / the `ResidentLayer` that guarantees residency
670 207264 : drop(read_extend_residency);
671 207264 : })
672 207264 : .await;
673 : }
674 181512 : }
675 :
676 756 : pub(crate) fn iter<'a>(&'a self, ctx: &'a RequestContext) -> ImageLayerIterator<'a> {
677 756 : let block_reader = FileBlockReader::new(&self.file, self.file_id);
678 756 : let tree_reader =
679 756 : DiskBtreeReader::new(self.index_start_blk, self.index_root_blk, block_reader);
680 756 : ImageLayerIterator {
681 756 : image_layer: self,
682 756 : ctx,
683 756 : index_iter: tree_reader.iter(&[0; KEY_SIZE], ctx),
684 756 : key_values_batch: VecDeque::new(),
685 756 : is_end: false,
686 756 : planner: StreamingVectoredReadPlanner::new(
687 756 : 1024 * 8192, // The default value. Unit tests might use a different value. 1024 * 8K = 8MB buffer.
688 756 : 1024, // The default value. Unit tests might use a different value
689 756 : ),
690 756 : }
691 756 : }
692 :
693 : /// NB: not super efficient, but not terrible either. Should prob be an iterator.
694 : //
695 : // We're reusing the index traversal logical in plan_reads; would be nice to
696 : // factor that out.
697 0 : pub(crate) async fn load_keys(&self, ctx: &RequestContext) -> anyhow::Result<Vec<Key>> {
698 0 : let plan = self
699 0 : .plan_reads(KeySpace::single(self.key_range.clone()), None, ctx)
700 0 : .await?;
701 0 : Ok(plan
702 0 : .into_iter()
703 0 : .flat_map(|read| read.blobs_at)
704 0 : .map(|(_, blob_meta)| blob_meta.key)
705 0 : .collect())
706 0 : }
707 : }
708 :
709 : /// A builder object for constructing a new image layer.
710 : ///
711 : /// Usage:
712 : ///
713 : /// 1. Create the ImageLayerWriter by calling ImageLayerWriter::new(...)
714 : ///
715 : /// 2. Write the contents by calling `put_page_image` for every key-value
716 : /// pair in the key range.
717 : ///
718 : /// 3. Call `finish`.
719 : ///
720 : struct ImageLayerWriterInner {
721 : conf: &'static PageServerConf,
722 : path: Utf8PathBuf,
723 : timeline_id: TimelineId,
724 : tenant_shard_id: TenantShardId,
725 : key_range: Range<Key>,
726 : lsn: Lsn,
727 :
728 : // Total uncompressed bytes passed into put_image
729 : uncompressed_bytes: u64,
730 :
731 : // Like `uncompressed_bytes`,
732 : // but only of images we might consider for compression
733 : uncompressed_bytes_eligible: u64,
734 :
735 : // Like `uncompressed_bytes`, but only of images
736 : // where we have chosen their compressed form
737 : uncompressed_bytes_chosen: u64,
738 :
739 : // Number of keys in the layer.
740 : num_keys: usize,
741 :
742 : blob_writer: BlobWriter<false>,
743 : tree: DiskBtreeBuilder<BlockBuf, KEY_SIZE>,
744 :
745 : #[cfg(feature = "testing")]
746 : last_written_key: Key,
747 : }
748 :
749 : impl ImageLayerWriterInner {
750 : ///
751 : /// Start building a new image layer.
752 : ///
753 : #[allow(clippy::too_many_arguments)]
754 3744 : async fn new(
755 3744 : conf: &'static PageServerConf,
756 3744 : timeline_id: TimelineId,
757 3744 : tenant_shard_id: TenantShardId,
758 3744 : key_range: &Range<Key>,
759 3744 : lsn: Lsn,
760 3744 : gate: &utils::sync::gate::Gate,
761 3744 : cancel: CancellationToken,
762 3744 : ctx: &RequestContext,
763 3744 : ) -> anyhow::Result<Self> {
764 3744 : // Create the file initially with a temporary filename.
765 3744 : // We'll atomically rename it to the final name when we're done.
766 3744 : let path = ImageLayer::temp_path_for(
767 3744 : conf,
768 3744 : timeline_id,
769 3744 : tenant_shard_id,
770 3744 : &ImageLayerName {
771 3744 : key_range: key_range.clone(),
772 3744 : lsn,
773 3744 : },
774 3744 : );
775 3744 : trace!("creating image layer {}", path);
776 3744 : let mut file = {
777 3744 : VirtualFile::open_with_options(
778 3744 : &path,
779 3744 : virtual_file::OpenOptions::new()
780 3744 : .write(true)
781 3744 : .create_new(true),
782 3744 : ctx,
783 3744 : )
784 3744 : .await?
785 : };
786 : // make room for the header block
787 3744 : file.seek(SeekFrom::Start(PAGE_SZ as u64)).await?;
788 3744 : let blob_writer = BlobWriter::new(file, PAGE_SZ as u64, gate, cancel, ctx);
789 3744 :
790 3744 : // Initialize the b-tree index builder
791 3744 : let block_buf = BlockBuf::new();
792 3744 : let tree_builder = DiskBtreeBuilder::new(block_buf);
793 3744 :
794 3744 : let writer = Self {
795 3744 : conf,
796 3744 : path,
797 3744 : timeline_id,
798 3744 : tenant_shard_id,
799 3744 : key_range: key_range.clone(),
800 3744 : lsn,
801 3744 : tree: tree_builder,
802 3744 : blob_writer,
803 3744 : uncompressed_bytes: 0,
804 3744 : uncompressed_bytes_eligible: 0,
805 3744 : uncompressed_bytes_chosen: 0,
806 3744 : num_keys: 0,
807 3744 : #[cfg(feature = "testing")]
808 3744 : last_written_key: Key::MIN,
809 3744 : };
810 3744 :
811 3744 : Ok(writer)
812 3744 : }
813 :
814 : ///
815 : /// Write next value to the file.
816 : ///
817 : /// The page versions must be appended in blknum order.
818 : ///
819 234828 : async fn put_image(
820 234828 : &mut self,
821 234828 : key: Key,
822 234828 : img: Bytes,
823 234828 : ctx: &RequestContext,
824 234828 : ) -> anyhow::Result<()> {
825 234828 : ensure!(self.key_range.contains(&key));
826 234828 : let compression = self.conf.image_compression;
827 234828 : let uncompressed_len = img.len() as u64;
828 234828 : self.uncompressed_bytes += uncompressed_len;
829 234828 : self.num_keys += 1;
830 234828 : let (_img, res) = self
831 234828 : .blob_writer
832 234828 : .write_blob_maybe_compressed(img.slice_len(), ctx, compression)
833 234828 : .await;
834 : // TODO: re-use the buffer for `img` further upstack
835 234828 : let (off, compression_info) = res?;
836 234828 : if compression_info.compressed_size.is_some() {
837 48012 : // The image has been considered for compression at least
838 48012 : self.uncompressed_bytes_eligible += uncompressed_len;
839 186816 : }
840 234828 : if compression_info.written_compressed {
841 0 : // The image has been compressed
842 0 : self.uncompressed_bytes_chosen += uncompressed_len;
843 234828 : }
844 :
845 234828 : let mut keybuf: [u8; KEY_SIZE] = [0u8; KEY_SIZE];
846 234828 : key.write_to_byte_slice(&mut keybuf);
847 234828 : self.tree.append(&keybuf, off)?;
848 :
849 : #[cfg(feature = "testing")]
850 234828 : {
851 234828 : self.last_written_key = key;
852 234828 : }
853 234828 :
854 234828 : Ok(())
855 234828 : }
856 :
857 : ///
858 : /// Write the next image to the file, as a raw blob header and data.
859 : ///
860 : /// The page versions must be appended in blknum order.
861 : ///
862 98304 : async fn put_image_raw(
863 98304 : &mut self,
864 98304 : key: Key,
865 98304 : raw_with_header: Bytes,
866 98304 : ctx: &RequestContext,
867 98304 : ) -> anyhow::Result<()> {
868 98304 : ensure!(self.key_range.contains(&key));
869 :
870 : // NB: we don't update the (un)compressed metrics, since we can't determine them without
871 : // decompressing the image. This seems okay.
872 98304 : self.num_keys += 1;
873 :
874 98304 : let (_, res) = self
875 98304 : .blob_writer
876 98304 : .write_blob_raw(raw_with_header.slice_len(), ctx)
877 98304 : .await;
878 98304 : let offset = res?;
879 :
880 98304 : let mut keybuf: [u8; KEY_SIZE] = [0u8; KEY_SIZE];
881 98304 : key.write_to_byte_slice(&mut keybuf);
882 98304 : self.tree.append(&keybuf, offset)?;
883 :
884 : #[cfg(feature = "testing")]
885 98304 : {
886 98304 : self.last_written_key = key;
887 98304 : }
888 98304 :
889 98304 : Ok(())
890 98304 : }
891 :
892 : ///
893 : /// Finish writing the image layer.
894 : ///
895 2280 : async fn finish(
896 2280 : self,
897 2280 : ctx: &RequestContext,
898 2280 : end_key: Option<Key>,
899 2280 : ) -> anyhow::Result<(PersistentLayerDesc, Utf8PathBuf)> {
900 2280 : let temp_path = self.path.clone();
901 2280 : let result = self.finish0(ctx, end_key).await;
902 2280 : if let Err(ref e) = result {
903 0 : tracing::info!(%temp_path, "cleaning up temporary file after error during writing: {e}");
904 0 : if let Err(e) = std::fs::remove_file(&temp_path) {
905 0 : tracing::warn!(error=%e, %temp_path, "error cleaning up temporary layer file after error during writing");
906 0 : }
907 2280 : }
908 2280 : result
909 2280 : }
910 :
911 : ///
912 : /// Finish writing the image layer.
913 : ///
914 2280 : async fn finish0(
915 2280 : self,
916 2280 : ctx: &RequestContext,
917 2280 : end_key: Option<Key>,
918 2280 : ) -> anyhow::Result<(PersistentLayerDesc, Utf8PathBuf)> {
919 2280 : let index_start_blk = self.blob_writer.size().div_ceil(PAGE_SZ as u64) as u32;
920 2280 :
921 2280 : // Calculate compression ratio
922 2280 : let compressed_size = self.blob_writer.size() - PAGE_SZ as u64; // Subtract PAGE_SZ for header
923 2280 : crate::metrics::COMPRESSION_IMAGE_INPUT_BYTES.inc_by(self.uncompressed_bytes);
924 2280 : crate::metrics::COMPRESSION_IMAGE_INPUT_BYTES_CONSIDERED
925 2280 : .inc_by(self.uncompressed_bytes_eligible);
926 2280 : crate::metrics::COMPRESSION_IMAGE_INPUT_BYTES_CHOSEN.inc_by(self.uncompressed_bytes_chosen);
927 2280 :
928 2280 : // NB: filter() may pass through raw pages from a different layer, without looking at
929 2280 : // whether these are compressed or not. We don't track metrics for these, so avoid
930 2280 : // increasing `COMPRESSION_IMAGE_OUTPUT_BYTES` in this case too.
931 2280 : if self.uncompressed_bytes > 0 {
932 2244 : crate::metrics::COMPRESSION_IMAGE_OUTPUT_BYTES.inc_by(compressed_size);
933 2244 : };
934 :
935 2280 : let mut file = self.blob_writer.into_inner();
936 2280 :
937 2280 : // Write out the index
938 2280 : file.seek(SeekFrom::Start(index_start_blk as u64 * PAGE_SZ as u64))
939 2280 : .await?;
940 2280 : let (index_root_blk, block_buf) = self.tree.finish()?;
941 4752 : for buf in block_buf.blocks {
942 2472 : let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
943 2472 : res?;
944 : }
945 :
946 2280 : let final_key_range = if let Some(end_key) = end_key {
947 1788 : self.key_range.start..end_key
948 : } else {
949 492 : self.key_range.clone()
950 : };
951 :
952 : // Fill in the summary on blk 0
953 2280 : let summary = Summary {
954 2280 : magic: IMAGE_FILE_MAGIC,
955 2280 : format_version: STORAGE_FORMAT_VERSION,
956 2280 : tenant_id: self.tenant_shard_id.tenant_id,
957 2280 : timeline_id: self.timeline_id,
958 2280 : key_range: final_key_range.clone(),
959 2280 : lsn: self.lsn,
960 2280 : index_start_blk,
961 2280 : index_root_blk,
962 2280 : };
963 2280 :
964 2280 : let mut buf = Vec::with_capacity(PAGE_SZ);
965 2280 : // TODO: could use smallvec here but it's a pain with Slice<T>
966 2280 : Summary::ser_into(&summary, &mut buf)?;
967 2280 : file.seek(SeekFrom::Start(0)).await?;
968 2280 : let (_buf, res) = file.write_all(buf.slice_len(), ctx).await;
969 2280 : res?;
970 :
971 2280 : let metadata = file
972 2280 : .metadata()
973 2280 : .await
974 2280 : .context("get metadata to determine file size")?;
975 :
976 2280 : let desc = PersistentLayerDesc::new_img(
977 2280 : self.tenant_shard_id,
978 2280 : self.timeline_id,
979 2280 : final_key_range,
980 2280 : self.lsn,
981 2280 : metadata.len(),
982 2280 : );
983 :
984 : #[cfg(feature = "testing")]
985 2280 : if let Some(end_key) = end_key {
986 1788 : assert!(
987 1788 : self.last_written_key < end_key,
988 0 : "written key violates end_key range"
989 : );
990 492 : }
991 :
992 : // Note: Because we open the file in write-only mode, we cannot
993 : // reuse the same VirtualFile for reading later. That's why we don't
994 : // set inner.file here. The first read will have to re-open it.
995 :
996 : // fsync the file
997 2280 : file.sync_all()
998 2280 : .await
999 2280 : .maybe_fatal_err("image_layer sync_all")?;
1000 :
1001 2280 : trace!("created image layer {}", self.path);
1002 :
1003 2280 : Ok((desc, self.path))
1004 2280 : }
1005 : }
1006 :
1007 : /// A builder object for constructing a new image layer.
1008 : ///
1009 : /// Usage:
1010 : ///
1011 : /// 1. Create the ImageLayerWriter by calling ImageLayerWriter::new(...)
1012 : ///
1013 : /// 2. Write the contents by calling `put_page_image` for every key-value
1014 : /// pair in the key range.
1015 : ///
1016 : /// 3. Call `finish`.
1017 : ///
1018 : /// # Note
1019 : ///
1020 : /// As described in <https://github.com/neondatabase/neon/issues/2650>, it's
1021 : /// possible for the writer to drop before `finish` is actually called. So this
1022 : /// could lead to odd temporary files in the directory, exhausting file system.
1023 : /// This structure wraps `ImageLayerWriterInner` and also contains `Drop`
1024 : /// implementation that cleans up the temporary file in failure. It's not
1025 : /// possible to do this directly in `ImageLayerWriterInner` since `finish` moves
1026 : /// out some fields, making it impossible to implement `Drop`.
1027 : ///
1028 : #[must_use]
1029 : pub struct ImageLayerWriter {
1030 : inner: Option<ImageLayerWriterInner>,
1031 : }
1032 :
1033 : impl ImageLayerWriter {
1034 : ///
1035 : /// Start building a new image layer.
1036 : ///
1037 : #[allow(clippy::too_many_arguments)]
1038 3744 : pub async fn new(
1039 3744 : conf: &'static PageServerConf,
1040 3744 : timeline_id: TimelineId,
1041 3744 : tenant_shard_id: TenantShardId,
1042 3744 : key_range: &Range<Key>,
1043 3744 : lsn: Lsn,
1044 3744 : gate: &utils::sync::gate::Gate,
1045 3744 : cancel: CancellationToken,
1046 3744 : ctx: &RequestContext,
1047 3744 : ) -> anyhow::Result<ImageLayerWriter> {
1048 3744 : Ok(Self {
1049 3744 : inner: Some(
1050 3744 : ImageLayerWriterInner::new(
1051 3744 : conf,
1052 3744 : timeline_id,
1053 3744 : tenant_shard_id,
1054 3744 : key_range,
1055 3744 : lsn,
1056 3744 : gate,
1057 3744 : cancel,
1058 3744 : ctx,
1059 3744 : )
1060 3744 : .await?,
1061 : ),
1062 : })
1063 3744 : }
1064 :
1065 : ///
1066 : /// Write next value to the file.
1067 : ///
1068 : /// The page versions must be appended in blknum order.
1069 : ///
1070 234828 : pub async fn put_image(
1071 234828 : &mut self,
1072 234828 : key: Key,
1073 234828 : img: Bytes,
1074 234828 : ctx: &RequestContext,
1075 234828 : ) -> anyhow::Result<()> {
1076 234828 : self.inner.as_mut().unwrap().put_image(key, img, ctx).await
1077 234828 : }
1078 :
1079 : ///
1080 : /// Write the next value to the file, as a raw header and data. This allows passing through a
1081 : /// raw, potentially compressed image from a different layer file without recompressing it.
1082 : ///
1083 : /// The page versions must be appended in blknum order.
1084 : ///
1085 98304 : pub async fn put_image_raw(
1086 98304 : &mut self,
1087 98304 : key: Key,
1088 98304 : raw_with_header: Bytes,
1089 98304 : ctx: &RequestContext,
1090 98304 : ) -> anyhow::Result<()> {
1091 98304 : self.inner
1092 98304 : .as_mut()
1093 98304 : .unwrap()
1094 98304 : .put_image_raw(key, raw_with_header, ctx)
1095 98304 : .await
1096 98304 : }
1097 :
1098 : /// Estimated size of the image layer.
1099 51324 : pub(crate) fn estimated_size(&self) -> u64 {
1100 51324 : let inner = self.inner.as_ref().unwrap();
1101 51324 : inner.blob_writer.size() + inner.tree.borrow_writer().size() + PAGE_SZ as u64
1102 51324 : }
1103 :
1104 51912 : pub(crate) fn num_keys(&self) -> usize {
1105 51912 : self.inner.as_ref().unwrap().num_keys
1106 51912 : }
1107 :
1108 : ///
1109 : /// Finish writing the image layer.
1110 : ///
1111 492 : pub(crate) async fn finish(
1112 492 : mut self,
1113 492 : ctx: &RequestContext,
1114 492 : ) -> anyhow::Result<(PersistentLayerDesc, Utf8PathBuf)> {
1115 492 : self.inner.take().unwrap().finish(ctx, None).await
1116 492 : }
1117 :
1118 : /// Finish writing the image layer with an end key, used in [`super::batch_split_writer::SplitImageLayerWriter`]. The end key determines the end of the image layer's covered range and is exclusive.
1119 1788 : pub(super) async fn finish_with_end_key(
1120 1788 : mut self,
1121 1788 : end_key: Key,
1122 1788 : ctx: &RequestContext,
1123 1788 : ) -> anyhow::Result<(PersistentLayerDesc, Utf8PathBuf)> {
1124 1788 : self.inner.take().unwrap().finish(ctx, Some(end_key)).await
1125 1788 : }
1126 : }
1127 :
1128 : impl Drop for ImageLayerWriter {
1129 3744 : fn drop(&mut self) {
1130 3744 : if let Some(inner) = self.inner.take() {
1131 1464 : inner.blob_writer.into_inner().remove();
1132 2280 : }
1133 3744 : }
1134 : }
1135 :
1136 : pub struct ImageLayerIterator<'a> {
1137 : image_layer: &'a ImageLayerInner,
1138 : ctx: &'a RequestContext,
1139 : planner: StreamingVectoredReadPlanner,
1140 : index_iter: DiskBtreeIterator<'a>,
1141 : key_values_batch: VecDeque<(Key, Lsn, Value)>,
1142 : is_end: bool,
1143 : }
1144 :
1145 : impl ImageLayerIterator<'_> {
1146 0 : pub(crate) fn layer_dbg_info(&self) -> String {
1147 0 : self.image_layer.layer_dbg_info()
1148 0 : }
1149 :
1150 : /// Retrieve a batch of key-value pairs into the iterator buffer.
1151 114780 : async fn next_batch(&mut self) -> anyhow::Result<()> {
1152 114780 : assert!(self.key_values_batch.is_empty());
1153 114780 : assert!(!self.is_end);
1154 :
1155 114780 : let plan = loop {
1156 174444 : if let Some(res) = self.index_iter.next().await {
1157 173856 : let (raw_key, offset) = res?;
1158 173856 : if let Some(batch_plan) = self.planner.handle(
1159 173856 : Key::from_slice(&raw_key[..KEY_SIZE]),
1160 173856 : self.image_layer.lsn,
1161 173856 : offset,
1162 173856 : true,
1163 173856 : ) {
1164 114192 : break batch_plan;
1165 59664 : }
1166 : } else {
1167 588 : self.is_end = true;
1168 588 : let payload_end = self.image_layer.index_start_blk as u64 * PAGE_SZ as u64;
1169 588 : if let Some(item) = self.planner.handle_range_end(payload_end) {
1170 588 : break item;
1171 : } else {
1172 0 : return Ok(()); // TODO: a test case on empty iterator
1173 : }
1174 : }
1175 : };
1176 114780 : let vectored_blob_reader = VectoredBlobReader::new(&self.image_layer.file);
1177 114780 : let mut next_batch = std::collections::VecDeque::new();
1178 114780 : let buf_size = plan.size();
1179 114780 : let buf = IoBufferMut::with_capacity(buf_size);
1180 114780 : let blobs_buf = vectored_blob_reader
1181 114780 : .read_blobs(&plan, buf, self.ctx)
1182 114780 : .await?;
1183 114780 : let view = BufView::new_slice(&blobs_buf.buf);
1184 173688 : for meta in blobs_buf.blobs.iter() {
1185 173688 : let img_buf = meta.read(&view).await?;
1186 173688 : next_batch.push_back((
1187 173688 : meta.meta.key,
1188 173688 : self.image_layer.lsn,
1189 173688 : Value::Image(img_buf.into_bytes()),
1190 173688 : ));
1191 : }
1192 114780 : self.key_values_batch = next_batch;
1193 114780 : Ok(())
1194 114780 : }
1195 :
1196 172968 : pub async fn next(&mut self) -> anyhow::Result<Option<(Key, Lsn, Value)>> {
1197 172968 : if self.key_values_batch.is_empty() {
1198 115248 : if self.is_end {
1199 972 : return Ok(None);
1200 114276 : }
1201 114276 : self.next_batch().await?;
1202 57720 : }
1203 171996 : Ok(Some(
1204 171996 : self.key_values_batch
1205 171996 : .pop_front()
1206 171996 : .expect("should not be empty"),
1207 171996 : ))
1208 172968 : }
1209 : }
1210 :
1211 : #[cfg(test)]
1212 : mod test {
1213 : use std::sync::Arc;
1214 : use std::time::Duration;
1215 :
1216 : use bytes::Bytes;
1217 : use itertools::Itertools;
1218 : use pageserver_api::key::Key;
1219 : use pageserver_api::shard::{ShardCount, ShardIdentity, ShardNumber, ShardStripeSize};
1220 : use pageserver_api::value::Value;
1221 : use utils::generation::Generation;
1222 : use utils::id::{TenantId, TimelineId};
1223 : use utils::lsn::Lsn;
1224 :
1225 : use super::{ImageLayerIterator, ImageLayerWriter};
1226 : use crate::DEFAULT_PG_VERSION;
1227 : use crate::context::RequestContext;
1228 : use crate::tenant::harness::{TIMELINE_ID, TenantHarness};
1229 : use crate::tenant::storage_layer::{Layer, ResidentLayer};
1230 : use crate::tenant::vectored_blob_io::StreamingVectoredReadPlanner;
1231 : use crate::tenant::{TenantShard, Timeline};
1232 :
1233 : #[tokio::test]
1234 12 : async fn image_layer_rewrite() {
1235 12 : let tenant_conf = pageserver_api::models::TenantConfig {
1236 12 : gc_period: Some(Duration::ZERO),
1237 12 : compaction_period: Some(Duration::ZERO),
1238 12 : ..Default::default()
1239 12 : };
1240 12 : let tenant_id = TenantId::generate();
1241 12 : let mut gen_ = Generation::new(0xdead0001);
1242 60 : let mut get_next_gen = || {
1243 60 : let ret = gen_;
1244 60 : gen_ = gen_.next();
1245 60 : ret
1246 60 : };
1247 12 : // The LSN at which we will create an image layer to filter
1248 12 : let lsn = Lsn(0xdeadbeef0000);
1249 12 : let timeline_id = TimelineId::generate();
1250 12 :
1251 12 : //
1252 12 : // Create an unsharded parent with a layer.
1253 12 : //
1254 12 :
1255 12 : let harness = TenantHarness::create_custom(
1256 12 : "test_image_layer_rewrite--parent",
1257 12 : tenant_conf.clone(),
1258 12 : tenant_id,
1259 12 : ShardIdentity::unsharded(),
1260 12 : get_next_gen(),
1261 12 : )
1262 12 : .await
1263 12 : .unwrap();
1264 12 : let (tenant, ctx) = harness.load().await;
1265 12 : let timeline = tenant
1266 12 : .create_test_timeline(timeline_id, lsn, DEFAULT_PG_VERSION, &ctx)
1267 12 : .await
1268 12 : .unwrap();
1269 12 :
1270 12 : // This key range contains several 0x8000 page stripes, only one of which belongs to shard zero
1271 12 : let input_start = Key::from_hex("000000067f00000001000000ae0000000000").unwrap();
1272 12 : let input_end = Key::from_hex("000000067f00000001000000ae0000002000").unwrap();
1273 12 : let range = input_start..input_end;
1274 12 :
1275 12 : // Build an image layer to filter
1276 12 : let resident = {
1277 12 : let mut writer = ImageLayerWriter::new(
1278 12 : harness.conf,
1279 12 : timeline_id,
1280 12 : harness.tenant_shard_id,
1281 12 : &range,
1282 12 : lsn,
1283 12 : &timeline.gate,
1284 12 : timeline.cancel.clone(),
1285 12 : &ctx,
1286 12 : )
1287 12 : .await
1288 12 : .unwrap();
1289 12 :
1290 12 : let foo_img = Bytes::from_static(&[1, 2, 3, 4]);
1291 12 : let mut key = range.start;
1292 98316 : while key < range.end {
1293 98304 : writer.put_image(key, foo_img.clone(), &ctx).await.unwrap();
1294 98304 :
1295 98304 : key = key.next();
1296 12 : }
1297 12 : let (desc, path) = writer.finish(&ctx).await.unwrap();
1298 12 : Layer::finish_creating(tenant.conf, &timeline, desc, &path).unwrap()
1299 12 : };
1300 12 : let original_size = resident.metadata().file_size;
1301 12 :
1302 12 : //
1303 12 : // Create child shards and do the rewrite, exercising filter().
1304 12 : // TODO: abstraction in TenantHarness for splits.
1305 12 : //
1306 12 :
1307 12 : // Filter for various shards: this exercises cases like values at start of key range, end of key
1308 12 : // range, middle of key range.
1309 12 : let shard_count = ShardCount::new(4);
1310 48 : for shard_number in 0..shard_count.count() {
1311 12 : //
1312 12 : // mimic the shard split
1313 12 : //
1314 48 : let shard_identity = ShardIdentity::new(
1315 48 : ShardNumber(shard_number),
1316 48 : shard_count,
1317 48 : ShardStripeSize(0x800),
1318 48 : )
1319 48 : .unwrap();
1320 48 : let harness = TenantHarness::create_custom(
1321 48 : Box::leak(Box::new(format!(
1322 48 : "test_image_layer_rewrite--child{}",
1323 48 : shard_identity.shard_slug()
1324 48 : ))),
1325 48 : tenant_conf.clone(),
1326 48 : tenant_id,
1327 48 : shard_identity,
1328 48 : // NB: in reality, the shards would each fork off their own gen number sequence from the parent.
1329 48 : // But here, all we care about is that the gen number is unique.
1330 48 : get_next_gen(),
1331 48 : )
1332 48 : .await
1333 48 : .unwrap();
1334 48 : let (tenant, ctx) = harness.load().await;
1335 48 : let timeline = tenant
1336 48 : .create_test_timeline(timeline_id, lsn, DEFAULT_PG_VERSION, &ctx)
1337 48 : .await
1338 48 : .unwrap();
1339 12 :
1340 12 : //
1341 12 : // use filter() and make assertions
1342 12 : //
1343 12 :
1344 48 : let mut filtered_writer = ImageLayerWriter::new(
1345 48 : harness.conf,
1346 48 : timeline_id,
1347 48 : harness.tenant_shard_id,
1348 48 : &range,
1349 48 : lsn,
1350 48 : &timeline.gate,
1351 48 : timeline.cancel.clone(),
1352 48 : &ctx,
1353 48 : )
1354 48 : .await
1355 48 : .unwrap();
1356 12 :
1357 48 : let wrote_keys = resident
1358 48 : .filter(&shard_identity, &mut filtered_writer, &ctx)
1359 48 : .await
1360 48 : .unwrap();
1361 48 : let replacement = if wrote_keys > 0 {
1362 36 : let (desc, path) = filtered_writer.finish(&ctx).await.unwrap();
1363 36 : let resident = Layer::finish_creating(tenant.conf, &timeline, desc, &path).unwrap();
1364 36 : Some(resident)
1365 12 : } else {
1366 12 : None
1367 12 : };
1368 12 :
1369 12 : // This exact size and those below will need updating as/when the layer encoding changes, but
1370 12 : // should be deterministic for a given version of the format, as we used no randomness generating the input.
1371 48 : assert_eq!(original_size, 122880);
1372 12 :
1373 48 : match shard_number {
1374 12 : 0 => {
1375 12 : // We should have written out just one stripe for our shard identity
1376 12 : assert_eq!(wrote_keys, 0x800);
1377 12 : let replacement = replacement.unwrap();
1378 12 :
1379 12 : // We should have dropped some of the data
1380 12 : assert!(replacement.metadata().file_size < original_size);
1381 12 : assert!(replacement.metadata().file_size > 0);
1382 12 :
1383 12 : // Assert that we dropped ~3/4 of the data.
1384 12 : assert_eq!(replacement.metadata().file_size, 49152);
1385 12 : }
1386 12 : 1 => {
1387 12 : // Shard 1 has no keys in our input range
1388 12 : assert_eq!(wrote_keys, 0x0);
1389 12 : assert!(replacement.is_none());
1390 12 : }
1391 12 : 2 => {
1392 12 : // Shard 2 has one stripes in the input range
1393 12 : assert_eq!(wrote_keys, 0x800);
1394 12 : let replacement = replacement.unwrap();
1395 12 : assert!(replacement.metadata().file_size < original_size);
1396 12 : assert!(replacement.metadata().file_size > 0);
1397 12 : assert_eq!(replacement.metadata().file_size, 49152);
1398 12 : }
1399 12 : 3 => {
1400 12 : // Shard 3 has two stripes in the input range
1401 12 : assert_eq!(wrote_keys, 0x1000);
1402 12 : let replacement = replacement.unwrap();
1403 12 : assert!(replacement.metadata().file_size < original_size);
1404 12 : assert!(replacement.metadata().file_size > 0);
1405 12 : assert_eq!(replacement.metadata().file_size, 73728);
1406 12 : }
1407 12 : _ => unreachable!(),
1408 12 : }
1409 12 : }
1410 12 : }
1411 :
1412 12 : async fn produce_image_layer(
1413 12 : tenant: &TenantShard,
1414 12 : tline: &Arc<Timeline>,
1415 12 : mut images: Vec<(Key, Bytes)>,
1416 12 : lsn: Lsn,
1417 12 : ctx: &RequestContext,
1418 12 : ) -> anyhow::Result<ResidentLayer> {
1419 12 : images.sort();
1420 12 : let (key_start, _) = images.first().unwrap();
1421 12 : let (key_last, _) = images.last().unwrap();
1422 12 : let key_end = key_last.next();
1423 12 : let key_range = *key_start..key_end;
1424 12 : let mut writer = ImageLayerWriter::new(
1425 12 : tenant.conf,
1426 12 : tline.timeline_id,
1427 12 : tenant.tenant_shard_id,
1428 12 : &key_range,
1429 12 : lsn,
1430 12 : &tline.gate,
1431 12 : tline.cancel.clone(),
1432 12 : ctx,
1433 12 : )
1434 12 : .await?;
1435 :
1436 12012 : for (key, img) in images {
1437 12000 : writer.put_image(key, img, ctx).await?;
1438 : }
1439 12 : let (desc, path) = writer.finish(ctx).await?;
1440 12 : let img_layer = Layer::finish_creating(tenant.conf, tline, desc, &path)?;
1441 :
1442 12 : Ok::<_, anyhow::Error>(img_layer)
1443 12 : }
1444 :
1445 168 : async fn assert_img_iter_equal(
1446 168 : img_iter: &mut ImageLayerIterator<'_>,
1447 168 : expect: &[(Key, Bytes)],
1448 168 : expect_lsn: Lsn,
1449 168 : ) {
1450 168 : let mut expect_iter = expect.iter();
1451 : loop {
1452 168168 : let o1 = img_iter.next().await.unwrap();
1453 168168 : let o2 = expect_iter.next();
1454 168168 : match (o1, o2) {
1455 168 : (None, None) => break,
1456 168000 : (Some((k1, l1, v1)), Some((k2, i2))) => {
1457 168000 : let Value::Image(i1) = v1 else {
1458 0 : panic!("expect Value::Image")
1459 : };
1460 168000 : assert_eq!(&k1, k2);
1461 168000 : assert_eq!(l1, expect_lsn);
1462 168000 : assert_eq!(&i1, i2);
1463 : }
1464 0 : (o1, o2) => panic!("iterators length mismatch: {:?}, {:?}", o1, o2),
1465 : }
1466 : }
1467 168 : }
1468 :
1469 : #[tokio::test]
1470 12 : async fn image_layer_iterator() {
1471 12 : let harness = TenantHarness::create("image_layer_iterator").await.unwrap();
1472 12 : let (tenant, ctx) = harness.load().await;
1473 12 :
1474 12 : let tline = tenant
1475 12 : .create_test_timeline(TIMELINE_ID, Lsn(0x10), DEFAULT_PG_VERSION, &ctx)
1476 12 : .await
1477 12 : .unwrap();
1478 12 :
1479 12000 : fn get_key(id: u32) -> Key {
1480 12000 : let mut key = Key::from_hex("000000000033333333444444445500000000").unwrap();
1481 12000 : key.field6 = id;
1482 12000 : key
1483 12000 : }
1484 12 : const N: usize = 1000;
1485 12 : let test_imgs = (0..N)
1486 12000 : .map(|idx| (get_key(idx as u32), Bytes::from(format!("img{idx:05}"))))
1487 12 : .collect_vec();
1488 12 : let resident_layer =
1489 12 : produce_image_layer(&tenant, &tline, test_imgs.clone(), Lsn(0x10), &ctx)
1490 12 : .await
1491 12 : .unwrap();
1492 12 : let img_layer = resident_layer.get_as_image(&ctx).await.unwrap();
1493 36 : for max_read_size in [1, 1024] {
1494 192 : for batch_size in [1, 2, 4, 8, 3, 7, 13] {
1495 168 : println!("running with batch_size={batch_size} max_read_size={max_read_size}");
1496 168 : // Test if the batch size is correctly determined
1497 168 : let mut iter = img_layer.iter(&ctx);
1498 168 : iter.planner = StreamingVectoredReadPlanner::new(max_read_size, batch_size);
1499 168 : let mut num_items = 0;
1500 672 : for _ in 0..3 {
1501 504 : iter.next_batch().await.unwrap();
1502 504 : num_items += iter.key_values_batch.len();
1503 504 : if max_read_size == 1 {
1504 12 : // every key should be a batch b/c the value is larger than max_read_size
1505 252 : assert_eq!(iter.key_values_batch.len(), 1);
1506 12 : } else {
1507 252 : assert!(iter.key_values_batch.len() <= batch_size);
1508 12 : }
1509 504 : if num_items >= N {
1510 12 : break;
1511 504 : }
1512 504 : iter.key_values_batch.clear();
1513 12 : }
1514 12 : // Test if the result is correct
1515 168 : let mut iter = img_layer.iter(&ctx);
1516 168 : iter.planner = StreamingVectoredReadPlanner::new(max_read_size, batch_size);
1517 168 : assert_img_iter_equal(&mut iter, &test_imgs, Lsn(0x10)).await;
1518 12 : }
1519 12 : }
1520 12 : }
1521 : }
|
