shardlog 0.2.0

Embedded crash-recoverable append log with durable extent packs
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
use std::fs::{self, File, OpenOptions};
use std::io::{Read, Write};
use std::path::{Component, Path, PathBuf};
use std::sync::atomic::{AtomicU64, Ordering};
use std::time::{SystemTime, UNIX_EPOCH};

use crate::{LogicalOffset, Reservation, ShardId, TopicPartition};
use fs2::FileExt;
use serde::{Deserialize, Serialize};

use crate::{StorageError, StorageResult};

const MANIFEST_VERSION: u8 = 3;
const CHECKSUM_BLAKE3: &str = "blake3";
const COPY_BUFFER_BYTES: usize = 1024 * 1024;
const MAX_MANIFEST_BYTES: u64 = 64 * 1024 * 1024;
static TEMP_SEQUENCE: AtomicU64 = AtomicU64::new(0);

#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct TierObject {
    pub pack_sequence: u64,
    pub object_key: String,
    pub bytes: u64,
    pub checksum_algorithm: String,
    pub checksum: String,
    pub created_unix_ms: u64,
    pub extents: Vec<TierExtent>,
}

#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct TierExtent {
    pub topic_id: String,
    pub partition_id: u32,
    pub first_offset: u64,
    pub last_offset: u64,
    pub virtual_lane_id: u32,
    pub ring_epoch: u64,
    pub leader_epoch: u64,
}

impl TierExtent {
    #[must_use]
    pub fn from_reservation(reservation: Reservation) -> Self {
        Self {
            topic_id: reservation.topic_id.to_string(),
            partition_id: reservation.partition_id.get(),
            first_offset: reservation.first_offset.get(),
            last_offset: reservation.last_offset.get(),
            virtual_lane_id: reservation.placement.virtual_lane_id.get(),
            ring_epoch: reservation.placement.ring_epoch.get(),
            leader_epoch: reservation.placement.leader_epoch.get(),
        }
    }
}

#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct TierManifest {
    pub format_version: u8,
    pub generation: u64,
    pub shard_id: u32,
    pub objects: Vec<TierObject>,
}

impl TierManifest {
    fn empty(shard_id: ShardId) -> Self {
        Self {
            format_version: MANIFEST_VERSION,
            generation: 0,
            shard_id: shard_id.get(),
            objects: Vec::new(),
        }
    }

    fn validate(&self, shard_id: ShardId) -> StorageResult<()> {
        if self.format_version != MANIFEST_VERSION {
            return Err(tier_corrupt(format!(
                "unsupported tier manifest version {}",
                self.format_version
            )));
        }
        if self.shard_id != shard_id.get() {
            return Err(tier_corrupt(format!(
                "tier manifest belongs to shard {}, expected {}",
                self.shard_id, shard_id
            )));
        }
        let mut previous = None;
        for object in &self.objects {
            validate_object_key(&object.object_key)?;
            if object.bytes == 0
                || object.checksum.len() != 64
                || object.checksum_algorithm != CHECKSUM_BLAKE3
                || object.extents.is_empty()
            {
                return Err(tier_corrupt(
                    "tier manifest contains invalid object metadata",
                ));
            }
            for extent in &object.extents {
                if extent.topic_id.parse::<u128>().is_err()
                    || extent.first_offset > extent.last_offset
                {
                    return Err(tier_corrupt(
                        "tier manifest contains invalid extent coverage",
                    ));
                }
            }
            if previous.is_some_and(|sequence| sequence >= object.pack_sequence) {
                return Err(tier_corrupt(
                    "tier manifest pack sequences are not strictly increasing",
                ));
            }
            previous = Some(object.pack_sequence);
        }
        Ok(())
    }
}

#[derive(Debug)]
pub struct LocalObjectTier {
    root: PathBuf,
    shard_id: ShardId,
    manifest: TierManifest,
}

impl LocalObjectTier {
    pub fn open(root: impl AsRef<Path>, shard_id: ShardId) -> StorageResult<Self> {
        let root = root.as_ref().to_path_buf();
        fs::create_dir_all(root.join("objects").join(shard_directory(shard_id)))?;
        fs::create_dir_all(root.join("manifests").join(shard_directory(shard_id)))?;
        let manifest = load_current_manifest(&root, shard_id)?;
        manifest.validate(shard_id)?;
        for object in &manifest.objects {
            verify_object(&root, object)?;
        }
        Ok(Self {
            root,
            shard_id,
            manifest,
        })
    }

    #[must_use]
    pub fn manifest(&self) -> &TierManifest {
        &self.manifest
    }

    pub fn offload_pack(
        &mut self,
        pack_sequence: u64,
        source: impl AsRef<Path>,
        extents: Vec<TierExtent>,
    ) -> StorageResult<TierObject> {
        if extents.is_empty() {
            return Err(StorageError::InvalidInput(
                "object manifest coverage cannot be empty".into(),
            ));
        }
        if let Some(existing) = self
            .manifest
            .objects
            .iter()
            .find(|object| object.pack_sequence == pack_sequence)
        {
            let observed = hash_file(source.as_ref())?;
            if existing.bytes != observed.0 || existing.checksum != observed.1 {
                return Err(tier_corrupt(format!(
                    "sealed pack {pack_sequence} differs from its authoritative tier manifest"
                )));
            }
            verify_object(&self.root, existing)?;
            if existing.extents != extents {
                return Err(tier_corrupt(format!(
                    "sealed pack {pack_sequence} has different extent coverage"
                )));
            }
            return Ok(existing.clone());
        }
        if self
            .manifest
            .objects
            .last()
            .is_some_and(|object| object.pack_sequence >= pack_sequence)
        {
            return Err(StorageError::InvalidInput(
                "tier packs must be published in increasing sequence order".into(),
            ));
        }

        let source = source.as_ref();
        let metadata = source.metadata()?;
        if !metadata.is_file() || metadata.len() == 0 {
            return Err(StorageError::InvalidInput(
                "only nonempty sealed pack files can be offloaded".into(),
            ));
        }
        let object_key = format!(
            "objects/{}/pack-{pack_sequence:020}.sse",
            shard_directory(self.shard_id)
        );
        let destination = object_path(&self.root, &object_key)?;
        let (bytes, checksum) = copy_object_atomically(source, &destination)?;
        let object = TierObject {
            pack_sequence,
            object_key,
            bytes,
            checksum_algorithm: CHECKSUM_BLAKE3.into(),
            checksum,
            created_unix_ms: now_unix_ms()?,
            extents,
        };
        verify_object(&self.root, &object)?;

        let mut next = self.manifest.clone();
        next.generation = next.generation.checked_add(1).ok_or_else(|| {
            StorageError::LimitExceeded("tier manifest generation exhausted".into())
        })?;
        next.objects.push(object.clone());
        publish_manifest(&self.root, self.shard_id, self.manifest.generation, &next)?;
        self.manifest = next;
        Ok(object)
    }

    /// Returns the exclusive object-backed high watermark for a partition
    /// within this physical-lane manifest. Callers combine all lane manifests
    /// before using the result as a partition-wide watermark.
    #[must_use]
    pub fn covered_through(&self, topic_partition: TopicPartition) -> LogicalOffset {
        let mut ranges = self
            .manifest
            .objects
            .iter()
            .flat_map(|object| &object.extents)
            .filter(|extent| {
                extent.topic_id == topic_partition.topic_id.to_string()
                    && extent.partition_id == topic_partition.partition_id.get()
            })
            .map(|extent| (extent.first_offset, extent.last_offset))
            .collect::<Vec<_>>();
        ranges.sort_unstable();
        let mut end = 0u64;
        for (first, last) in ranges {
            if first > end {
                break;
            }
            end = end.max(last.saturating_add(1));
        }
        LogicalOffset::new(end)
    }
}

fn load_current_manifest(root: &Path, shard_id: ShardId) -> StorageResult<TierManifest> {
    let current = current_path(root, shard_id);
    let generation = match fs::read_to_string(&current) {
        Ok(value) => value
            .trim()
            .parse::<u64>()
            .map_err(|_| tier_corrupt("invalid tier CURRENT generation"))?,
        Err(error) if error.kind() == std::io::ErrorKind::NotFound => {
            return Ok(TierManifest::empty(shard_id));
        }
        Err(error) => return Err(error.into()),
    };
    let path = manifest_path(root, shard_id, generation);
    let metadata = path.metadata()?;
    if metadata.len() > MAX_MANIFEST_BYTES {
        return Err(StorageError::LimitExceeded(
            "tier manifest exceeds the read limit".into(),
        ));
    }
    let bytes = fs::read(path)?;
    let manifest: TierManifest = serde_json::from_slice(&bytes)
        .map_err(|error| tier_corrupt(format!("invalid tier manifest: {error}")))?;
    if manifest.generation != generation {
        return Err(tier_corrupt(
            "tier CURRENT and manifest generation disagree",
        ));
    }
    Ok(manifest)
}

fn publish_manifest(
    root: &Path,
    shard_id: ShardId,
    expected_generation: u64,
    manifest: &TierManifest,
) -> StorageResult<()> {
    manifest.validate(shard_id)?;
    let _lock = ManifestUpdateLock::acquire(root, shard_id)?;
    let observed = load_current_manifest(root, shard_id)?;
    if observed.generation != expected_generation
        || manifest.generation != expected_generation.saturating_add(1)
    {
        return Err(StorageError::InvalidInput(format!(
            "conditional manifest update failed: expected generation {expected_generation}, observed {}",
            observed.generation
        )));
    }
    let encoded = serde_json::to_vec(manifest).map_err(|error| {
        StorageError::InvalidInput(format!("manifest encoding failed: {error}"))
    })?;
    if encoded.len() as u64 > MAX_MANIFEST_BYTES {
        return Err(StorageError::LimitExceeded(
            "tier manifest exceeds the write limit".into(),
        ));
    }
    let path = manifest_path(root, shard_id, manifest.generation);
    match OpenOptions::new().create_new(true).write(true).open(&path) {
        Ok(mut file) => {
            file.write_all(&encoded)?;
            file.sync_all()?;
            sync_directory(path.parent().expect("manifest has parent"))?;
        }
        Err(error) if error.kind() == std::io::ErrorKind::AlreadyExists => {
            if fs::read(&path)? != encoded {
                return Err(StorageError::corrupt(
                    &path,
                    0,
                    "immutable manifest generation has different content",
                ));
            }
        }
        Err(error) => return Err(error.into()),
    }

    let current = current_path(root, shard_id);
    let temporary = temporary_path(&current);
    let mut pointer = OpenOptions::new()
        .create_new(true)
        .write(true)
        .open(&temporary)?;
    writeln!(pointer, "{}", manifest.generation)?;
    pointer.sync_all()?;
    fs::rename(&temporary, &current)?;
    sync_directory(current.parent().expect("CURRENT has parent"))?;
    Ok(())
}

struct ManifestUpdateLock {
    file: File,
}

impl ManifestUpdateLock {
    fn acquire(root: &Path, shard_id: ShardId) -> StorageResult<Self> {
        let path = root
            .join("manifests")
            .join(shard_directory(shard_id))
            .join("UPDATE.lock");
        let file = OpenOptions::new()
            .create(true)
            .truncate(false)
            .read(true)
            .write(true)
            .open(path)?;
        file.lock_exclusive()?;
        Ok(Self { file })
    }
}

impl Drop for ManifestUpdateLock {
    fn drop(&mut self) {
        let _ = FileExt::unlock(&self.file);
    }
}

fn copy_object_atomically(source: &Path, destination: &Path) -> StorageResult<(u64, String)> {
    if destination.exists() {
        let source_digest = hash_file(source)?;
        let destination_digest = hash_file(destination)?;
        if source_digest != destination_digest {
            return Err(StorageError::corrupt(
                destination,
                0,
                "immutable object already exists with different content",
            ));
        }
        return Ok(destination_digest);
    }
    let temporary = temporary_path(destination);
    let result = (|| {
        let mut input = File::open(source)?;
        let mut output = OpenOptions::new()
            .create_new(true)
            .write(true)
            .open(&temporary)?;
        let mut digest = blake3::Hasher::new();
        let mut bytes = 0u64;
        let mut buffer = vec![0u8; COPY_BUFFER_BYTES];
        loop {
            let read = input.read(&mut buffer)?;
            if read == 0 {
                break;
            }
            output.write_all(&buffer[..read])?;
            digest.update(&buffer[..read]);
            bytes = bytes
                .checked_add(read as u64)
                .ok_or_else(|| StorageError::LimitExceeded("object size overflow".into()))?;
        }
        output.sync_all()?;
        fs::rename(&temporary, destination)?;
        sync_directory(destination.parent().expect("object has parent"))?;
        Ok((bytes, digest.finalize().to_hex().to_string()))
    })();
    if result.is_err() {
        let _ = fs::remove_file(&temporary);
    }
    result
}

fn verify_object(root: &Path, object: &TierObject) -> StorageResult<()> {
    let path = object_path(root, &object.object_key)?;
    let (bytes, checksum) = hash_file(&path)?;
    if bytes != object.bytes || checksum != object.checksum {
        return Err(StorageError::corrupt(
            &path,
            0,
            "tier object length or checksum mismatch",
        ));
    }
    Ok(())
}

fn hash_file(path: &Path) -> StorageResult<(u64, String)> {
    let mut file = File::open(path)?;
    let mut digest = blake3::Hasher::new();
    let mut bytes = 0u64;
    let mut buffer = vec![0u8; COPY_BUFFER_BYTES];
    loop {
        let read = file.read(&mut buffer)?;
        if read == 0 {
            break;
        }
        digest.update(&buffer[..read]);
        bytes = checked_object_bytes(bytes, read)?;
    }
    Ok((bytes, digest.finalize().to_hex().to_string()))
}

fn checked_object_bytes(current: u64, read: usize) -> StorageResult<u64> {
    current
        .checked_add(read as u64)
        .ok_or_else(|| StorageError::LimitExceeded("object size overflow".into()))
}

fn object_path(root: &Path, key: &str) -> StorageResult<PathBuf> {
    validate_object_key(key)?;
    Ok(root.join(key))
}

fn validate_object_key(key: &str) -> StorageResult<()> {
    if key.is_empty()
        || Path::new(key)
            .components()
            .any(|component| !matches!(component, Component::Normal(_)))
    {
        return Err(tier_corrupt("tier manifest contains an unsafe object key"));
    }
    Ok(())
}

fn manifest_path(root: &Path, shard_id: ShardId, generation: u64) -> PathBuf {
    root.join("manifests")
        .join(shard_directory(shard_id))
        .join(format!("manifest-{generation:020}.json"))
}

fn current_path(root: &Path, shard_id: ShardId) -> PathBuf {
    root.join("manifests")
        .join(shard_directory(shard_id))
        .join("CURRENT")
}

fn shard_directory(shard_id: ShardId) -> String {
    format!("shard-{}", shard_id.get())
}

fn temporary_path(path: &Path) -> PathBuf {
    let sequence = TEMP_SEQUENCE.fetch_add(1, Ordering::Relaxed);
    let file_name = path
        .file_name()
        .and_then(|name| name.to_str())
        .unwrap_or("tier");
    path.with_file_name(format!(
        ".{file_name}.{}.{}.tmp",
        std::process::id(),
        sequence
    ))
}

fn sync_directory(path: &Path) -> StorageResult<()> {
    File::open(path)?.sync_all()?;
    Ok(())
}

fn now_unix_ms() -> StorageResult<u64> {
    let millis = SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .map_err(|error| StorageError::InvalidInput(format!("system clock before epoch: {error}")))?
        .as_millis();
    u64::try_from(millis)
        .map_err(|_| StorageError::LimitExceeded("system time exceeds u64 milliseconds".into()))
}

fn tier_corrupt(reason: impl Into<String>) -> StorageError {
    StorageError::Corrupt {
        path: "tier-manifest".into(),
        offset: 0,
        reason: reason.into(),
    }
}

#[cfg(test)]
mod tests {
    use std::fs;
    use std::time::{SystemTime, UNIX_EPOCH};

    use super::*;

    fn coverage() -> Vec<TierExtent> {
        vec![TierExtent {
            topic_id: "1".into(),
            partition_id: 0,
            first_offset: 0,
            last_offset: 0,
            virtual_lane_id: 0,
            ring_epoch: 1,
            leader_epoch: 1,
        }]
    }

    struct TempDir(PathBuf);

    impl TempDir {
        fn new() -> Self {
            let nonce = SystemTime::now()
                .duration_since(UNIX_EPOCH)
                .expect("clock")
                .as_nanos();
            let sequence = TEMP_SEQUENCE.fetch_add(1, Ordering::Relaxed);
            let path = std::env::temp_dir().join(format!(
                "shard-stream-tier-{}-{nonce}-{sequence}",
                std::process::id()
            ));
            fs::create_dir_all(&path).expect("temp dir");
            Self(path)
        }
    }

    impl Drop for TempDir {
        fn drop(&mut self) {
            let _ = fs::remove_dir_all(&self.0);
        }
    }

    #[test]
    fn publishes_immutable_objects_and_recovers_from_current_manifest() {
        let temp = TempDir::new();
        let source = temp.0.join("source.sse");
        fs::write(&source, b"sealed pack bytes").expect("source");
        let root = temp.0.join("tier");
        let mut tier = LocalObjectTier::open(&root, ShardId::new(2)).expect("open");
        let object = tier.offload_pack(7, &source, coverage()).expect("offload");
        assert_eq!(object.bytes, 17);
        assert_eq!(tier.manifest().generation, 1);

        let reopened = LocalObjectTier::open(&root, ShardId::new(2)).expect("reopen");
        assert_eq!(reopened.manifest(), tier.manifest());
        assert_eq!(reopened.manifest().objects, vec![object]);
    }

    #[test]
    fn rejects_object_corruption_on_recovery() {
        let temp = TempDir::new();
        let source = temp.0.join("source.sse");
        fs::write(&source, b"sealed pack bytes").expect("source");
        let root = temp.0.join("tier");
        let mut tier = LocalObjectTier::open(&root, ShardId::new(0)).expect("open");
        let object = tier.offload_pack(0, &source, coverage()).expect("offload");
        fs::write(root.join(object.object_key), b"corrupt").expect("corrupt");
        assert!(matches!(
            LocalObjectTier::open(&root, ShardId::new(0)),
            Err(StorageError::Corrupt { .. })
        ));
    }

    #[test]
    fn conditional_manifest_publish_rejects_a_stale_writer() {
        let temp = TempDir::new();
        let root = temp.0.join("tier");
        let first_source = temp.0.join("first.sse");
        let second_source = temp.0.join("second.sse");
        fs::write(&first_source, b"first sealed pack").expect("first");
        fs::write(&second_source, b"second sealed pack").expect("second");
        let mut first = LocalObjectTier::open(&root, ShardId::new(0)).expect("first open");
        let mut stale = LocalObjectTier::open(&root, ShardId::new(0)).expect("stale open");
        first
            .offload_pack(0, &first_source, coverage())
            .expect("first publish");
        assert!(matches!(
            stale.offload_pack(1, &second_source, coverage()),
            Err(StorageError::InvalidInput(_))
        ));
    }
}