kevy_persist/aof.rs
1//! Append-only command log. Split out from `lib.rs` to keep that file
2//! under the 500-LOC house rule; the snapshot writer/reader stays there.
3
4use std::fs::{File, OpenOptions};
5use std::io::{self, BufWriter, Seek, SeekFrom, Write};
6use std::path::{Path, PathBuf};
7use std::time::{Duration, Instant};
8
9use kevy_resp::ArgvView;
10use kevy_store::Store;
11
12use crate::{
13 dump_store_to_aof, estimate_multibulk_bytes, write_multibulk,
14};
15
16/// 9-byte file-format header written at the start of every kevy-managed
17/// AOF as of v1.2.0. `replay_aof` strips it before parsing RESP, so
18/// non-kevy bytes accidentally written into the AOF path (e.g. a deploy
19/// pipeline redirecting shell stderr into the file) get the same loud
20/// rejection as any other corrupt frame. Pre-1.2 AOFs (no magic) still
21/// replay — the parser only consumes the magic if it sees it.
22pub(crate) const AOF_MAGIC: &[u8; 9] = b"KEVYAOF1\n";
23
24/// When to fsync the AOF to disk.
25#[derive(Debug, Clone, Copy, PartialEq, Eq)]
26pub enum Fsync {
27 /// fsync after every write — safest, slowest.
28 Always,
29 /// fsync at most once per second (call [`Aof::maybe_sync`] periodically).
30 EverySec,
31 /// Never fsync explicitly; leave it to the OS.
32 No,
33}
34
35/// An append-only command log. Each write command is appended as a RESP
36/// multi-bulk frame; [`crate::replay_aof`] re-applies them on startup.
37///
38/// Durability model (paired with snapshots): a snapshot taken at T0 plus
39/// the AOF of writes in (T0, now] reconstructs the current state. `SAVE`
40/// writes the snapshot then [`Aof::truncate`]s the log, so replay never
41/// double-applies.
42///
43/// Sizes (`size_bytes`, `size_at_last_rewrite`) drive auto-trigger of
44/// [`Aof::rewrite_from`] (BGREWRITEAOF) via the
45/// `auto_aof_rewrite_percentage` + `auto_aof_rewrite_min_size` knobs in
46/// `kevy_config`.
47pub struct Aof {
48 file: BufWriter<File>,
49 path: PathBuf,
50 fsync: Fsync,
51 dirty: bool,
52 last_sync: Instant,
53 /// Estimated bytes currently in the AOF file (existing + appended since
54 /// open). Maintained without fstat() syscalls per append.
55 size_bytes: u64,
56 /// File size right after the most recent [`Self::rewrite_from`] (or
57 /// `Self::open` if never rewritten). Anchor for `auto_aof_rewrite_*`.
58 size_at_last_rewrite: u64,
59 /// Total rewrites successfully completed since open. Surfaced via INFO.
60 rewrites_total: u64,
61 /// Group-commit window: while `true`, an `Fsync::Always` `append` only
62 /// buffers (sets `dirty`) instead of fsyncing per command. The caller
63 /// brackets a batch of writes with [`Self::begin_group`] /
64 /// [`Self::end_group`] and `end_group` does the single fsync **before**
65 /// the batch's replies are sent — preserving "durable before reply"
66 /// while amortizing the per-command `flush()+sync_data()` syscalls.
67 /// Only the multi-command reactor entry points (pipelined socket reads,
68 /// cross-shard request batches) open a group; every other path keeps
69 /// the per-command fsync, so the default is always the safe one.
70 deferred: bool,
71}
72
73/// Result of an [`Aof::rewrite_from`] call. Surfaced by `BGREWRITEAOF` /
74/// `INFO persistence`.
75#[derive(Debug, Clone, Copy)]
76pub struct RewriteStats {
77 /// Keys dumped into the new AOF.
78 pub keys: u64,
79 /// New AOF size in bytes.
80 pub bytes: u64,
81}
82
83impl Aof {
84 /// Open (creating if needed) `path` for appending. New files get the
85 /// 9-byte [`AOF_MAGIC`] header so replays can identify the file as
86 /// kevy-managed. Pre-existing files (legacy bare-RESP or already-
87 /// magic'd) are left untouched.
88 pub fn open(path: &Path, fsync: Fsync) -> io::Result<Self> {
89 let mut file = OpenOptions::new().create(true).append(true).open(path)?;
90 let mut size = file.metadata().map(|m| m.len()).unwrap_or(0);
91 if size == 0 {
92 // Fresh file: stamp the magic header so the replayer can
93 // distinguish kevy-written AOFs from accidental writes.
94 file.write_all(AOF_MAGIC)?;
95 file.sync_data()?;
96 size = AOF_MAGIC.len() as u64;
97 }
98 Ok(Aof {
99 file: BufWriter::new(file),
100 path: path.to_path_buf(),
101 fsync,
102 dirty: false,
103 last_sync: Instant::now(),
104 size_bytes: size,
105 size_at_last_rewrite: size,
106 rewrites_total: 0,
107 deferred: false,
108 })
109 }
110
111 /// The fsync policy this AOF was opened with (or last switched to).
112 /// Mostly for tests / INFO output; the hot path doesn't read this.
113 #[inline]
114 pub fn fsync_policy(&self) -> Fsync {
115 self.fsync
116 }
117
118 /// Switch the fsync policy at runtime (called by `CONFIG SET
119 /// appendfsync`). When tightening to `Always`, also flushes + fsyncs
120 /// any bytes still in the BufWriter so the new "every write is on
121 /// disk before reply" contract is honoured starting on the next
122 /// append, not after the dirty backlog clears.
123 pub fn set_fsync(&mut self, fsync: Fsync) -> io::Result<()> {
124 let upgrading_to_always = matches!(fsync, Fsync::Always) && !matches!(self.fsync, Fsync::Always);
125 self.fsync = fsync;
126 if upgrading_to_always && self.dirty {
127 self.file.flush()?;
128 self.file.get_ref().sync_data()?;
129 self.dirty = false;
130 self.last_sync = Instant::now();
131 }
132 Ok(())
133 }
134
135 /// Append one command, applying the fsync policy.
136 pub fn append<A: ArgvView + ?Sized>(&mut self, args: &A) -> io::Result<()> {
137 write_multibulk(&mut self.file, args)?;
138 self.size_bytes = self
139 .size_bytes
140 .saturating_add(estimate_multibulk_bytes(args));
141 match self.fsync {
142 // Inside a group-commit window, defer the fsync to `end_group`
143 // (one per batch, still before the batch's replies). Outside
144 // one, fsync per command — the safe default for every path.
145 Fsync::Always if self.deferred => self.dirty = true,
146 Fsync::Always => {
147 self.file.flush()?;
148 self.file.get_ref().sync_data()?;
149 }
150 Fsync::EverySec | Fsync::No => self.dirty = true,
151 }
152 Ok(())
153 }
154
155 /// Open a group-commit window (no-op unless the policy is `Always`):
156 /// subsequent `append`s buffer instead of fsyncing per command. Pair
157 /// with [`Self::end_group`] **before** sending the batch's replies.
158 #[inline]
159 pub fn begin_group(&mut self) {
160 if matches!(self.fsync, Fsync::Always) {
161 self.deferred = true;
162 }
163 }
164
165 /// Close the group-commit window: one `flush()+sync_data()` for the
166 /// whole batch (if anything was buffered), then resume per-command
167 /// fsync. Must be called before the batch's replies leave the shard.
168 #[inline]
169 pub fn end_group(&mut self) -> io::Result<()> {
170 if self.deferred {
171 self.deferred = false;
172 if self.dirty {
173 self.file.flush()?;
174 self.file.get_ref().sync_data()?;
175 self.dirty = false;
176 self.last_sync = Instant::now();
177 }
178 }
179 Ok(())
180 }
181
182 /// Flush+fsync if the `EverySec` window has elapsed. Call once per loop tick.
183 pub fn maybe_sync(&mut self) -> io::Result<()> {
184 if matches!(self.fsync, Fsync::EverySec)
185 && self.dirty
186 && self.last_sync.elapsed() >= Duration::from_secs(1)
187 {
188 self.file.flush()?;
189 self.file.get_ref().sync_data()?;
190 self.dirty = false;
191 self.last_sync = Instant::now();
192 }
193 Ok(())
194 }
195
196 /// Empty the log (after a snapshot has captured the full state). The
197 /// post-truncate file keeps the [`AOF_MAGIC`] header so replays of
198 /// the freshly-trimmed log still identify as kevy-managed.
199 pub fn truncate(&mut self) -> io::Result<()> {
200 self.file.flush()?;
201 let f = self.file.get_mut();
202 f.set_len(0)?;
203 f.seek(SeekFrom::Start(0))?; // harmless under O_APPEND; keeps len/pos coherent
204 f.write_all(AOF_MAGIC)?;
205 f.sync_all()?;
206 self.dirty = false;
207 self.size_bytes = AOF_MAGIC.len() as u64;
208 self.size_at_last_rewrite = AOF_MAGIC.len() as u64;
209 Ok(())
210 }
211
212 /// Estimated current AOF size in bytes (file content as of last append).
213 #[inline]
214 pub fn size_bytes(&self) -> u64 {
215 self.size_bytes
216 }
217
218 /// AOF size at the most recent rewrite (or open). Auto-trigger compares
219 /// `(size_bytes - size_at_last_rewrite) * 100 / size_at_last_rewrite` to
220 /// the `auto_aof_rewrite_percentage` knob.
221 #[inline]
222 pub fn size_at_last_rewrite(&self) -> u64 {
223 self.size_at_last_rewrite
224 }
225
226 /// Successful rewrite count since `Self::open`. Surfaced in INFO.
227 #[inline]
228 pub fn rewrites_total(&self) -> u64 {
229 self.rewrites_total
230 }
231
232 /// BGREWRITEAOF: rebuild a compact AOF from `store`'s current state and
233 /// atomically swap it in.
234 ///
235 /// **v1.0 is synchronous** — the calling shard blocks for the rewrite's
236 /// duration. Each shard owns its own AOF, so the shards' rewrites
237 /// proceed independently; per-shard blocking matches Redis's `BGSAVE`
238 /// cost in a typical single-key-per-shard workload. Concurrent
239 /// (rewrite-during-writes) incrementalisation is a v1.x perf item.
240 ///
241 /// Writes to a `<path>.rewrite` temp file with fsync, then `rename(2)`s
242 /// it over the live AOF. The append handle is reopened against the new
243 /// file before this call returns, so subsequent `append` calls land in
244 /// the rewritten log.
245 pub fn rewrite_from(&mut self, store: &Store) -> io::Result<RewriteStats> {
246 // Flush any pending writes to the OLD file first so the snapshot
247 // accounts for everything the caller intended to durabilise.
248 self.file.flush()?;
249
250 let tmp = rewrite_tmp_path(&self.path);
251 let (keys, bytes) = dump_store_to_aof(&tmp, store)?;
252
253 // Atomic replacement. After this, the OLD file descriptor in
254 // `self.file` is open against an unlinked inode; new writes would
255 // go nowhere visible. Reopen against the new path.
256 std::fs::rename(&tmp, &self.path)?;
257 let f = OpenOptions::new().append(true).open(&self.path)?;
258 self.file = BufWriter::new(f);
259 self.size_bytes = bytes;
260 self.size_at_last_rewrite = bytes;
261 self.dirty = false;
262 self.rewrites_total = self.rewrites_total.saturating_add(1);
263 Ok(RewriteStats { keys, bytes })
264 }
265}
266
267/// `<aof>.rewrite` — same-directory temp path so `rename(2)` stays atomic.
268fn rewrite_tmp_path(path: &Path) -> PathBuf {
269 let mut p = path.to_path_buf();
270 let new_name = match path.file_name() {
271 Some(n) => {
272 let mut s = n.to_os_string();
273 s.push(".rewrite");
274 s
275 }
276 None => std::ffi::OsString::from("aof.rewrite"),
277 };
278 p.set_file_name(new_name);
279 p
280}