Line data Source code
1 : //!
2 : //! Parse PostgreSQL WAL records and store them in a neon Timeline.
3 : //!
4 : //! The pipeline for ingesting WAL looks like this:
5 : //!
6 : //! WAL receiver -> [`wal_decoder`] -> WalIngest -> Repository
7 : //!
8 : //! The WAL receiver receives a stream of WAL from the WAL safekeepers.
9 : //! Records get decoded and interpreted in the [`wal_decoder`] module
10 : //! and then stored to the Repository by WalIngest.
11 : //!
12 : //! The neon Repository can store page versions in two formats: as
13 : //! page images, or a WAL records. [`wal_decoder::models::InterpretedWalRecord::from_bytes_filtered`]
14 : //! extracts page images out of some WAL records, but mostly it's WAL
15 : //! records. If a WAL record modifies multiple pages, WalIngest
16 : //! will call Repository::put_rel_wal_record or put_rel_page_image functions
17 : //! separately for each modified page.
18 : //!
19 : //! To reconstruct a page using a WAL record, the Repository calls the
20 : //! code in walredo.rs. walredo.rs passes most WAL records to the WAL
21 : //! redo Postgres process, but some records it can handle directly with
22 : //! bespoken Rust code.
23 :
24 : use std::collections::HashMap;
25 : use std::sync::Arc;
26 : use std::sync::OnceLock;
27 : use std::time::Duration;
28 : use std::time::Instant;
29 : use std::time::SystemTime;
30 :
31 : use pageserver_api::shard::ShardIdentity;
32 : use postgres_ffi::fsm_logical_to_physical;
33 : use postgres_ffi::walrecord::*;
34 : use postgres_ffi::{dispatch_pgversion, enum_pgversion, enum_pgversion_dispatch, TimestampTz};
35 : use wal_decoder::models::*;
36 :
37 : use anyhow::{bail, Result};
38 : use bytes::{Buf, Bytes};
39 : use tracing::*;
40 : use utils::failpoint_support;
41 : use utils::rate_limit::RateLimit;
42 :
43 : use crate::context::RequestContext;
44 : use crate::metrics::WAL_INGEST;
45 : use crate::pgdatadir_mapping::{DatadirModification, Version};
46 : use crate::span::debug_assert_current_span_has_tenant_and_timeline_id;
47 : use crate::tenant::PageReconstructError;
48 : use crate::tenant::Timeline;
49 : use crate::ZERO_PAGE;
50 : use pageserver_api::key::rel_block_to_key;
51 : use pageserver_api::record::NeonWalRecord;
52 : use pageserver_api::reltag::{BlockNumber, RelTag, SlruKind};
53 : use postgres_ffi::pg_constants;
54 : use postgres_ffi::relfile_utils::{FSM_FORKNUM, INIT_FORKNUM, MAIN_FORKNUM, VISIBILITYMAP_FORKNUM};
55 : use postgres_ffi::TransactionId;
56 : use utils::bin_ser::SerializeError;
57 : use utils::lsn::Lsn;
58 :
59 : enum_pgversion! {CheckPoint, pgv::CheckPoint}
60 :
61 : impl CheckPoint {
62 6 : fn encode(&self) -> Result<Bytes, SerializeError> {
63 6 : enum_pgversion_dispatch!(self, CheckPoint, cp, { cp.encode() })
64 6 : }
65 :
66 145834 : fn update_next_xid(&mut self, xid: u32) -> bool {
67 145834 : enum_pgversion_dispatch!(self, CheckPoint, cp, { cp.update_next_xid(xid) })
68 145834 : }
69 :
70 0 : pub fn update_next_multixid(&mut self, multi_xid: u32, multi_offset: u32) -> bool {
71 0 : enum_pgversion_dispatch!(self, CheckPoint, cp, {
72 0 : cp.update_next_multixid(multi_xid, multi_offset)
73 : })
74 0 : }
75 : }
76 :
77 : /// Temporary limitation of WAL lag warnings after attach
78 : ///
79 : /// After tenant attach, we want to limit WAL lag warnings because
80 : /// we don't look at the WAL until the attach is complete, which
81 : /// might take a while.
82 : pub struct WalLagCooldown {
83 : /// Until when should this limitation apply at all
84 : active_until: std::time::Instant,
85 : /// The maximum lag to suppress. Lags above this limit get reported anyways.
86 : max_lag: Duration,
87 : }
88 :
89 : impl WalLagCooldown {
90 0 : pub fn new(attach_start: Instant, attach_duration: Duration) -> Self {
91 0 : Self {
92 0 : active_until: attach_start + attach_duration * 3 + Duration::from_secs(120),
93 0 : max_lag: attach_duration * 2 + Duration::from_secs(60),
94 0 : }
95 0 : }
96 : }
97 :
98 : pub struct WalIngest {
99 : attach_wal_lag_cooldown: Arc<OnceLock<WalLagCooldown>>,
100 : shard: ShardIdentity,
101 : checkpoint: CheckPoint,
102 : checkpoint_modified: bool,
103 : warn_ingest_lag: WarnIngestLag,
104 : }
105 :
106 : struct WarnIngestLag {
107 : lag_msg_ratelimit: RateLimit,
108 : future_lsn_msg_ratelimit: RateLimit,
109 : timestamp_invalid_msg_ratelimit: RateLimit,
110 : }
111 :
112 : impl WalIngest {
113 12 : pub async fn new(
114 12 : timeline: &Timeline,
115 12 : startpoint: Lsn,
116 12 : ctx: &RequestContext,
117 12 : ) -> anyhow::Result<WalIngest> {
118 : // Fetch the latest checkpoint into memory, so that we can compare with it
119 : // quickly in `ingest_record` and update it when it changes.
120 12 : let checkpoint_bytes = timeline.get_checkpoint(startpoint, ctx).await?;
121 12 : let pgversion = timeline.pg_version;
122 :
123 12 : let checkpoint = dispatch_pgversion!(pgversion, {
124 0 : let checkpoint = pgv::CheckPoint::decode(&checkpoint_bytes)?;
125 0 : trace!("CheckPoint.nextXid = {}", checkpoint.nextXid.value);
126 0 : <pgv::CheckPoint as Into<CheckPoint>>::into(checkpoint)
127 : });
128 :
129 12 : Ok(WalIngest {
130 12 : shard: *timeline.get_shard_identity(),
131 12 : checkpoint,
132 12 : checkpoint_modified: false,
133 12 : attach_wal_lag_cooldown: timeline.attach_wal_lag_cooldown.clone(),
134 12 : warn_ingest_lag: WarnIngestLag {
135 12 : lag_msg_ratelimit: RateLimit::new(std::time::Duration::from_secs(10)),
136 12 : future_lsn_msg_ratelimit: RateLimit::new(std::time::Duration::from_secs(10)),
137 12 : timestamp_invalid_msg_ratelimit: RateLimit::new(std::time::Duration::from_secs(10)),
138 12 : },
139 12 : })
140 12 : }
141 :
142 : /// Ingest an interpreted PostgreSQL WAL record by doing writes to the underlying key value
143 : /// storage of a given timeline.
144 : ///
145 : /// This function updates `lsn` field of `DatadirModification`
146 : ///
147 : /// This function returns `true` if the record was ingested, and `false` if it was filtered out
148 145852 : pub async fn ingest_record(
149 145852 : &mut self,
150 145852 : interpreted: InterpretedWalRecord,
151 145852 : modification: &mut DatadirModification<'_>,
152 145852 : ctx: &RequestContext,
153 145852 : ) -> anyhow::Result<bool> {
154 145852 : WAL_INGEST.records_received.inc();
155 145852 : let prev_len = modification.len();
156 145852 :
157 145852 : modification.set_lsn(interpreted.end_lsn)?;
158 :
159 145852 : if matches!(interpreted.flush_uncommitted, FlushUncommittedRecords::Yes) {
160 : // Records of this type should always be preceded by a commit(), as they
161 : // rely on reading data pages back from the Timeline.
162 0 : assert!(!modification.has_dirty_data());
163 145852 : }
164 :
165 145852 : assert!(!self.checkpoint_modified);
166 145852 : if interpreted.xid != pg_constants::INVALID_TRANSACTION_ID
167 145834 : && self.checkpoint.update_next_xid(interpreted.xid)
168 2 : {
169 2 : self.checkpoint_modified = true;
170 145850 : }
171 :
172 145852 : failpoint_support::sleep_millis_async!("wal-ingest-record-sleep");
173 :
174 66 : match interpreted.metadata_record {
175 12 : Some(MetadataRecord::Heapam(rec)) => match rec {
176 12 : HeapamRecord::ClearVmBits(clear_vm_bits) => {
177 12 : self.ingest_clear_vm_bits(clear_vm_bits, modification, ctx)
178 0 : .await?;
179 : }
180 : },
181 0 : Some(MetadataRecord::Neonrmgr(rec)) => match rec {
182 0 : NeonrmgrRecord::ClearVmBits(clear_vm_bits) => {
183 0 : self.ingest_clear_vm_bits(clear_vm_bits, modification, ctx)
184 0 : .await?;
185 : }
186 : },
187 16 : Some(MetadataRecord::Smgr(rec)) => match rec {
188 16 : SmgrRecord::Create(create) => {
189 16 : self.ingest_xlog_smgr_create(create, modification, ctx)
190 12 : .await?;
191 : }
192 0 : SmgrRecord::Truncate(truncate) => {
193 0 : self.ingest_xlog_smgr_truncate(truncate, modification, ctx)
194 0 : .await?;
195 : }
196 : },
197 0 : Some(MetadataRecord::Dbase(rec)) => match rec {
198 0 : DbaseRecord::Create(create) => {
199 0 : self.ingest_xlog_dbase_create(create, modification, ctx)
200 0 : .await?;
201 : }
202 0 : DbaseRecord::Drop(drop) => {
203 0 : self.ingest_xlog_dbase_drop(drop, modification, ctx).await?;
204 : }
205 : },
206 0 : Some(MetadataRecord::Clog(rec)) => match rec {
207 0 : ClogRecord::ZeroPage(zero_page) => {
208 0 : self.ingest_clog_zero_page(zero_page, modification, ctx)
209 0 : .await?;
210 : }
211 0 : ClogRecord::Truncate(truncate) => {
212 0 : self.ingest_clog_truncate(truncate, modification, ctx)
213 0 : .await?;
214 : }
215 : },
216 8 : Some(MetadataRecord::Xact(rec)) => {
217 8 : self.ingest_xact_record(rec, modification, ctx).await?;
218 : }
219 0 : Some(MetadataRecord::MultiXact(rec)) => match rec {
220 0 : MultiXactRecord::ZeroPage(zero_page) => {
221 0 : self.ingest_multixact_zero_page(zero_page, modification, ctx)
222 0 : .await?;
223 : }
224 0 : MultiXactRecord::Create(create) => {
225 0 : self.ingest_multixact_create(modification, &create)?;
226 : }
227 0 : MultiXactRecord::Truncate(truncate) => {
228 0 : self.ingest_multixact_truncate(modification, &truncate, ctx)
229 0 : .await?;
230 : }
231 : },
232 0 : Some(MetadataRecord::Relmap(rec)) => match rec {
233 0 : RelmapRecord::Update(update) => {
234 0 : self.ingest_relmap_update(update, modification, ctx).await?;
235 : }
236 : },
237 30 : Some(MetadataRecord::Xlog(rec)) => match rec {
238 30 : XlogRecord::Raw(raw) => {
239 30 : self.ingest_raw_xlog_record(raw, modification, ctx).await?;
240 : }
241 : },
242 0 : Some(MetadataRecord::LogicalMessage(rec)) => match rec {
243 0 : LogicalMessageRecord::Put(put) => {
244 0 : self.ingest_logical_message_put(put, modification, ctx)
245 0 : .await?;
246 : }
247 : #[cfg(feature = "testing")]
248 : LogicalMessageRecord::Failpoint => {
249 : // This is a convenient way to make the WAL ingestion pause at
250 : // particular point in the WAL. For more fine-grained control,
251 : // we could peek into the message and only pause if it contains
252 : // a particular string, for example, but this is enough for now.
253 0 : failpoint_support::sleep_millis_async!(
254 0 : "pageserver-wal-ingest-logical-message-sleep"
255 0 : );
256 : }
257 : },
258 0 : Some(MetadataRecord::Standby(rec)) => {
259 0 : self.ingest_standby_record(rec).unwrap();
260 0 : }
261 0 : Some(MetadataRecord::Replorigin(rec)) => {
262 0 : self.ingest_replorigin_record(rec, modification).await?;
263 : }
264 145786 : None => {
265 145786 : // There are two cases through which we end up here:
266 145786 : // 1. The resource manager for the original PG WAL record
267 145786 : // is [`pg_constants::RM_TBLSPC_ID`]. This is not a supported
268 145786 : // record type within Neon.
269 145786 : // 2. The resource manager id was unknown to
270 145786 : // [`wal_decoder::decoder::MetadataRecord::from_decoded`].
271 145786 : // TODO(vlad): Tighten this up more once we build confidence
272 145786 : // that case (2) does not happen in the field.
273 145786 : }
274 : }
275 :
276 145852 : modification
277 145852 : .ingest_batch(interpreted.batch, &self.shard, ctx)
278 284 : .await?;
279 :
280 : // If checkpoint data was updated, store the new version in the repository
281 145852 : if self.checkpoint_modified {
282 6 : let new_checkpoint_bytes = self.checkpoint.encode()?;
283 :
284 6 : modification.put_checkpoint(new_checkpoint_bytes)?;
285 6 : self.checkpoint_modified = false;
286 145846 : }
287 :
288 : // Note that at this point this record is only cached in the modification
289 : // until commit() is called to flush the data into the repository and update
290 : // the latest LSN.
291 :
292 145852 : Ok(modification.len() > prev_len)
293 145852 : }
294 :
295 : /// This is the same as AdjustToFullTransactionId(xid) in PostgreSQL
296 0 : fn adjust_to_full_transaction_id(&self, xid: TransactionId) -> Result<u64> {
297 0 : let next_full_xid =
298 0 : enum_pgversion_dispatch!(&self.checkpoint, CheckPoint, cp, { cp.nextXid.value });
299 :
300 0 : let next_xid = (next_full_xid) as u32;
301 0 : let mut epoch = (next_full_xid >> 32) as u32;
302 0 :
303 0 : if xid > next_xid {
304 : // Wraparound occurred, must be from a prev epoch.
305 0 : if epoch == 0 {
306 0 : bail!("apparent XID wraparound with prepared transaction XID {xid}, nextXid is {next_full_xid}");
307 0 : }
308 0 : epoch -= 1;
309 0 : }
310 :
311 0 : Ok((epoch as u64) << 32 | xid as u64)
312 0 : }
313 :
314 12 : async fn ingest_clear_vm_bits(
315 12 : &mut self,
316 12 : clear_vm_bits: ClearVmBits,
317 12 : modification: &mut DatadirModification<'_>,
318 12 : ctx: &RequestContext,
319 12 : ) -> anyhow::Result<()> {
320 12 : let ClearVmBits {
321 12 : new_heap_blkno,
322 12 : old_heap_blkno,
323 12 : flags,
324 12 : vm_rel,
325 12 : } = clear_vm_bits;
326 12 : // Clear the VM bits if required.
327 12 : let mut new_vm_blk = new_heap_blkno.map(pg_constants::HEAPBLK_TO_MAPBLOCK);
328 12 : let mut old_vm_blk = old_heap_blkno.map(pg_constants::HEAPBLK_TO_MAPBLOCK);
329 :
330 : // Sometimes, Postgres seems to create heap WAL records with the
331 : // ALL_VISIBLE_CLEARED flag set, even though the bit in the VM page is
332 : // not set. In fact, it's possible that the VM page does not exist at all.
333 : // In that case, we don't want to store a record to clear the VM bit;
334 : // replaying it would fail to find the previous image of the page, because
335 : // it doesn't exist. So check if the VM page(s) exist, and skip the WAL
336 : // record if it doesn't.
337 12 : let vm_size = get_relsize(modification, vm_rel, ctx).await?;
338 12 : if let Some(blknum) = new_vm_blk {
339 12 : if blknum >= vm_size {
340 0 : new_vm_blk = None;
341 12 : }
342 0 : }
343 12 : if let Some(blknum) = old_vm_blk {
344 0 : if blknum >= vm_size {
345 0 : old_vm_blk = None;
346 0 : }
347 12 : }
348 :
349 12 : if new_vm_blk.is_some() || old_vm_blk.is_some() {
350 12 : if new_vm_blk == old_vm_blk {
351 : // An UPDATE record that needs to clear the bits for both old and the
352 : // new page, both of which reside on the same VM page.
353 0 : self.put_rel_wal_record(
354 0 : modification,
355 0 : vm_rel,
356 0 : new_vm_blk.unwrap(),
357 0 : NeonWalRecord::ClearVisibilityMapFlags {
358 0 : new_heap_blkno,
359 0 : old_heap_blkno,
360 0 : flags,
361 0 : },
362 0 : ctx,
363 0 : )
364 0 : .await?;
365 : } else {
366 : // Clear VM bits for one heap page, or for two pages that reside on
367 : // different VM pages.
368 12 : if let Some(new_vm_blk) = new_vm_blk {
369 12 : self.put_rel_wal_record(
370 12 : modification,
371 12 : vm_rel,
372 12 : new_vm_blk,
373 12 : NeonWalRecord::ClearVisibilityMapFlags {
374 12 : new_heap_blkno,
375 12 : old_heap_blkno: None,
376 12 : flags,
377 12 : },
378 12 : ctx,
379 12 : )
380 0 : .await?;
381 0 : }
382 12 : if let Some(old_vm_blk) = old_vm_blk {
383 0 : self.put_rel_wal_record(
384 0 : modification,
385 0 : vm_rel,
386 0 : old_vm_blk,
387 0 : NeonWalRecord::ClearVisibilityMapFlags {
388 0 : new_heap_blkno: None,
389 0 : old_heap_blkno,
390 0 : flags,
391 0 : },
392 0 : ctx,
393 0 : )
394 0 : .await?;
395 12 : }
396 : }
397 0 : }
398 :
399 12 : Ok(())
400 12 : }
401 :
402 : /// Subroutine of ingest_record(), to handle an XLOG_DBASE_CREATE record.
403 0 : async fn ingest_xlog_dbase_create(
404 0 : &mut self,
405 0 : create: DbaseCreate,
406 0 : modification: &mut DatadirModification<'_>,
407 0 : ctx: &RequestContext,
408 0 : ) -> anyhow::Result<()> {
409 0 : let DbaseCreate {
410 0 : db_id,
411 0 : tablespace_id,
412 0 : src_db_id,
413 0 : src_tablespace_id,
414 0 : } = create;
415 :
416 0 : let rels = modification
417 0 : .tline
418 0 : .list_rels(
419 0 : src_tablespace_id,
420 0 : src_db_id,
421 0 : Version::Modified(modification),
422 0 : ctx,
423 0 : )
424 0 : .await?;
425 :
426 0 : debug!("ingest_xlog_dbase_create: {} rels", rels.len());
427 :
428 : // Copy relfilemap
429 0 : let filemap = modification
430 0 : .tline
431 0 : .get_relmap_file(
432 0 : src_tablespace_id,
433 0 : src_db_id,
434 0 : Version::Modified(modification),
435 0 : ctx,
436 0 : )
437 0 : .await?;
438 0 : modification
439 0 : .put_relmap_file(tablespace_id, db_id, filemap, ctx)
440 0 : .await?;
441 :
442 0 : let mut num_rels_copied = 0;
443 0 : let mut num_blocks_copied = 0;
444 0 : for src_rel in rels {
445 0 : assert_eq!(src_rel.spcnode, src_tablespace_id);
446 0 : assert_eq!(src_rel.dbnode, src_db_id);
447 :
448 0 : let nblocks = modification
449 0 : .tline
450 0 : .get_rel_size(src_rel, Version::Modified(modification), ctx)
451 0 : .await?;
452 0 : let dst_rel = RelTag {
453 0 : spcnode: tablespace_id,
454 0 : dbnode: db_id,
455 0 : relnode: src_rel.relnode,
456 0 : forknum: src_rel.forknum,
457 0 : };
458 0 :
459 0 : modification.put_rel_creation(dst_rel, nblocks, ctx).await?;
460 :
461 : // Copy content
462 0 : debug!("copying rel {} to {}, {} blocks", src_rel, dst_rel, nblocks);
463 0 : for blknum in 0..nblocks {
464 : // Sharding:
465 : // - src and dst are always on the same shard, because they differ only by dbNode, and
466 : // dbNode is not included in the hash inputs for sharding.
467 : // - This WAL command is replayed on all shards, but each shard only copies the blocks
468 : // that belong to it.
469 0 : let src_key = rel_block_to_key(src_rel, blknum);
470 0 : if !self.shard.is_key_local(&src_key) {
471 0 : debug!(
472 0 : "Skipping non-local key {} during XLOG_DBASE_CREATE",
473 : src_key
474 : );
475 0 : continue;
476 0 : }
477 0 : debug!(
478 0 : "copying block {} from {} ({}) to {}",
479 : blknum, src_rel, src_key, dst_rel
480 : );
481 :
482 0 : let content = modification
483 0 : .tline
484 0 : .get_rel_page_at_lsn(src_rel, blknum, Version::Modified(modification), ctx)
485 0 : .await?;
486 0 : modification.put_rel_page_image(dst_rel, blknum, content)?;
487 0 : num_blocks_copied += 1;
488 : }
489 :
490 0 : num_rels_copied += 1;
491 : }
492 :
493 0 : info!(
494 0 : "Created database {}/{}, copied {} blocks in {} rels",
495 : tablespace_id, db_id, num_blocks_copied, num_rels_copied
496 : );
497 0 : Ok(())
498 0 : }
499 :
500 0 : async fn ingest_xlog_dbase_drop(
501 0 : &mut self,
502 0 : dbase_drop: DbaseDrop,
503 0 : modification: &mut DatadirModification<'_>,
504 0 : ctx: &RequestContext,
505 0 : ) -> anyhow::Result<()> {
506 0 : let DbaseDrop {
507 0 : db_id,
508 0 : tablespace_ids,
509 0 : } = dbase_drop;
510 0 : for tablespace_id in tablespace_ids {
511 0 : trace!("Drop db {}, {}", tablespace_id, db_id);
512 0 : modification.drop_dbdir(tablespace_id, db_id, ctx).await?;
513 : }
514 :
515 0 : Ok(())
516 0 : }
517 :
518 16 : async fn ingest_xlog_smgr_create(
519 16 : &mut self,
520 16 : create: SmgrCreate,
521 16 : modification: &mut DatadirModification<'_>,
522 16 : ctx: &RequestContext,
523 16 : ) -> anyhow::Result<()> {
524 16 : let SmgrCreate { rel } = create;
525 16 : self.put_rel_creation(modification, rel, ctx).await?;
526 16 : Ok(())
527 16 : }
528 :
529 : /// Subroutine of ingest_record(), to handle an XLOG_SMGR_TRUNCATE record.
530 : ///
531 : /// This is the same logic as in PostgreSQL's smgr_redo() function.
532 0 : async fn ingest_xlog_smgr_truncate(
533 0 : &mut self,
534 0 : truncate: XlSmgrTruncate,
535 0 : modification: &mut DatadirModification<'_>,
536 0 : ctx: &RequestContext,
537 0 : ) -> anyhow::Result<()> {
538 0 : let XlSmgrTruncate {
539 0 : blkno,
540 0 : rnode,
541 0 : flags,
542 0 : } = truncate;
543 0 :
544 0 : let spcnode = rnode.spcnode;
545 0 : let dbnode = rnode.dbnode;
546 0 : let relnode = rnode.relnode;
547 0 :
548 0 : if flags & pg_constants::SMGR_TRUNCATE_HEAP != 0 {
549 0 : let rel = RelTag {
550 0 : spcnode,
551 0 : dbnode,
552 0 : relnode,
553 0 : forknum: MAIN_FORKNUM,
554 0 : };
555 0 :
556 0 : self.put_rel_truncation(modification, rel, blkno, ctx)
557 0 : .await?;
558 0 : }
559 0 : if flags & pg_constants::SMGR_TRUNCATE_FSM != 0 {
560 0 : let rel = RelTag {
561 0 : spcnode,
562 0 : dbnode,
563 0 : relnode,
564 0 : forknum: FSM_FORKNUM,
565 0 : };
566 0 :
567 0 : let fsm_logical_page_no = blkno / pg_constants::SLOTS_PER_FSM_PAGE;
568 0 : let mut fsm_physical_page_no = fsm_logical_to_physical(fsm_logical_page_no);
569 0 : if blkno % pg_constants::SLOTS_PER_FSM_PAGE != 0 {
570 : // Tail of last remaining FSM page has to be zeroed.
571 : // We are not precise here and instead of digging in FSM bitmap format just clear the whole page.
572 0 : modification.put_rel_page_image_zero(rel, fsm_physical_page_no)?;
573 0 : fsm_physical_page_no += 1;
574 0 : }
575 0 : let nblocks = get_relsize(modification, rel, ctx).await?;
576 0 : if nblocks > fsm_physical_page_no {
577 : // check if something to do: FSM is larger than truncate position
578 0 : self.put_rel_truncation(modification, rel, fsm_physical_page_no, ctx)
579 0 : .await?;
580 0 : }
581 0 : }
582 0 : if flags & pg_constants::SMGR_TRUNCATE_VM != 0 {
583 0 : let rel = RelTag {
584 0 : spcnode,
585 0 : dbnode,
586 0 : relnode,
587 0 : forknum: VISIBILITYMAP_FORKNUM,
588 0 : };
589 0 :
590 0 : let mut vm_page_no = blkno / pg_constants::VM_HEAPBLOCKS_PER_PAGE;
591 0 : if blkno % pg_constants::VM_HEAPBLOCKS_PER_PAGE != 0 {
592 : // Tail of last remaining vm page has to be zeroed.
593 : // We are not precise here and instead of digging in VM bitmap format just clear the whole page.
594 0 : modification.put_rel_page_image_zero(rel, vm_page_no)?;
595 0 : vm_page_no += 1;
596 0 : }
597 0 : let nblocks = get_relsize(modification, rel, ctx).await?;
598 0 : if nblocks > vm_page_no {
599 : // check if something to do: VM is larger than truncate position
600 0 : self.put_rel_truncation(modification, rel, vm_page_no, ctx)
601 0 : .await?;
602 0 : }
603 0 : }
604 0 : Ok(())
605 0 : }
606 :
607 8 : fn warn_on_ingest_lag(
608 8 : &mut self,
609 8 : conf: &crate::config::PageServerConf,
610 8 : wal_timestamp: TimestampTz,
611 8 : ) {
612 8 : debug_assert_current_span_has_tenant_and_timeline_id();
613 8 : let now = SystemTime::now();
614 8 : let rate_limits = &mut self.warn_ingest_lag;
615 :
616 8 : let ts = enum_pgversion_dispatch!(&self.checkpoint, CheckPoint, _cp, {
617 0 : pgv::xlog_utils::try_from_pg_timestamp(wal_timestamp)
618 : });
619 :
620 8 : match ts {
621 8 : Ok(ts) => {
622 8 : match now.duration_since(ts) {
623 8 : Ok(lag) => {
624 8 : if lag > conf.wait_lsn_timeout {
625 8 : rate_limits.lag_msg_ratelimit.call2(|rate_limit_stats| {
626 2 : if let Some(cooldown) = self.attach_wal_lag_cooldown.get() {
627 0 : if std::time::Instant::now() < cooldown.active_until && lag <= cooldown.max_lag {
628 0 : return;
629 0 : }
630 2 : } else {
631 2 : // Still loading? We shouldn't be here
632 2 : }
633 2 : let lag = humantime::format_duration(lag);
634 2 : warn!(%rate_limit_stats, %lag, "ingesting record with timestamp lagging more than wait_lsn_timeout");
635 8 : })
636 0 : }
637 : }
638 0 : Err(e) => {
639 0 : let delta_t = e.duration();
640 : // determined by prod victoriametrics query: 1000 * (timestamp(node_time_seconds{neon_service="pageserver"}) - node_time_seconds)
641 : // => https://www.robustperception.io/time-metric-from-the-node-exporter/
642 : const IGNORED_DRIFT: Duration = Duration::from_millis(100);
643 0 : if delta_t > IGNORED_DRIFT {
644 0 : let delta_t = humantime::format_duration(delta_t);
645 0 : rate_limits.future_lsn_msg_ratelimit.call2(|rate_limit_stats| {
646 0 : warn!(%rate_limit_stats, %delta_t, "ingesting record with timestamp from future");
647 0 : })
648 0 : }
649 : }
650 : };
651 : }
652 0 : Err(error) => {
653 0 : rate_limits.timestamp_invalid_msg_ratelimit.call2(|rate_limit_stats| {
654 0 : warn!(%rate_limit_stats, %error, "ingesting record with invalid timestamp, cannot calculate lag and will fail find-lsn-for-timestamp type queries");
655 0 : })
656 : }
657 : }
658 8 : }
659 :
660 : /// Subroutine of ingest_record(), to handle an XLOG_XACT_* records.
661 : ///
662 8 : async fn ingest_xact_record(
663 8 : &mut self,
664 8 : record: XactRecord,
665 8 : modification: &mut DatadirModification<'_>,
666 8 : ctx: &RequestContext,
667 8 : ) -> anyhow::Result<()> {
668 8 : let (xact_common, is_commit, is_prepared) = match record {
669 0 : XactRecord::Prepare(XactPrepare { xl_xid, data }) => {
670 0 : let xid: u64 = if modification.tline.pg_version >= 17 {
671 0 : self.adjust_to_full_transaction_id(xl_xid)?
672 : } else {
673 0 : xl_xid as u64
674 : };
675 0 : return modification.put_twophase_file(xid, data, ctx).await;
676 : }
677 8 : XactRecord::Commit(common) => (common, true, false),
678 0 : XactRecord::Abort(common) => (common, false, false),
679 0 : XactRecord::CommitPrepared(common) => (common, true, true),
680 0 : XactRecord::AbortPrepared(common) => (common, false, true),
681 : };
682 :
683 : let XactCommon {
684 8 : parsed,
685 8 : origin_id,
686 8 : xl_xid,
687 8 : lsn,
688 8 : } = xact_common;
689 8 :
690 8 : // Record update of CLOG pages
691 8 : let mut pageno = parsed.xid / pg_constants::CLOG_XACTS_PER_PAGE;
692 8 : let mut segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
693 8 : let mut rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
694 8 : let mut page_xids: Vec<TransactionId> = vec![parsed.xid];
695 8 :
696 8 : self.warn_on_ingest_lag(modification.tline.conf, parsed.xact_time);
697 :
698 8 : for subxact in &parsed.subxacts {
699 0 : let subxact_pageno = subxact / pg_constants::CLOG_XACTS_PER_PAGE;
700 0 : if subxact_pageno != pageno {
701 : // This subxact goes to different page. Write the record
702 : // for all the XIDs on the previous page, and continue
703 : // accumulating XIDs on this new page.
704 0 : modification.put_slru_wal_record(
705 0 : SlruKind::Clog,
706 0 : segno,
707 0 : rpageno,
708 0 : if is_commit {
709 0 : NeonWalRecord::ClogSetCommitted {
710 0 : xids: page_xids,
711 0 : timestamp: parsed.xact_time,
712 0 : }
713 : } else {
714 0 : NeonWalRecord::ClogSetAborted { xids: page_xids }
715 : },
716 0 : )?;
717 0 : page_xids = Vec::new();
718 0 : }
719 0 : pageno = subxact_pageno;
720 0 : segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
721 0 : rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
722 0 : page_xids.push(*subxact);
723 : }
724 8 : modification.put_slru_wal_record(
725 8 : SlruKind::Clog,
726 8 : segno,
727 8 : rpageno,
728 8 : if is_commit {
729 8 : NeonWalRecord::ClogSetCommitted {
730 8 : xids: page_xids,
731 8 : timestamp: parsed.xact_time,
732 8 : }
733 : } else {
734 0 : NeonWalRecord::ClogSetAborted { xids: page_xids }
735 : },
736 0 : )?;
737 :
738 : // Group relations to drop by dbNode. This map will contain all relations that _might_
739 : // exist, we will reduce it to which ones really exist later. This map can be huge if
740 : // the transaction touches a huge number of relations (there is no bound on this in
741 : // postgres).
742 8 : let mut drop_relations: HashMap<(u32, u32), Vec<RelTag>> = HashMap::new();
743 :
744 8 : for xnode in &parsed.xnodes {
745 0 : for forknum in MAIN_FORKNUM..=INIT_FORKNUM {
746 0 : let rel = RelTag {
747 0 : forknum,
748 0 : spcnode: xnode.spcnode,
749 0 : dbnode: xnode.dbnode,
750 0 : relnode: xnode.relnode,
751 0 : };
752 0 : drop_relations
753 0 : .entry((xnode.spcnode, xnode.dbnode))
754 0 : .or_default()
755 0 : .push(rel);
756 0 : }
757 : }
758 :
759 : // Execute relation drops in a batch: the number may be huge, so deleting individually is prohibitively expensive
760 8 : modification.put_rel_drops(drop_relations, ctx).await?;
761 :
762 8 : if origin_id != 0 {
763 0 : modification
764 0 : .set_replorigin(origin_id, parsed.origin_lsn)
765 0 : .await?;
766 8 : }
767 :
768 8 : if is_prepared {
769 : // Remove twophase file. see RemoveTwoPhaseFile() in postgres code
770 0 : trace!(
771 0 : "Drop twophaseFile for xid {} parsed_xact.xid {} here at {}",
772 : xl_xid,
773 : parsed.xid,
774 : lsn,
775 : );
776 :
777 0 : let xid: u64 = if modification.tline.pg_version >= 17 {
778 0 : self.adjust_to_full_transaction_id(parsed.xid)?
779 : } else {
780 0 : parsed.xid as u64
781 : };
782 0 : modification.drop_twophase_file(xid, ctx).await?;
783 8 : }
784 :
785 8 : Ok(())
786 8 : }
787 :
788 0 : async fn ingest_clog_truncate(
789 0 : &mut self,
790 0 : truncate: ClogTruncate,
791 0 : modification: &mut DatadirModification<'_>,
792 0 : ctx: &RequestContext,
793 0 : ) -> anyhow::Result<()> {
794 0 : let ClogTruncate {
795 0 : pageno,
796 0 : oldest_xid,
797 0 : oldest_xid_db,
798 0 : } = truncate;
799 0 :
800 0 : info!(
801 0 : "RM_CLOG_ID truncate pageno {} oldestXid {} oldestXidDB {}",
802 : pageno, oldest_xid, oldest_xid_db
803 : );
804 :
805 : // In Postgres, oldestXid and oldestXidDB are updated in memory when the CLOG is
806 : // truncated, but a checkpoint record with the updated values isn't written until
807 : // later. In Neon, a server can start at any LSN, not just on a checkpoint record,
808 : // so we keep the oldestXid and oldestXidDB up-to-date.
809 0 : enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
810 0 : cp.oldestXid = oldest_xid;
811 0 : cp.oldestXidDB = oldest_xid_db;
812 0 : });
813 0 : self.checkpoint_modified = true;
814 :
815 : // TODO Treat AdvanceOldestClogXid() or write a comment why we don't need it
816 :
817 0 : let latest_page_number =
818 0 : enum_pgversion_dispatch!(self.checkpoint, CheckPoint, cp, { cp.nextXid.value }) as u32
819 : / pg_constants::CLOG_XACTS_PER_PAGE;
820 :
821 : // Now delete all segments containing pages between xlrec.pageno
822 : // and latest_page_number.
823 :
824 : // First, make an important safety check:
825 : // the current endpoint page must not be eligible for removal.
826 : // See SimpleLruTruncate() in slru.c
827 0 : if dispatch_pgversion!(modification.tline.pg_version, {
828 0 : pgv::nonrelfile_utils::clogpage_precedes(latest_page_number, pageno)
829 : }) {
830 0 : info!("could not truncate directory pg_xact apparent wraparound");
831 0 : return Ok(());
832 0 : }
833 :
834 : // Iterate via SLRU CLOG segments and drop segments that we're ready to truncate
835 : //
836 : // We cannot pass 'lsn' to the Timeline.list_nonrels(), or it
837 : // will block waiting for the last valid LSN to advance up to
838 : // it. So we use the previous record's LSN in the get calls
839 : // instead.
840 0 : for segno in modification
841 0 : .tline
842 0 : .list_slru_segments(SlruKind::Clog, Version::Modified(modification), ctx)
843 0 : .await?
844 : {
845 0 : let segpage = segno * pg_constants::SLRU_PAGES_PER_SEGMENT;
846 :
847 0 : let may_delete = dispatch_pgversion!(modification.tline.pg_version, {
848 0 : pgv::nonrelfile_utils::slru_may_delete_clogsegment(segpage, pageno)
849 : });
850 :
851 0 : if may_delete {
852 0 : modification
853 0 : .drop_slru_segment(SlruKind::Clog, segno, ctx)
854 0 : .await?;
855 0 : trace!("Drop CLOG segment {:>04X}", segno);
856 0 : }
857 : }
858 :
859 0 : Ok(())
860 0 : }
861 :
862 0 : async fn ingest_clog_zero_page(
863 0 : &mut self,
864 0 : zero_page: ClogZeroPage,
865 0 : modification: &mut DatadirModification<'_>,
866 0 : ctx: &RequestContext,
867 0 : ) -> anyhow::Result<()> {
868 0 : let ClogZeroPage { segno, rpageno } = zero_page;
869 0 :
870 0 : self.put_slru_page_image(
871 0 : modification,
872 0 : SlruKind::Clog,
873 0 : segno,
874 0 : rpageno,
875 0 : ZERO_PAGE.clone(),
876 0 : ctx,
877 0 : )
878 0 : .await
879 0 : }
880 :
881 0 : fn ingest_multixact_create(
882 0 : &mut self,
883 0 : modification: &mut DatadirModification,
884 0 : xlrec: &XlMultiXactCreate,
885 0 : ) -> Result<()> {
886 0 : // Create WAL record for updating the multixact-offsets page
887 0 : let pageno = xlrec.mid / pg_constants::MULTIXACT_OFFSETS_PER_PAGE as u32;
888 0 : let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
889 0 : let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
890 0 :
891 0 : modification.put_slru_wal_record(
892 0 : SlruKind::MultiXactOffsets,
893 0 : segno,
894 0 : rpageno,
895 0 : NeonWalRecord::MultixactOffsetCreate {
896 0 : mid: xlrec.mid,
897 0 : moff: xlrec.moff,
898 0 : },
899 0 : )?;
900 :
901 : // Create WAL records for the update of each affected multixact-members page
902 0 : let mut members = xlrec.members.iter();
903 0 : let mut offset = xlrec.moff;
904 : loop {
905 0 : let pageno = offset / pg_constants::MULTIXACT_MEMBERS_PER_PAGE as u32;
906 0 :
907 0 : // How many members fit on this page?
908 0 : let page_remain = pg_constants::MULTIXACT_MEMBERS_PER_PAGE as u32
909 0 : - offset % pg_constants::MULTIXACT_MEMBERS_PER_PAGE as u32;
910 0 :
911 0 : let mut this_page_members: Vec<MultiXactMember> = Vec::new();
912 0 : for _ in 0..page_remain {
913 0 : if let Some(m) = members.next() {
914 0 : this_page_members.push(m.clone());
915 0 : } else {
916 0 : break;
917 : }
918 : }
919 0 : if this_page_members.is_empty() {
920 : // all done
921 0 : break;
922 0 : }
923 0 : let n_this_page = this_page_members.len();
924 0 :
925 0 : modification.put_slru_wal_record(
926 0 : SlruKind::MultiXactMembers,
927 0 : pageno / pg_constants::SLRU_PAGES_PER_SEGMENT,
928 0 : pageno % pg_constants::SLRU_PAGES_PER_SEGMENT,
929 0 : NeonWalRecord::MultixactMembersCreate {
930 0 : moff: offset,
931 0 : members: this_page_members,
932 0 : },
933 0 : )?;
934 :
935 : // Note: The multixact members can wrap around, even within one WAL record.
936 0 : offset = offset.wrapping_add(n_this_page as u32);
937 : }
938 0 : let next_offset = offset;
939 0 : assert!(xlrec.moff.wrapping_add(xlrec.nmembers) == next_offset);
940 :
941 : // Update next-multi-xid and next-offset
942 : //
943 : // NB: In PostgreSQL, the next-multi-xid stored in the control file is allowed to
944 : // go to 0, and it's fixed up by skipping to FirstMultiXactId in functions that
945 : // read it, like GetNewMultiXactId(). This is different from how nextXid is
946 : // incremented! nextXid skips over < FirstNormalTransactionId when the the value
947 : // is stored, so it's never 0 in a checkpoint.
948 : //
949 : // I don't know why it's done that way, it seems less error-prone to skip over 0
950 : // when the value is stored rather than when it's read. But let's do it the same
951 : // way here.
952 0 : let next_multi_xid = xlrec.mid.wrapping_add(1);
953 0 :
954 0 : if self
955 0 : .checkpoint
956 0 : .update_next_multixid(next_multi_xid, next_offset)
957 0 : {
958 0 : self.checkpoint_modified = true;
959 0 : }
960 :
961 : // Also update the next-xid with the highest member. According to the comments in
962 : // multixact_redo(), this shouldn't be necessary, but let's do the same here.
963 0 : let max_mbr_xid = xlrec.members.iter().fold(None, |acc, mbr| {
964 0 : if let Some(max_xid) = acc {
965 0 : if mbr.xid.wrapping_sub(max_xid) as i32 > 0 {
966 0 : Some(mbr.xid)
967 : } else {
968 0 : acc
969 : }
970 : } else {
971 0 : Some(mbr.xid)
972 : }
973 0 : });
974 :
975 0 : if let Some(max_xid) = max_mbr_xid {
976 0 : if self.checkpoint.update_next_xid(max_xid) {
977 0 : self.checkpoint_modified = true;
978 0 : }
979 0 : }
980 0 : Ok(())
981 0 : }
982 :
983 0 : async fn ingest_multixact_truncate(
984 0 : &mut self,
985 0 : modification: &mut DatadirModification<'_>,
986 0 : xlrec: &XlMultiXactTruncate,
987 0 : ctx: &RequestContext,
988 0 : ) -> Result<()> {
989 0 : let (maxsegment, startsegment, endsegment) =
990 0 : enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
991 0 : cp.oldestMulti = xlrec.end_trunc_off;
992 0 : cp.oldestMultiDB = xlrec.oldest_multi_db;
993 0 : let maxsegment: i32 = pgv::nonrelfile_utils::mx_offset_to_member_segment(
994 0 : pg_constants::MAX_MULTIXACT_OFFSET,
995 0 : );
996 0 : let startsegment: i32 =
997 0 : pgv::nonrelfile_utils::mx_offset_to_member_segment(xlrec.start_trunc_memb);
998 0 : let endsegment: i32 =
999 0 : pgv::nonrelfile_utils::mx_offset_to_member_segment(xlrec.end_trunc_memb);
1000 0 : (maxsegment, startsegment, endsegment)
1001 : });
1002 :
1003 0 : self.checkpoint_modified = true;
1004 0 :
1005 0 : // PerformMembersTruncation
1006 0 : let mut segment: i32 = startsegment;
1007 :
1008 : // Delete all the segments except the last one. The last segment can still
1009 : // contain, possibly partially, valid data.
1010 0 : while segment != endsegment {
1011 0 : modification
1012 0 : .drop_slru_segment(SlruKind::MultiXactMembers, segment as u32, ctx)
1013 0 : .await?;
1014 :
1015 : /* move to next segment, handling wraparound correctly */
1016 0 : if segment == maxsegment {
1017 0 : segment = 0;
1018 0 : } else {
1019 0 : segment += 1;
1020 0 : }
1021 : }
1022 :
1023 : // Truncate offsets
1024 : // FIXME: this did not handle wraparound correctly
1025 :
1026 0 : Ok(())
1027 0 : }
1028 :
1029 0 : async fn ingest_multixact_zero_page(
1030 0 : &mut self,
1031 0 : zero_page: MultiXactZeroPage,
1032 0 : modification: &mut DatadirModification<'_>,
1033 0 : ctx: &RequestContext,
1034 0 : ) -> Result<()> {
1035 0 : let MultiXactZeroPage {
1036 0 : slru_kind,
1037 0 : segno,
1038 0 : rpageno,
1039 0 : } = zero_page;
1040 0 : self.put_slru_page_image(
1041 0 : modification,
1042 0 : slru_kind,
1043 0 : segno,
1044 0 : rpageno,
1045 0 : ZERO_PAGE.clone(),
1046 0 : ctx,
1047 0 : )
1048 0 : .await
1049 0 : }
1050 :
1051 0 : async fn ingest_relmap_update(
1052 0 : &mut self,
1053 0 : update: RelmapUpdate,
1054 0 : modification: &mut DatadirModification<'_>,
1055 0 : ctx: &RequestContext,
1056 0 : ) -> Result<()> {
1057 0 : let RelmapUpdate { update, buf } = update;
1058 0 :
1059 0 : modification
1060 0 : .put_relmap_file(update.tsid, update.dbid, buf, ctx)
1061 0 : .await
1062 0 : }
1063 :
1064 30 : async fn ingest_raw_xlog_record(
1065 30 : &mut self,
1066 30 : raw_record: RawXlogRecord,
1067 30 : modification: &mut DatadirModification<'_>,
1068 30 : ctx: &RequestContext,
1069 30 : ) -> Result<()> {
1070 30 : let RawXlogRecord { info, lsn, mut buf } = raw_record;
1071 30 : let pg_version = modification.tline.pg_version;
1072 30 :
1073 30 : if info == pg_constants::XLOG_PARAMETER_CHANGE {
1074 2 : if let CheckPoint::V17(cp) = &mut self.checkpoint {
1075 0 : let rec = v17::XlParameterChange::decode(&mut buf);
1076 0 : cp.wal_level = rec.wal_level;
1077 0 : self.checkpoint_modified = true;
1078 2 : }
1079 28 : } else if info == pg_constants::XLOG_END_OF_RECOVERY {
1080 0 : if let CheckPoint::V17(cp) = &mut self.checkpoint {
1081 0 : let rec = v17::XlEndOfRecovery::decode(&mut buf);
1082 0 : cp.wal_level = rec.wal_level;
1083 0 : self.checkpoint_modified = true;
1084 0 : }
1085 28 : }
1086 :
1087 30 : enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
1088 0 : if info == pg_constants::XLOG_NEXTOID {
1089 0 : let next_oid = buf.get_u32_le();
1090 0 : if cp.nextOid != next_oid {
1091 0 : cp.nextOid = next_oid;
1092 0 : self.checkpoint_modified = true;
1093 0 : }
1094 0 : } else if info == pg_constants::XLOG_CHECKPOINT_ONLINE
1095 0 : || info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
1096 : {
1097 0 : let mut checkpoint_bytes = [0u8; pgv::xlog_utils::SIZEOF_CHECKPOINT];
1098 0 : buf.copy_to_slice(&mut checkpoint_bytes);
1099 0 : let xlog_checkpoint = pgv::CheckPoint::decode(&checkpoint_bytes)?;
1100 0 : trace!(
1101 0 : "xlog_checkpoint.oldestXid={}, checkpoint.oldestXid={}",
1102 : xlog_checkpoint.oldestXid,
1103 : cp.oldestXid
1104 : );
1105 0 : if (cp.oldestXid.wrapping_sub(xlog_checkpoint.oldestXid) as i32) < 0 {
1106 0 : cp.oldestXid = xlog_checkpoint.oldestXid;
1107 0 : }
1108 0 : trace!(
1109 0 : "xlog_checkpoint.oldestActiveXid={}, checkpoint.oldestActiveXid={}",
1110 : xlog_checkpoint.oldestActiveXid,
1111 : cp.oldestActiveXid
1112 : );
1113 :
1114 : // A shutdown checkpoint has `oldestActiveXid == InvalidTransactionid`,
1115 : // because at shutdown, all in-progress transactions will implicitly
1116 : // end. Postgres startup code knows that, and allows hot standby to start
1117 : // immediately from a shutdown checkpoint.
1118 : //
1119 : // In Neon, Postgres hot standby startup always behaves as if starting from
1120 : // an online checkpoint. It needs a valid `oldestActiveXid` value, so
1121 : // instead of overwriting self.checkpoint.oldestActiveXid with
1122 : // InvalidTransactionid from the checkpoint WAL record, update it to a
1123 : // proper value, knowing that there are no in-progress transactions at this
1124 : // point, except for prepared transactions.
1125 : //
1126 : // See also the neon code changes in the InitWalRecovery() function.
1127 0 : if xlog_checkpoint.oldestActiveXid == pg_constants::INVALID_TRANSACTION_ID
1128 0 : && info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
1129 : {
1130 0 : let oldest_active_xid = if pg_version >= 17 {
1131 0 : let mut oldest_active_full_xid = cp.nextXid.value;
1132 0 : for xid in modification.tline.list_twophase_files(lsn, ctx).await? {
1133 0 : if xid < oldest_active_full_xid {
1134 0 : oldest_active_full_xid = xid;
1135 0 : }
1136 : }
1137 0 : oldest_active_full_xid as u32
1138 : } else {
1139 0 : let mut oldest_active_xid = cp.nextXid.value as u32;
1140 0 : for xid in modification.tline.list_twophase_files(lsn, ctx).await? {
1141 0 : let narrow_xid = xid as u32;
1142 0 : if (narrow_xid.wrapping_sub(oldest_active_xid) as i32) < 0 {
1143 0 : oldest_active_xid = narrow_xid;
1144 0 : }
1145 : }
1146 0 : oldest_active_xid
1147 : };
1148 0 : cp.oldestActiveXid = oldest_active_xid;
1149 0 : } else {
1150 0 : cp.oldestActiveXid = xlog_checkpoint.oldestActiveXid;
1151 0 : }
1152 :
1153 : // Write a new checkpoint key-value pair on every checkpoint record, even
1154 : // if nothing really changed. Not strictly required, but it seems nice to
1155 : // have some trace of the checkpoint records in the layer files at the same
1156 : // LSNs.
1157 0 : self.checkpoint_modified = true;
1158 0 : }
1159 : });
1160 :
1161 30 : Ok(())
1162 30 : }
1163 :
1164 0 : async fn ingest_logical_message_put(
1165 0 : &mut self,
1166 0 : put: PutLogicalMessage,
1167 0 : modification: &mut DatadirModification<'_>,
1168 0 : ctx: &RequestContext,
1169 0 : ) -> Result<()> {
1170 0 : let PutLogicalMessage { path, buf } = put;
1171 0 : modification.put_file(path.as_str(), &buf, ctx).await
1172 0 : }
1173 :
1174 0 : fn ingest_standby_record(&mut self, record: StandbyRecord) -> Result<()> {
1175 0 : match record {
1176 0 : StandbyRecord::RunningXacts(running_xacts) => {
1177 0 : enum_pgversion_dispatch!(&mut self.checkpoint, CheckPoint, cp, {
1178 0 : cp.oldestActiveXid = running_xacts.oldest_running_xid;
1179 0 : });
1180 :
1181 0 : self.checkpoint_modified = true;
1182 0 : }
1183 0 : }
1184 0 :
1185 0 : Ok(())
1186 0 : }
1187 :
1188 0 : async fn ingest_replorigin_record(
1189 0 : &mut self,
1190 0 : record: ReploriginRecord,
1191 0 : modification: &mut DatadirModification<'_>,
1192 0 : ) -> Result<()> {
1193 0 : match record {
1194 0 : ReploriginRecord::Set(set) => {
1195 0 : modification
1196 0 : .set_replorigin(set.node_id, set.remote_lsn)
1197 0 : .await?;
1198 : }
1199 0 : ReploriginRecord::Drop(drop) => {
1200 0 : modification.drop_replorigin(drop.node_id).await?;
1201 : }
1202 : }
1203 :
1204 0 : Ok(())
1205 0 : }
1206 :
1207 18 : async fn put_rel_creation(
1208 18 : &mut self,
1209 18 : modification: &mut DatadirModification<'_>,
1210 18 : rel: RelTag,
1211 18 : ctx: &RequestContext,
1212 18 : ) -> Result<()> {
1213 18 : modification.put_rel_creation(rel, 0, ctx).await?;
1214 18 : Ok(())
1215 18 : }
1216 :
1217 : #[cfg(test)]
1218 272402 : async fn put_rel_page_image(
1219 272402 : &mut self,
1220 272402 : modification: &mut DatadirModification<'_>,
1221 272402 : rel: RelTag,
1222 272402 : blknum: BlockNumber,
1223 272402 : img: Bytes,
1224 272402 : ctx: &RequestContext,
1225 272402 : ) -> Result<(), PageReconstructError> {
1226 272402 : self.handle_rel_extend(modification, rel, blknum, ctx)
1227 5420 : .await?;
1228 272402 : modification.put_rel_page_image(rel, blknum, img)?;
1229 272402 : Ok(())
1230 272402 : }
1231 :
1232 12 : async fn put_rel_wal_record(
1233 12 : &mut self,
1234 12 : modification: &mut DatadirModification<'_>,
1235 12 : rel: RelTag,
1236 12 : blknum: BlockNumber,
1237 12 : rec: NeonWalRecord,
1238 12 : ctx: &RequestContext,
1239 12 : ) -> Result<()> {
1240 12 : self.handle_rel_extend(modification, rel, blknum, ctx)
1241 0 : .await?;
1242 12 : modification.put_rel_wal_record(rel, blknum, rec)?;
1243 12 : Ok(())
1244 12 : }
1245 :
1246 6012 : async fn put_rel_truncation(
1247 6012 : &mut self,
1248 6012 : modification: &mut DatadirModification<'_>,
1249 6012 : rel: RelTag,
1250 6012 : nblocks: BlockNumber,
1251 6012 : ctx: &RequestContext,
1252 6012 : ) -> anyhow::Result<()> {
1253 6012 : modification.put_rel_truncation(rel, nblocks, ctx).await?;
1254 6012 : Ok(())
1255 6012 : }
1256 :
1257 272414 : async fn handle_rel_extend(
1258 272414 : &mut self,
1259 272414 : modification: &mut DatadirModification<'_>,
1260 272414 : rel: RelTag,
1261 272414 : blknum: BlockNumber,
1262 272414 : ctx: &RequestContext,
1263 272414 : ) -> Result<(), PageReconstructError> {
1264 272414 : let new_nblocks = blknum + 1;
1265 : // Check if the relation exists. We implicitly create relations on first
1266 : // record.
1267 272414 : let old_nblocks = modification.create_relation_if_required(rel, ctx).await?;
1268 :
1269 272414 : if new_nblocks > old_nblocks {
1270 : //info!("extending {} {} to {}", rel, old_nblocks, new_nblocks);
1271 272398 : modification.put_rel_extend(rel, new_nblocks, ctx).await?;
1272 :
1273 272398 : let mut key = rel_block_to_key(rel, blknum);
1274 272398 :
1275 272398 : // fill the gap with zeros
1276 272398 : let mut gap_blocks_filled: u64 = 0;
1277 272398 : for gap_blknum in old_nblocks..blknum {
1278 2998 : key.field6 = gap_blknum;
1279 2998 :
1280 2998 : if self.shard.get_shard_number(&key) != self.shard.number {
1281 0 : continue;
1282 2998 : }
1283 2998 :
1284 2998 : modification.put_rel_page_image_zero(rel, gap_blknum)?;
1285 2998 : gap_blocks_filled += 1;
1286 : }
1287 :
1288 272398 : WAL_INGEST
1289 272398 : .gap_blocks_zeroed_on_rel_extend
1290 272398 : .inc_by(gap_blocks_filled);
1291 272398 :
1292 272398 : // Log something when relation extends cause use to fill gaps
1293 272398 : // with zero pages. Logging is rate limited per pg version to
1294 272398 : // avoid skewing.
1295 272398 : if gap_blocks_filled > 0 {
1296 : use once_cell::sync::Lazy;
1297 : use std::sync::Mutex;
1298 : use utils::rate_limit::RateLimit;
1299 :
1300 : struct RateLimitPerPgVersion {
1301 : rate_limiters: [Lazy<Mutex<RateLimit>>; 4],
1302 : }
1303 :
1304 : impl RateLimitPerPgVersion {
1305 0 : const fn new() -> Self {
1306 0 : Self {
1307 0 : rate_limiters: [const {
1308 2 : Lazy::new(|| Mutex::new(RateLimit::new(Duration::from_secs(30))))
1309 0 : }; 4],
1310 0 : }
1311 0 : }
1312 :
1313 4 : const fn rate_limiter(
1314 4 : &self,
1315 4 : pg_version: u32,
1316 4 : ) -> Option<&Lazy<Mutex<RateLimit>>> {
1317 : const MIN_PG_VERSION: u32 = 14;
1318 : const MAX_PG_VERSION: u32 = 17;
1319 :
1320 4 : if pg_version < MIN_PG_VERSION || pg_version > MAX_PG_VERSION {
1321 0 : return None;
1322 4 : }
1323 4 :
1324 4 : Some(&self.rate_limiters[(pg_version - MIN_PG_VERSION) as usize])
1325 4 : }
1326 : }
1327 :
1328 : static LOGGED: RateLimitPerPgVersion = RateLimitPerPgVersion::new();
1329 4 : if let Some(rate_limiter) = LOGGED.rate_limiter(modification.tline.pg_version) {
1330 4 : if let Ok(mut locked) = rate_limiter.try_lock() {
1331 4 : locked.call(|| {
1332 2 : info!(
1333 0 : lsn=%modification.get_lsn(),
1334 0 : pg_version=%modification.tline.pg_version,
1335 0 : rel=%rel,
1336 0 : "Filled {} gap blocks on rel extend to {} from {}",
1337 : gap_blocks_filled,
1338 : new_nblocks,
1339 : old_nblocks);
1340 4 : });
1341 4 : }
1342 0 : }
1343 272394 : }
1344 16 : }
1345 272414 : Ok(())
1346 272414 : }
1347 :
1348 0 : async fn put_slru_page_image(
1349 0 : &mut self,
1350 0 : modification: &mut DatadirModification<'_>,
1351 0 : kind: SlruKind,
1352 0 : segno: u32,
1353 0 : blknum: BlockNumber,
1354 0 : img: Bytes,
1355 0 : ctx: &RequestContext,
1356 0 : ) -> Result<()> {
1357 0 : self.handle_slru_extend(modification, kind, segno, blknum, ctx)
1358 0 : .await?;
1359 0 : modification.put_slru_page_image(kind, segno, blknum, img)?;
1360 0 : Ok(())
1361 0 : }
1362 :
1363 0 : async fn handle_slru_extend(
1364 0 : &mut self,
1365 0 : modification: &mut DatadirModification<'_>,
1366 0 : kind: SlruKind,
1367 0 : segno: u32,
1368 0 : blknum: BlockNumber,
1369 0 : ctx: &RequestContext,
1370 0 : ) -> anyhow::Result<()> {
1371 0 : // we don't use a cache for this like we do for relations. SLRUS are explcitly
1372 0 : // extended with ZEROPAGE records, not with commit records, so it happens
1373 0 : // a lot less frequently.
1374 0 :
1375 0 : let new_nblocks = blknum + 1;
1376 : // Check if the relation exists. We implicitly create relations on first
1377 : // record.
1378 : // TODO: would be nice if to be more explicit about it
1379 0 : let old_nblocks = if !modification
1380 0 : .tline
1381 0 : .get_slru_segment_exists(kind, segno, Version::Modified(modification), ctx)
1382 0 : .await?
1383 : {
1384 : // create it with 0 size initially, the logic below will extend it
1385 0 : modification
1386 0 : .put_slru_segment_creation(kind, segno, 0, ctx)
1387 0 : .await?;
1388 0 : 0
1389 : } else {
1390 0 : modification
1391 0 : .tline
1392 0 : .get_slru_segment_size(kind, segno, Version::Modified(modification), ctx)
1393 0 : .await?
1394 : };
1395 :
1396 0 : if new_nblocks > old_nblocks {
1397 0 : trace!(
1398 0 : "extending SLRU {:?} seg {} from {} to {} blocks",
1399 : kind,
1400 : segno,
1401 : old_nblocks,
1402 : new_nblocks
1403 : );
1404 0 : modification.put_slru_extend(kind, segno, new_nblocks)?;
1405 :
1406 : // fill the gap with zeros
1407 0 : for gap_blknum in old_nblocks..blknum {
1408 0 : modification.put_slru_page_image_zero(kind, segno, gap_blknum)?;
1409 : }
1410 0 : }
1411 0 : Ok(())
1412 0 : }
1413 : }
1414 :
1415 12 : async fn get_relsize(
1416 12 : modification: &DatadirModification<'_>,
1417 12 : rel: RelTag,
1418 12 : ctx: &RequestContext,
1419 12 : ) -> Result<BlockNumber, PageReconstructError> {
1420 12 : let nblocks = if !modification
1421 12 : .tline
1422 12 : .get_rel_exists(rel, Version::Modified(modification), ctx)
1423 0 : .await?
1424 : {
1425 0 : 0
1426 : } else {
1427 12 : modification
1428 12 : .tline
1429 12 : .get_rel_size(rel, Version::Modified(modification), ctx)
1430 0 : .await?
1431 : };
1432 12 : Ok(nblocks)
1433 12 : }
1434 :
1435 : #[allow(clippy::bool_assert_comparison)]
1436 : #[cfg(test)]
1437 : mod tests {
1438 : use super::*;
1439 : use crate::tenant::harness::*;
1440 : use crate::tenant::remote_timeline_client::{remote_initdb_archive_path, INITDB_PATH};
1441 : use postgres_ffi::RELSEG_SIZE;
1442 :
1443 : use crate::DEFAULT_PG_VERSION;
1444 :
1445 : /// Arbitrary relation tag, for testing.
1446 : const TESTREL_A: RelTag = RelTag {
1447 : spcnode: 0,
1448 : dbnode: 111,
1449 : relnode: 1000,
1450 : forknum: 0,
1451 : };
1452 :
1453 12 : fn assert_current_logical_size(_timeline: &Timeline, _lsn: Lsn) {
1454 12 : // TODO
1455 12 : }
1456 :
1457 : #[tokio::test]
1458 2 : async fn test_zeroed_checkpoint_decodes_correctly() -> Result<()> {
1459 8 : for i in 14..=16 {
1460 6 : dispatch_pgversion!(i, {
1461 2 : pgv::CheckPoint::decode(&pgv::ZERO_CHECKPOINT)?;
1462 2 : });
1463 2 : }
1464 2 :
1465 2 : Ok(())
1466 2 : }
1467 :
1468 8 : async fn init_walingest_test(tline: &Timeline, ctx: &RequestContext) -> Result<WalIngest> {
1469 8 : let mut m = tline.begin_modification(Lsn(0x10));
1470 8 : m.put_checkpoint(dispatch_pgversion!(
1471 8 : tline.pg_version,
1472 0 : pgv::ZERO_CHECKPOINT.clone()
1473 0 : ))?;
1474 16 : m.put_relmap_file(0, 111, Bytes::from(""), ctx).await?; // dummy relmapper file
1475 8 : m.commit(ctx).await?;
1476 8 : let walingest = WalIngest::new(tline, Lsn(0x10), ctx).await?;
1477 :
1478 8 : Ok(walingest)
1479 8 : }
1480 :
1481 : #[tokio::test]
1482 2 : async fn test_relsize() -> Result<()> {
1483 20 : let (tenant, ctx) = TenantHarness::create("test_relsize").await?.load().await;
1484 2 : let tline = tenant
1485 2 : .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
1486 6 : .await?;
1487 5 : let mut walingest = init_walingest_test(&tline, &ctx).await?;
1488 2 :
1489 2 : let mut m = tline.begin_modification(Lsn(0x20));
1490 2 : walingest.put_rel_creation(&mut m, TESTREL_A, &ctx).await?;
1491 2 : walingest
1492 2 : .put_rel_page_image(&mut m, TESTREL_A, 0, test_img("foo blk 0 at 2"), &ctx)
1493 2 : .await?;
1494 2 : m.commit(&ctx).await?;
1495 2 : let mut m = tline.begin_modification(Lsn(0x30));
1496 2 : walingest
1497 2 : .put_rel_page_image(&mut m, TESTREL_A, 0, test_img("foo blk 0 at 3"), &ctx)
1498 2 : .await?;
1499 2 : m.commit(&ctx).await?;
1500 2 : let mut m = tline.begin_modification(Lsn(0x40));
1501 2 : walingest
1502 2 : .put_rel_page_image(&mut m, TESTREL_A, 1, test_img("foo blk 1 at 4"), &ctx)
1503 2 : .await?;
1504 2 : m.commit(&ctx).await?;
1505 2 : let mut m = tline.begin_modification(Lsn(0x50));
1506 2 : walingest
1507 2 : .put_rel_page_image(&mut m, TESTREL_A, 2, test_img("foo blk 2 at 5"), &ctx)
1508 2 : .await?;
1509 2 : m.commit(&ctx).await?;
1510 2 :
1511 2 : assert_current_logical_size(&tline, Lsn(0x50));
1512 2 :
1513 2 : // The relation was created at LSN 2, not visible at LSN 1 yet.
1514 2 : assert_eq!(
1515 2 : tline
1516 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x10)), &ctx)
1517 2 : .await?,
1518 2 : false
1519 2 : );
1520 2 : assert!(tline
1521 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x10)), &ctx)
1522 2 : .await
1523 2 : .is_err());
1524 2 : assert_eq!(
1525 2 : tline
1526 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x20)), &ctx)
1527 2 : .await?,
1528 2 : true
1529 2 : );
1530 2 : assert_eq!(
1531 2 : tline
1532 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x20)), &ctx)
1533 2 : .await?,
1534 2 : 1
1535 2 : );
1536 2 : assert_eq!(
1537 2 : tline
1538 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x50)), &ctx)
1539 2 : .await?,
1540 2 : 3
1541 2 : );
1542 2 :
1543 2 : // Check page contents at each LSN
1544 2 : assert_eq!(
1545 2 : tline
1546 2 : .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x20)), &ctx)
1547 2 : .await?,
1548 2 : test_img("foo blk 0 at 2")
1549 2 : );
1550 2 :
1551 2 : assert_eq!(
1552 2 : tline
1553 2 : .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x30)), &ctx)
1554 2 : .await?,
1555 2 : test_img("foo blk 0 at 3")
1556 2 : );
1557 2 :
1558 2 : assert_eq!(
1559 2 : tline
1560 2 : .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x40)), &ctx)
1561 2 : .await?,
1562 2 : test_img("foo blk 0 at 3")
1563 2 : );
1564 2 : assert_eq!(
1565 2 : tline
1566 2 : .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x40)), &ctx)
1567 2 : .await?,
1568 2 : test_img("foo blk 1 at 4")
1569 2 : );
1570 2 :
1571 2 : assert_eq!(
1572 2 : tline
1573 2 : .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x50)), &ctx)
1574 2 : .await?,
1575 2 : test_img("foo blk 0 at 3")
1576 2 : );
1577 2 : assert_eq!(
1578 2 : tline
1579 2 : .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x50)), &ctx)
1580 2 : .await?,
1581 2 : test_img("foo blk 1 at 4")
1582 2 : );
1583 2 : assert_eq!(
1584 2 : tline
1585 2 : .get_rel_page_at_lsn(TESTREL_A, 2, Version::Lsn(Lsn(0x50)), &ctx)
1586 2 : .await?,
1587 2 : test_img("foo blk 2 at 5")
1588 2 : );
1589 2 :
1590 2 : // Truncate last block
1591 2 : let mut m = tline.begin_modification(Lsn(0x60));
1592 2 : walingest
1593 2 : .put_rel_truncation(&mut m, TESTREL_A, 2, &ctx)
1594 2 : .await?;
1595 2 : m.commit(&ctx).await?;
1596 2 : assert_current_logical_size(&tline, Lsn(0x60));
1597 2 :
1598 2 : // Check reported size and contents after truncation
1599 2 : assert_eq!(
1600 2 : tline
1601 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x60)), &ctx)
1602 2 : .await?,
1603 2 : 2
1604 2 : );
1605 2 : assert_eq!(
1606 2 : tline
1607 2 : .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x60)), &ctx)
1608 2 : .await?,
1609 2 : test_img("foo blk 0 at 3")
1610 2 : );
1611 2 : assert_eq!(
1612 2 : tline
1613 2 : .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x60)), &ctx)
1614 2 : .await?,
1615 2 : test_img("foo blk 1 at 4")
1616 2 : );
1617 2 :
1618 2 : // should still see the truncated block with older LSN
1619 2 : assert_eq!(
1620 2 : tline
1621 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x50)), &ctx)
1622 2 : .await?,
1623 2 : 3
1624 2 : );
1625 2 : assert_eq!(
1626 2 : tline
1627 2 : .get_rel_page_at_lsn(TESTREL_A, 2, Version::Lsn(Lsn(0x50)), &ctx)
1628 2 : .await?,
1629 2 : test_img("foo blk 2 at 5")
1630 2 : );
1631 2 :
1632 2 : // Truncate to zero length
1633 2 : let mut m = tline.begin_modification(Lsn(0x68));
1634 2 : walingest
1635 2 : .put_rel_truncation(&mut m, TESTREL_A, 0, &ctx)
1636 2 : .await?;
1637 2 : m.commit(&ctx).await?;
1638 2 : assert_eq!(
1639 2 : tline
1640 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x68)), &ctx)
1641 2 : .await?,
1642 2 : 0
1643 2 : );
1644 2 :
1645 2 : // Extend from 0 to 2 blocks, leaving a gap
1646 2 : let mut m = tline.begin_modification(Lsn(0x70));
1647 2 : walingest
1648 2 : .put_rel_page_image(&mut m, TESTREL_A, 1, test_img("foo blk 1"), &ctx)
1649 2 : .await?;
1650 2 : m.commit(&ctx).await?;
1651 2 : assert_eq!(
1652 2 : tline
1653 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x70)), &ctx)
1654 2 : .await?,
1655 2 : 2
1656 2 : );
1657 2 : assert_eq!(
1658 2 : tline
1659 2 : .get_rel_page_at_lsn(TESTREL_A, 0, Version::Lsn(Lsn(0x70)), &ctx)
1660 2 : .await?,
1661 2 : ZERO_PAGE
1662 2 : );
1663 2 : assert_eq!(
1664 2 : tline
1665 2 : .get_rel_page_at_lsn(TESTREL_A, 1, Version::Lsn(Lsn(0x70)), &ctx)
1666 2 : .await?,
1667 2 : test_img("foo blk 1")
1668 2 : );
1669 2 :
1670 2 : // Extend a lot more, leaving a big gap that spans across segments
1671 2 : let mut m = tline.begin_modification(Lsn(0x80));
1672 2 : walingest
1673 2 : .put_rel_page_image(&mut m, TESTREL_A, 1500, test_img("foo blk 1500"), &ctx)
1674 2 : .await?;
1675 190 : m.commit(&ctx).await?;
1676 2 : assert_eq!(
1677 2 : tline
1678 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x80)), &ctx)
1679 2 : .await?,
1680 2 : 1501
1681 2 : );
1682 2998 : for blk in 2..1500 {
1683 2996 : assert_eq!(
1684 2996 : tline
1685 2996 : .get_rel_page_at_lsn(TESTREL_A, blk, Version::Lsn(Lsn(0x80)), &ctx)
1686 1540 : .await?,
1687 2996 : ZERO_PAGE
1688 2 : );
1689 2 : }
1690 2 : assert_eq!(
1691 2 : tline
1692 2 : .get_rel_page_at_lsn(TESTREL_A, 1500, Version::Lsn(Lsn(0x80)), &ctx)
1693 2 : .await?,
1694 2 : test_img("foo blk 1500")
1695 2 : );
1696 2 :
1697 2 : Ok(())
1698 2 : }
1699 :
1700 : // Test what happens if we dropped a relation
1701 : // and then created it again within the same layer.
1702 : #[tokio::test]
1703 2 : async fn test_drop_extend() -> Result<()> {
1704 2 : let (tenant, ctx) = TenantHarness::create("test_drop_extend")
1705 2 : .await?
1706 2 : .load()
1707 20 : .await;
1708 2 : let tline = tenant
1709 2 : .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
1710 6 : .await?;
1711 5 : let mut walingest = init_walingest_test(&tline, &ctx).await?;
1712 2 :
1713 2 : let mut m = tline.begin_modification(Lsn(0x20));
1714 2 : walingest
1715 2 : .put_rel_page_image(&mut m, TESTREL_A, 0, test_img("foo blk 0 at 2"), &ctx)
1716 2 : .await?;
1717 2 : m.commit(&ctx).await?;
1718 2 :
1719 2 : // Check that rel exists and size is correct
1720 2 : assert_eq!(
1721 2 : tline
1722 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x20)), &ctx)
1723 2 : .await?,
1724 2 : true
1725 2 : );
1726 2 : assert_eq!(
1727 2 : tline
1728 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x20)), &ctx)
1729 2 : .await?,
1730 2 : 1
1731 2 : );
1732 2 :
1733 2 : // Drop rel
1734 2 : let mut m = tline.begin_modification(Lsn(0x30));
1735 2 : let mut rel_drops = HashMap::new();
1736 2 : rel_drops.insert((TESTREL_A.spcnode, TESTREL_A.dbnode), vec![TESTREL_A]);
1737 2 : m.put_rel_drops(rel_drops, &ctx).await?;
1738 2 : m.commit(&ctx).await?;
1739 2 :
1740 2 : // Check that rel is not visible anymore
1741 2 : assert_eq!(
1742 2 : tline
1743 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x30)), &ctx)
1744 2 : .await?,
1745 2 : false
1746 2 : );
1747 2 :
1748 2 : // FIXME: should fail
1749 2 : //assert!(tline.get_rel_size(TESTREL_A, Lsn(0x30), false)?.is_none());
1750 2 :
1751 2 : // Re-create it
1752 2 : let mut m = tline.begin_modification(Lsn(0x40));
1753 2 : walingest
1754 2 : .put_rel_page_image(&mut m, TESTREL_A, 0, test_img("foo blk 0 at 4"), &ctx)
1755 2 : .await?;
1756 2 : m.commit(&ctx).await?;
1757 2 :
1758 2 : // Check that rel exists and size is correct
1759 2 : assert_eq!(
1760 2 : tline
1761 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x40)), &ctx)
1762 2 : .await?,
1763 2 : true
1764 2 : );
1765 2 : assert_eq!(
1766 2 : tline
1767 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x40)), &ctx)
1768 2 : .await?,
1769 2 : 1
1770 2 : );
1771 2 :
1772 2 : Ok(())
1773 2 : }
1774 :
1775 : // Test what happens if we truncated a relation
1776 : // so that one of its segments was dropped
1777 : // and then extended it again within the same layer.
1778 : #[tokio::test]
1779 2 : async fn test_truncate_extend() -> Result<()> {
1780 2 : let (tenant, ctx) = TenantHarness::create("test_truncate_extend")
1781 2 : .await?
1782 2 : .load()
1783 20 : .await;
1784 2 : let tline = tenant
1785 2 : .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
1786 6 : .await?;
1787 5 : let mut walingest = init_walingest_test(&tline, &ctx).await?;
1788 2 :
1789 2 : // Create a 20 MB relation (the size is arbitrary)
1790 2 : let relsize = 20 * 1024 * 1024 / 8192;
1791 2 : let mut m = tline.begin_modification(Lsn(0x20));
1792 5120 : for blkno in 0..relsize {
1793 5120 : let data = format!("foo blk {} at {}", blkno, Lsn(0x20));
1794 5120 : walingest
1795 5120 : .put_rel_page_image(&mut m, TESTREL_A, blkno, test_img(&data), &ctx)
1796 2 : .await?;
1797 2 : }
1798 2 : m.commit(&ctx).await?;
1799 2 :
1800 2 : // The relation was created at LSN 20, not visible at LSN 1 yet.
1801 2 : assert_eq!(
1802 2 : tline
1803 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x10)), &ctx)
1804 2 : .await?,
1805 2 : false
1806 2 : );
1807 2 : assert!(tline
1808 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x10)), &ctx)
1809 2 : .await
1810 2 : .is_err());
1811 2 :
1812 2 : assert_eq!(
1813 2 : tline
1814 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x20)), &ctx)
1815 2 : .await?,
1816 2 : true
1817 2 : );
1818 2 : assert_eq!(
1819 2 : tline
1820 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x20)), &ctx)
1821 2 : .await?,
1822 2 : relsize
1823 2 : );
1824 2 :
1825 2 : // Check relation content
1826 5120 : for blkno in 0..relsize {
1827 5120 : let lsn = Lsn(0x20);
1828 5120 : let data = format!("foo blk {} at {}", blkno, lsn);
1829 5120 : assert_eq!(
1830 5120 : tline
1831 5120 : .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(lsn), &ctx)
1832 1803 : .await?,
1833 5120 : test_img(&data)
1834 2 : );
1835 2 : }
1836 2 :
1837 2 : // Truncate relation so that second segment was dropped
1838 2 : // - only leave one page
1839 2 : let mut m = tline.begin_modification(Lsn(0x60));
1840 2 : walingest
1841 2 : .put_rel_truncation(&mut m, TESTREL_A, 1, &ctx)
1842 2 : .await?;
1843 2 : m.commit(&ctx).await?;
1844 2 :
1845 2 : // Check reported size and contents after truncation
1846 2 : assert_eq!(
1847 2 : tline
1848 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x60)), &ctx)
1849 2 : .await?,
1850 2 : 1
1851 2 : );
1852 2 :
1853 4 : for blkno in 0..1 {
1854 2 : let lsn = Lsn(0x20);
1855 2 : let data = format!("foo blk {} at {}", blkno, lsn);
1856 2 : assert_eq!(
1857 2 : tline
1858 2 : .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(Lsn(0x60)), &ctx)
1859 2 : .await?,
1860 2 : test_img(&data)
1861 2 : );
1862 2 : }
1863 2 :
1864 2 : // should still see all blocks with older LSN
1865 2 : assert_eq!(
1866 2 : tline
1867 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x50)), &ctx)
1868 2 : .await?,
1869 2 : relsize
1870 2 : );
1871 5120 : for blkno in 0..relsize {
1872 5120 : let lsn = Lsn(0x20);
1873 5120 : let data = format!("foo blk {} at {}", blkno, lsn);
1874 5120 : assert_eq!(
1875 5120 : tline
1876 5120 : .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(Lsn(0x50)), &ctx)
1877 1856 : .await?,
1878 5120 : test_img(&data)
1879 2 : );
1880 2 : }
1881 2 :
1882 2 : // Extend relation again.
1883 2 : // Add enough blocks to create second segment
1884 2 : let lsn = Lsn(0x80);
1885 2 : let mut m = tline.begin_modification(lsn);
1886 5120 : for blkno in 0..relsize {
1887 5120 : let data = format!("foo blk {} at {}", blkno, lsn);
1888 5120 : walingest
1889 5120 : .put_rel_page_image(&mut m, TESTREL_A, blkno, test_img(&data), &ctx)
1890 2 : .await?;
1891 2 : }
1892 3 : m.commit(&ctx).await?;
1893 2 :
1894 2 : assert_eq!(
1895 2 : tline
1896 2 : .get_rel_exists(TESTREL_A, Version::Lsn(Lsn(0x80)), &ctx)
1897 2 : .await?,
1898 2 : true
1899 2 : );
1900 2 : assert_eq!(
1901 2 : tline
1902 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(0x80)), &ctx)
1903 2 : .await?,
1904 2 : relsize
1905 2 : );
1906 2 : // Check relation content
1907 5120 : for blkno in 0..relsize {
1908 5120 : let lsn = Lsn(0x80);
1909 5120 : let data = format!("foo blk {} at {}", blkno, lsn);
1910 5120 : assert_eq!(
1911 5120 : tline
1912 5120 : .get_rel_page_at_lsn(TESTREL_A, blkno, Version::Lsn(Lsn(0x80)), &ctx)
1913 1828 : .await?,
1914 5120 : test_img(&data)
1915 2 : );
1916 2 : }
1917 2 :
1918 2 : Ok(())
1919 2 : }
1920 :
1921 : /// Test get_relsize() and truncation with a file larger than 1 GB, so that it's
1922 : /// split into multiple 1 GB segments in Postgres.
1923 : #[tokio::test]
1924 2 : async fn test_large_rel() -> Result<()> {
1925 20 : let (tenant, ctx) = TenantHarness::create("test_large_rel").await?.load().await;
1926 2 : let tline = tenant
1927 2 : .create_test_timeline(TIMELINE_ID, Lsn(8), DEFAULT_PG_VERSION, &ctx)
1928 6 : .await?;
1929 5 : let mut walingest = init_walingest_test(&tline, &ctx).await?;
1930 2 :
1931 2 : let mut lsn = 0x10;
1932 262146 : for blknum in 0..RELSEG_SIZE + 1 {
1933 262146 : lsn += 0x10;
1934 262146 : let mut m = tline.begin_modification(Lsn(lsn));
1935 262146 : let img = test_img(&format!("foo blk {} at {}", blknum, Lsn(lsn)));
1936 262146 : walingest
1937 262146 : .put_rel_page_image(&mut m, TESTREL_A, blknum as BlockNumber, img, &ctx)
1938 5420 : .await?;
1939 262146 : m.commit(&ctx).await?;
1940 2 : }
1941 2 :
1942 2 : assert_current_logical_size(&tline, Lsn(lsn));
1943 2 :
1944 2 : assert_eq!(
1945 2 : tline
1946 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(lsn)), &ctx)
1947 2 : .await?,
1948 2 : RELSEG_SIZE + 1
1949 2 : );
1950 2 :
1951 2 : // Truncate one block
1952 2 : lsn += 0x10;
1953 2 : let mut m = tline.begin_modification(Lsn(lsn));
1954 2 : walingest
1955 2 : .put_rel_truncation(&mut m, TESTREL_A, RELSEG_SIZE, &ctx)
1956 2 : .await?;
1957 2 : m.commit(&ctx).await?;
1958 2 : assert_eq!(
1959 2 : tline
1960 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(lsn)), &ctx)
1961 2 : .await?,
1962 2 : RELSEG_SIZE
1963 2 : );
1964 2 : assert_current_logical_size(&tline, Lsn(lsn));
1965 2 :
1966 2 : // Truncate another block
1967 2 : lsn += 0x10;
1968 2 : let mut m = tline.begin_modification(Lsn(lsn));
1969 2 : walingest
1970 2 : .put_rel_truncation(&mut m, TESTREL_A, RELSEG_SIZE - 1, &ctx)
1971 2 : .await?;
1972 2 : m.commit(&ctx).await?;
1973 2 : assert_eq!(
1974 2 : tline
1975 2 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(lsn)), &ctx)
1976 2 : .await?,
1977 2 : RELSEG_SIZE - 1
1978 2 : );
1979 2 : assert_current_logical_size(&tline, Lsn(lsn));
1980 2 :
1981 2 : // Truncate to 1500, and then truncate all the way down to 0, one block at a time
1982 2 : // This tests the behavior at segment boundaries
1983 2 : let mut size: i32 = 3000;
1984 6004 : while size >= 0 {
1985 6002 : lsn += 0x10;
1986 6002 : let mut m = tline.begin_modification(Lsn(lsn));
1987 6002 : walingest
1988 6002 : .put_rel_truncation(&mut m, TESTREL_A, size as BlockNumber, &ctx)
1989 2 : .await?;
1990 6002 : m.commit(&ctx).await?;
1991 6002 : assert_eq!(
1992 6002 : tline
1993 6002 : .get_rel_size(TESTREL_A, Version::Lsn(Lsn(lsn)), &ctx)
1994 2 : .await?,
1995 6002 : size as BlockNumber
1996 2 : );
1997 2 :
1998 6002 : size -= 1;
1999 2 : }
2000 2 : assert_current_logical_size(&tline, Lsn(lsn));
2001 2 :
2002 2 : Ok(())
2003 2 : }
2004 :
2005 : /// Replay a wal segment file taken directly from safekeepers.
2006 : ///
2007 : /// This test is useful for benchmarking since it allows us to profile only
2008 : /// the walingest code in a single-threaded executor, and iterate more quickly
2009 : /// without waiting for unrelated steps.
2010 : #[tokio::test]
2011 2 : async fn test_ingest_real_wal() {
2012 2 : use crate::tenant::harness::*;
2013 2 : use postgres_ffi::waldecoder::WalStreamDecoder;
2014 2 : use postgres_ffi::WAL_SEGMENT_SIZE;
2015 2 :
2016 2 : // Define test data path and constants.
2017 2 : //
2018 2 : // Steps to reconstruct the data, if needed:
2019 2 : // 1. Run the pgbench python test
2020 2 : // 2. Take the first wal segment file from safekeeper
2021 2 : // 3. Compress it using `zstd --long input_file`
2022 2 : // 4. Copy initdb.tar.zst from local_fs_remote_storage
2023 2 : // 5. Grep sk logs for "restart decoder" to get startpoint
2024 2 : // 6. Run just the decoder from this test to get the endpoint.
2025 2 : // It's the last LSN the decoder will output.
2026 2 : let pg_version = 15; // The test data was generated by pg15
2027 2 : let path = "test_data/sk_wal_segment_from_pgbench";
2028 2 : let wal_segment_path = format!("{path}/000000010000000000000001.zst");
2029 2 : let source_initdb_path = format!("{path}/{INITDB_PATH}");
2030 2 : let startpoint = Lsn::from_hex("14AEC08").unwrap();
2031 2 : let _endpoint = Lsn::from_hex("1FFFF98").unwrap();
2032 2 :
2033 2 : let harness = TenantHarness::create("test_ingest_real_wal").await.unwrap();
2034 2 : let span = harness
2035 2 : .span()
2036 2 : .in_scope(|| info_span!("timeline_span", timeline_id=%TIMELINE_ID));
2037 20 : let (tenant, ctx) = harness.load().await;
2038 2 :
2039 2 : let remote_initdb_path =
2040 2 : remote_initdb_archive_path(&tenant.tenant_shard_id().tenant_id, &TIMELINE_ID);
2041 2 : let initdb_path = harness.remote_fs_dir.join(remote_initdb_path.get_path());
2042 2 :
2043 2 : std::fs::create_dir_all(initdb_path.parent().unwrap())
2044 2 : .expect("creating test dir should work");
2045 2 : std::fs::copy(source_initdb_path, initdb_path).expect("copying the initdb.tar.zst works");
2046 2 :
2047 2 : // Bootstrap a real timeline. We can't use create_test_timeline because
2048 2 : // it doesn't create a real checkpoint, and Walingest::new tries to parse
2049 2 : // the garbage data.
2050 2 : let tline = tenant
2051 2 : .bootstrap_timeline_test(TIMELINE_ID, pg_version, Some(TIMELINE_ID), &ctx)
2052 21296 : .await
2053 2 : .unwrap();
2054 2 :
2055 2 : // We fully read and decompress this into memory before decoding
2056 2 : // to get a more accurate perf profile of the decoder.
2057 2 : let bytes = {
2058 2 : use async_compression::tokio::bufread::ZstdDecoder;
2059 2 : let file = tokio::fs::File::open(wal_segment_path).await.unwrap();
2060 2 : let reader = tokio::io::BufReader::new(file);
2061 2 : let decoder = ZstdDecoder::new(reader);
2062 2 : let mut reader = tokio::io::BufReader::new(decoder);
2063 2 : let mut buffer = Vec::new();
2064 224 : tokio::io::copy_buf(&mut reader, &mut buffer).await.unwrap();
2065 2 : buffer
2066 2 : };
2067 2 :
2068 2 : // TODO start a profiler too
2069 2 : let started_at = std::time::Instant::now();
2070 2 :
2071 2 : // Initialize walingest
2072 2 : let xlogoff: usize = startpoint.segment_offset(WAL_SEGMENT_SIZE);
2073 2 : let mut decoder = WalStreamDecoder::new(startpoint, pg_version);
2074 2 : let mut walingest = WalIngest::new(tline.as_ref(), startpoint, &ctx)
2075 5 : .await
2076 2 : .unwrap();
2077 2 : let mut modification = tline.begin_modification(startpoint);
2078 2 : println!("decoding {} bytes", bytes.len() - xlogoff);
2079 2 :
2080 2 : // Decode and ingest wal. We process the wal in chunks because
2081 2 : // that's what happens when we get bytes from safekeepers.
2082 474686 : for chunk in bytes[xlogoff..].chunks(50) {
2083 474686 : decoder.feed_bytes(chunk);
2084 620536 : while let Some((lsn, recdata)) = decoder.poll_decode().unwrap() {
2085 145850 : let interpreted = InterpretedWalRecord::from_bytes_filtered(
2086 145850 : recdata,
2087 145850 : modification.tline.get_shard_identity(),
2088 145850 : lsn,
2089 145850 : modification.tline.pg_version,
2090 145850 : )
2091 145850 : .unwrap();
2092 145850 :
2093 145850 : walingest
2094 145850 : .ingest_record(interpreted, &mut modification, &ctx)
2095 145850 : .instrument(span.clone())
2096 296 : .await
2097 145850 : .unwrap();
2098 2 : }
2099 474686 : modification.commit(&ctx).await.unwrap();
2100 2 : }
2101 2 :
2102 2 : let duration = started_at.elapsed();
2103 2 : println!("done in {:?}", duration);
2104 2 : }
2105 : }
|
