Line data Source code
1 : pub mod detach_ancestor;
2 : pub mod partitioning;
3 : pub mod utilization;
4 :
5 : use core::ops::Range;
6 : use std::collections::HashMap;
7 : use std::fmt::Display;
8 : use std::io::{BufRead, Read};
9 : use std::num::{NonZeroU32, NonZeroU64, NonZeroUsize};
10 : use std::str::FromStr;
11 : use std::time::{Duration, SystemTime};
12 :
13 : use byteorder::{BigEndian, ReadBytesExt};
14 : use bytes::{Buf, BufMut, Bytes, BytesMut};
15 : #[cfg(feature = "testing")]
16 : use camino::Utf8PathBuf;
17 : use postgres_ffi::BLCKSZ;
18 : use serde::{Deserialize, Deserializer, Serialize, Serializer};
19 : use serde_with::serde_as;
20 : pub use utilization::PageserverUtilization;
21 : use utils::id::{NodeId, TenantId, TimelineId};
22 : use utils::lsn::Lsn;
23 : use utils::postgres_client::PostgresClientProtocol;
24 : use utils::{completion, serde_system_time};
25 :
26 : use crate::config::Ratio;
27 : use crate::key::{CompactKey, Key};
28 : use crate::reltag::RelTag;
29 : use crate::shard::{DEFAULT_STRIPE_SIZE, ShardCount, ShardStripeSize, TenantShardId};
30 :
31 : /// The state of a tenant in this pageserver.
32 : ///
33 : /// ```mermaid
34 : /// stateDiagram-v2
35 : ///
36 : /// [*] --> Attaching: spawn_attach()
37 : ///
38 : /// Attaching --> Activating: activate()
39 : /// Activating --> Active: infallible
40 : ///
41 : /// Attaching --> Broken: attach() failure
42 : ///
43 : /// Active --> Stopping: set_stopping(), part of shutdown & detach
44 : /// Stopping --> Broken: late error in remove_tenant_from_memory
45 : ///
46 : /// Broken --> [*]: ignore / detach / shutdown
47 : /// Stopping --> [*]: remove_from_memory complete
48 : ///
49 : /// Active --> Broken: cfg(testing)-only tenant break point
50 : /// ```
51 : #[derive(
52 : Clone,
53 : PartialEq,
54 : Eq,
55 0 : serde::Serialize,
56 1 : serde::Deserialize,
57 : strum_macros::Display,
58 : strum_macros::VariantNames,
59 : strum_macros::AsRefStr,
60 : strum_macros::IntoStaticStr,
61 : )]
62 : #[serde(tag = "slug", content = "data")]
63 : pub enum TenantState {
64 : /// This tenant is being attached to the pageserver.
65 : ///
66 : /// `set_stopping()` and `set_broken()` do not work in this state and wait for it to pass.
67 : Attaching,
68 : /// The tenant is transitioning from Loading/Attaching to Active.
69 : ///
70 : /// While in this state, the individual timelines are being activated.
71 : ///
72 : /// `set_stopping()` and `set_broken()` do not work in this state and wait for it to pass.
73 : Activating(ActivatingFrom),
74 : /// The tenant has finished activating and is open for business.
75 : ///
76 : /// Transitions out of this state are possible through `set_stopping()` and `set_broken()`.
77 : Active,
78 : /// The tenant is recognized by pageserver, but it is being detached or the
79 : /// system is being shut down.
80 : ///
81 : /// Transitions out of this state are possible through `set_broken()`.
82 : Stopping {
83 : /// The barrier can be used to wait for shutdown to complete. The first caller to set
84 : /// Some(Barrier) is responsible for driving shutdown to completion. Subsequent callers
85 : /// will wait for the first caller's existing barrier.
86 : ///
87 : /// None is set when an attach is cancelled, to signal to shutdown that the attach has in
88 : /// fact cancelled:
89 : ///
90 : /// 1. `shutdown` sees `TenantState::Attaching`, and cancels the tenant.
91 : /// 2. `attach` sets `TenantState::Stopping(None)` and exits.
92 : /// 3. `set_stopping` waits for `TenantState::Stopping(None)` and sets
93 : /// `TenantState::Stopping(Some)` to claim the barrier as the shutdown owner.
94 : //
95 : // Because of https://github.com/serde-rs/serde/issues/2105 this has to be a named field,
96 : // otherwise it will not be skipped during deserialization
97 : #[serde(skip)]
98 : progress: Option<completion::Barrier>,
99 : },
100 : /// The tenant is recognized by the pageserver, but can no longer be used for
101 : /// any operations.
102 : ///
103 : /// If the tenant fails to load or attach, it will transition to this state
104 : /// and it is guaranteed that no background tasks are running in its name.
105 : ///
106 : /// The other way to transition into this state is from `Stopping` state
107 : /// through `set_broken()` called from `remove_tenant_from_memory()`. That happens
108 : /// if the cleanup future executed by `remove_tenant_from_memory()` fails.
109 : Broken { reason: String, backtrace: String },
110 : }
111 :
112 : impl TenantState {
113 0 : pub fn attachment_status(&self) -> TenantAttachmentStatus {
114 : use TenantAttachmentStatus::*;
115 :
116 : // Below TenantState::Activating is used as "transient" or "transparent" state for
117 : // attachment_status determining.
118 0 : match self {
119 : // The attach procedure writes the marker file before adding the Attaching tenant to the tenants map.
120 : // So, technically, we can return Attached here.
121 : // However, as soon as Console observes Attached, it will proceed with the Postgres-level health check.
122 : // But, our attach task might still be fetching the remote timelines, etc.
123 : // So, return `Maybe` while Attaching, making Console wait for the attach task to finish.
124 0 : Self::Attaching | Self::Activating(ActivatingFrom::Attaching) => Maybe,
125 : // We only reach Active after successful load / attach.
126 : // So, call atttachment status Attached.
127 0 : Self::Active => Attached,
128 : // If the (initial or resumed) attach procedure fails, the tenant becomes Broken.
129 : // However, it also becomes Broken if the regular load fails.
130 : // From Console's perspective there's no practical difference
131 : // because attachment_status is polled by console only during attach operation execution.
132 0 : Self::Broken { reason, .. } => Failed {
133 0 : reason: reason.to_owned(),
134 0 : },
135 : // Why is Stopping a Maybe case? Because, during pageserver shutdown,
136 : // we set the Stopping state irrespective of whether the tenant
137 : // has finished attaching or not.
138 0 : Self::Stopping { .. } => Maybe,
139 : }
140 0 : }
141 :
142 0 : pub fn broken_from_reason(reason: String) -> Self {
143 0 : let backtrace_str: String = format!("{}", std::backtrace::Backtrace::force_capture());
144 0 : Self::Broken {
145 0 : reason,
146 0 : backtrace: backtrace_str,
147 0 : }
148 0 : }
149 : }
150 :
151 : impl std::fmt::Debug for TenantState {
152 2 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
153 2 : match self {
154 2 : Self::Broken { reason, backtrace } if !reason.is_empty() => {
155 2 : write!(f, "Broken due to: {reason}. Backtrace:\n{backtrace}")
156 : }
157 0 : _ => write!(f, "{self}"),
158 : }
159 2 : }
160 : }
161 :
162 : /// A temporary lease to a specific lsn inside a timeline.
163 : /// Access to the lsn is guaranteed by the pageserver until the expiration indicated by `valid_until`.
164 : #[serde_as]
165 0 : #[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
166 : pub struct LsnLease {
167 : #[serde_as(as = "SystemTimeAsRfc3339Millis")]
168 : pub valid_until: SystemTime,
169 : }
170 :
171 : serde_with::serde_conv!(
172 : SystemTimeAsRfc3339Millis,
173 : SystemTime,
174 0 : |time: &SystemTime| humantime::format_rfc3339_millis(*time).to_string(),
175 0 : |value: String| -> Result<_, humantime::TimestampError> { humantime::parse_rfc3339(&value) }
176 : );
177 :
178 : impl LsnLease {
179 : /// The default length for an explicit LSN lease request (10 minutes).
180 : pub const DEFAULT_LENGTH: Duration = Duration::from_secs(10 * 60);
181 :
182 : /// The default length for an implicit LSN lease granted during
183 : /// `get_lsn_by_timestamp` request (1 minutes).
184 : pub const DEFAULT_LENGTH_FOR_TS: Duration = Duration::from_secs(60);
185 :
186 : /// Checks whether the lease is expired.
187 3 : pub fn is_expired(&self, now: &SystemTime) -> bool {
188 3 : now > &self.valid_until
189 3 : }
190 : }
191 :
192 : /// Controls the detach ancestor behavior.
193 : /// - When set to `NoAncestorAndReparent`, we will only detach a branch if its ancestor is a root branch. It will automatically reparent any children of the ancestor before and at the branch point.
194 : /// - When set to `MultiLevelAndNoReparent`, we will detach a branch from multiple levels of ancestors, and no reparenting will happen at all.
195 : #[derive(Debug, Clone, Copy, Default)]
196 : pub enum DetachBehavior {
197 : #[default]
198 : NoAncestorAndReparent,
199 : MultiLevelAndNoReparent,
200 : }
201 :
202 : impl std::str::FromStr for DetachBehavior {
203 : type Err = &'static str;
204 :
205 0 : fn from_str(s: &str) -> Result<Self, Self::Err> {
206 0 : match s {
207 0 : "no_ancestor_and_reparent" => Ok(DetachBehavior::NoAncestorAndReparent),
208 0 : "multi_level_and_no_reparent" => Ok(DetachBehavior::MultiLevelAndNoReparent),
209 0 : "v1" => Ok(DetachBehavior::NoAncestorAndReparent),
210 0 : "v2" => Ok(DetachBehavior::MultiLevelAndNoReparent),
211 0 : _ => Err("cannot parse detach behavior"),
212 : }
213 0 : }
214 : }
215 :
216 : impl std::fmt::Display for DetachBehavior {
217 0 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
218 0 : match self {
219 0 : DetachBehavior::NoAncestorAndReparent => write!(f, "no_ancestor_and_reparent"),
220 0 : DetachBehavior::MultiLevelAndNoReparent => write!(f, "multi_level_and_no_reparent"),
221 : }
222 0 : }
223 : }
224 :
225 : /// The only [`TenantState`] variants we could be `TenantState::Activating` from.
226 : ///
227 : /// XXX: We used to have more variants here, but now it's just one, which makes this rather
228 : /// useless. Remove, once we've checked that there's no client code left that looks at this.
229 1 : #[derive(Clone, Copy, Debug, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
230 : pub enum ActivatingFrom {
231 : /// Arrived to [`TenantState::Activating`] from [`TenantState::Attaching`]
232 : Attaching,
233 : }
234 :
235 : /// A state of a timeline in pageserver's memory.
236 0 : #[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
237 : pub enum TimelineState {
238 : /// The timeline is recognized by the pageserver but is not yet operational.
239 : /// In particular, the walreceiver connection loop is not running for this timeline.
240 : /// It will eventually transition to state Active or Broken.
241 : Loading,
242 : /// The timeline is fully operational.
243 : /// It can be queried, and the walreceiver connection loop is running.
244 : Active,
245 : /// The timeline was previously Loading or Active but is shutting down.
246 : /// It cannot transition back into any other state.
247 : Stopping,
248 : /// The timeline is broken and not operational (previous states: Loading or Active).
249 : Broken { reason: String, backtrace: String },
250 : }
251 :
252 : #[serde_with::serde_as]
253 0 : #[derive(Debug, Clone, serde::Deserialize, serde::Serialize)]
254 : pub struct CompactLsnRange {
255 : pub start: Lsn,
256 : pub end: Lsn,
257 : }
258 :
259 : #[serde_with::serde_as]
260 0 : #[derive(Debug, Clone, serde::Deserialize, serde::Serialize)]
261 : pub struct CompactKeyRange {
262 : #[serde_as(as = "serde_with::DisplayFromStr")]
263 : pub start: Key,
264 : #[serde_as(as = "serde_with::DisplayFromStr")]
265 : pub end: Key,
266 : }
267 :
268 : impl From<Range<Lsn>> for CompactLsnRange {
269 3 : fn from(range: Range<Lsn>) -> Self {
270 3 : Self {
271 3 : start: range.start,
272 3 : end: range.end,
273 3 : }
274 3 : }
275 : }
276 :
277 : impl From<Range<Key>> for CompactKeyRange {
278 8 : fn from(range: Range<Key>) -> Self {
279 8 : Self {
280 8 : start: range.start,
281 8 : end: range.end,
282 8 : }
283 8 : }
284 : }
285 :
286 : impl From<CompactLsnRange> for Range<Lsn> {
287 5 : fn from(range: CompactLsnRange) -> Self {
288 5 : range.start..range.end
289 5 : }
290 : }
291 :
292 : impl From<CompactKeyRange> for Range<Key> {
293 8 : fn from(range: CompactKeyRange) -> Self {
294 8 : range.start..range.end
295 8 : }
296 : }
297 :
298 : impl CompactLsnRange {
299 2 : pub fn above(lsn: Lsn) -> Self {
300 2 : Self {
301 2 : start: lsn,
302 2 : end: Lsn::MAX,
303 2 : }
304 2 : }
305 : }
306 :
307 : #[derive(Debug, Clone, Serialize)]
308 : pub struct CompactInfoResponse {
309 : pub compact_key_range: Option<CompactKeyRange>,
310 : pub compact_lsn_range: Option<CompactLsnRange>,
311 : pub sub_compaction: bool,
312 : pub running: bool,
313 : pub job_id: usize,
314 : }
315 :
316 0 : #[derive(Serialize, Deserialize, Clone)]
317 : pub struct TimelineCreateRequest {
318 : pub new_timeline_id: TimelineId,
319 : #[serde(flatten)]
320 : pub mode: TimelineCreateRequestMode,
321 : }
322 :
323 : impl TimelineCreateRequest {
324 0 : pub fn mode_tag(&self) -> &'static str {
325 0 : match &self.mode {
326 0 : TimelineCreateRequestMode::Branch { .. } => "branch",
327 0 : TimelineCreateRequestMode::ImportPgdata { .. } => "import",
328 0 : TimelineCreateRequestMode::Bootstrap { .. } => "bootstrap",
329 : }
330 0 : }
331 :
332 0 : pub fn is_import(&self) -> bool {
333 0 : matches!(self.mode, TimelineCreateRequestMode::ImportPgdata { .. })
334 0 : }
335 : }
336 :
337 0 : #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq)]
338 : pub enum ShardImportStatus {
339 : InProgress(Option<ShardImportProgress>),
340 : Done,
341 : Error(String),
342 : }
343 :
344 0 : #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq)]
345 : pub enum ShardImportProgress {
346 : V1(ShardImportProgressV1),
347 : }
348 :
349 0 : #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq)]
350 : pub struct ShardImportProgressV1 {
351 : /// Total number of jobs in the import plan
352 : pub jobs: usize,
353 : /// Number of jobs completed
354 : pub completed: usize,
355 : /// Hash of the plan
356 : pub import_plan_hash: u64,
357 : }
358 :
359 : impl ShardImportStatus {
360 0 : pub fn is_terminal(&self) -> bool {
361 0 : match self {
362 0 : ShardImportStatus::InProgress(_) => false,
363 0 : ShardImportStatus::Done | ShardImportStatus::Error(_) => true,
364 : }
365 0 : }
366 : }
367 :
368 : /// Storage controller specific extensions to [`TimelineInfo`].
369 0 : #[derive(Serialize, Deserialize, Clone)]
370 : pub struct TimelineCreateResponseStorcon {
371 : #[serde(flatten)]
372 : pub timeline_info: TimelineInfo,
373 :
374 : pub safekeepers: Option<SafekeepersInfo>,
375 : }
376 :
377 : /// Safekeepers as returned in timeline creation request to storcon or pushed to
378 : /// cplane in the post migration hook.
379 0 : #[derive(Serialize, Deserialize, Clone)]
380 : pub struct SafekeepersInfo {
381 : pub tenant_id: TenantId,
382 : pub timeline_id: TimelineId,
383 : pub generation: u32,
384 : pub safekeepers: Vec<SafekeeperInfo>,
385 : }
386 :
387 0 : #[derive(Serialize, Deserialize, Clone)]
388 : pub struct SafekeeperInfo {
389 : pub id: NodeId,
390 : pub hostname: String,
391 : }
392 :
393 0 : #[derive(Serialize, Deserialize, Clone)]
394 : #[serde(untagged)]
395 : pub enum TimelineCreateRequestMode {
396 : Branch {
397 : ancestor_timeline_id: TimelineId,
398 : #[serde(default)]
399 : ancestor_start_lsn: Option<Lsn>,
400 : // TODO: cplane sets this, but, the branching code always
401 : // inherits the ancestor's pg_version. Earlier code wasn't
402 : // using a flattened enum, so, it was an accepted field, and
403 : // we continue to accept it by having it here.
404 : pg_version: Option<u32>,
405 : },
406 : ImportPgdata {
407 : import_pgdata: TimelineCreateRequestModeImportPgdata,
408 : },
409 : // NB: Bootstrap is all-optional, and thus the serde(untagged) will cause serde to stop at Bootstrap.
410 : // (serde picks the first matching enum variant, in declaration order).
411 : Bootstrap {
412 : #[serde(default)]
413 : existing_initdb_timeline_id: Option<TimelineId>,
414 : pg_version: Option<u32>,
415 : },
416 : }
417 :
418 0 : #[derive(Serialize, Deserialize, Clone)]
419 : pub struct TimelineCreateRequestModeImportPgdata {
420 : pub location: ImportPgdataLocation,
421 : pub idempotency_key: ImportPgdataIdempotencyKey,
422 : }
423 :
424 0 : #[derive(Serialize, Deserialize, Clone, Debug)]
425 : pub enum ImportPgdataLocation {
426 : #[cfg(feature = "testing")]
427 : LocalFs { path: Utf8PathBuf },
428 : AwsS3 {
429 : region: String,
430 : bucket: String,
431 : /// A better name for this would be `prefix`; changing requires coordination with cplane.
432 : /// See <https://github.com/neondatabase/cloud/issues/20646>.
433 : key: String,
434 : },
435 : }
436 :
437 0 : #[derive(Serialize, Deserialize, Clone)]
438 : #[serde(transparent)]
439 : pub struct ImportPgdataIdempotencyKey(pub String);
440 :
441 : impl ImportPgdataIdempotencyKey {
442 0 : pub fn random() -> Self {
443 : use rand::Rng;
444 : use rand::distributions::Alphanumeric;
445 0 : Self(
446 0 : rand::thread_rng()
447 0 : .sample_iter(&Alphanumeric)
448 0 : .take(20)
449 0 : .map(char::from)
450 0 : .collect(),
451 0 : )
452 0 : }
453 : }
454 :
455 0 : #[derive(Serialize, Deserialize, Clone)]
456 : pub struct LsnLeaseRequest {
457 : pub lsn: Lsn,
458 : }
459 :
460 0 : #[derive(Serialize, Deserialize)]
461 : pub struct TenantShardSplitRequest {
462 : pub new_shard_count: u8,
463 :
464 : // A tenant's stripe size is only meaningful the first time their shard count goes
465 : // above 1: therefore during a split from 1->N shards, we may modify the stripe size.
466 : //
467 : // If this is set while the stripe count is being increased from an already >1 value,
468 : // then the request will fail with 400.
469 : pub new_stripe_size: Option<ShardStripeSize>,
470 : }
471 :
472 0 : #[derive(Serialize, Deserialize)]
473 : pub struct TenantShardSplitResponse {
474 : pub new_shards: Vec<TenantShardId>,
475 : }
476 :
477 : /// Parameters that apply to all shards in a tenant. Used during tenant creation.
478 0 : #[derive(Serialize, Deserialize, Debug)]
479 : #[serde(deny_unknown_fields)]
480 : pub struct ShardParameters {
481 : pub count: ShardCount,
482 : pub stripe_size: ShardStripeSize,
483 : }
484 :
485 : impl ShardParameters {
486 0 : pub fn is_unsharded(&self) -> bool {
487 0 : self.count.is_unsharded()
488 0 : }
489 : }
490 :
491 : impl Default for ShardParameters {
492 118 : fn default() -> Self {
493 118 : Self {
494 118 : count: ShardCount::new(0),
495 118 : stripe_size: DEFAULT_STRIPE_SIZE,
496 118 : }
497 118 : }
498 : }
499 :
500 : #[derive(Debug, Default, Clone, Eq, PartialEq)]
501 : pub enum FieldPatch<T> {
502 : Upsert(T),
503 : Remove,
504 : #[default]
505 : Noop,
506 : }
507 :
508 : impl<T> FieldPatch<T> {
509 78 : fn is_noop(&self) -> bool {
510 78 : matches!(self, FieldPatch::Noop)
511 78 : }
512 :
513 39 : pub fn apply(self, target: &mut Option<T>) {
514 39 : match self {
515 1 : Self::Upsert(v) => *target = Some(v),
516 1 : Self::Remove => *target = None,
517 37 : Self::Noop => {}
518 : }
519 39 : }
520 :
521 10 : pub fn map<U, E, F: FnOnce(T) -> Result<U, E>>(self, map: F) -> Result<FieldPatch<U>, E> {
522 10 : match self {
523 0 : Self::Upsert(v) => Ok(FieldPatch::<U>::Upsert(map(v)?)),
524 0 : Self::Remove => Ok(FieldPatch::<U>::Remove),
525 10 : Self::Noop => Ok(FieldPatch::<U>::Noop),
526 : }
527 10 : }
528 : }
529 :
530 : impl<'de, T: Deserialize<'de>> Deserialize<'de> for FieldPatch<T> {
531 2 : fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
532 2 : where
533 2 : D: Deserializer<'de>,
534 2 : {
535 2 : Option::deserialize(deserializer).map(|opt| match opt {
536 1 : None => FieldPatch::Remove,
537 1 : Some(val) => FieldPatch::Upsert(val),
538 2 : })
539 2 : }
540 : }
541 :
542 : impl<T: Serialize> Serialize for FieldPatch<T> {
543 2 : fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
544 2 : where
545 2 : S: Serializer,
546 2 : {
547 2 : match self {
548 1 : FieldPatch::Upsert(val) => serializer.serialize_some(val),
549 1 : FieldPatch::Remove => serializer.serialize_none(),
550 0 : FieldPatch::Noop => unreachable!(),
551 : }
552 2 : }
553 : }
554 :
555 2 : #[derive(Serialize, Deserialize, Debug, Default, Clone, Eq, PartialEq)]
556 : #[serde(default)]
557 : pub struct TenantConfigPatch {
558 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
559 : pub checkpoint_distance: FieldPatch<u64>,
560 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
561 : pub checkpoint_timeout: FieldPatch<String>,
562 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
563 : pub compaction_target_size: FieldPatch<u64>,
564 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
565 : pub compaction_period: FieldPatch<String>,
566 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
567 : pub compaction_threshold: FieldPatch<usize>,
568 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
569 : pub compaction_upper_limit: FieldPatch<usize>,
570 : // defer parsing compaction_algorithm, like eviction_policy
571 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
572 : pub compaction_algorithm: FieldPatch<CompactionAlgorithmSettings>,
573 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
574 : pub compaction_shard_ancestor: FieldPatch<bool>,
575 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
576 : pub compaction_l0_first: FieldPatch<bool>,
577 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
578 : pub compaction_l0_semaphore: FieldPatch<bool>,
579 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
580 : pub l0_flush_delay_threshold: FieldPatch<usize>,
581 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
582 : pub l0_flush_stall_threshold: FieldPatch<usize>,
583 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
584 : pub gc_horizon: FieldPatch<u64>,
585 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
586 : pub gc_period: FieldPatch<String>,
587 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
588 : pub image_creation_threshold: FieldPatch<usize>,
589 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
590 : pub pitr_interval: FieldPatch<String>,
591 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
592 : pub walreceiver_connect_timeout: FieldPatch<String>,
593 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
594 : pub lagging_wal_timeout: FieldPatch<String>,
595 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
596 : pub max_lsn_wal_lag: FieldPatch<NonZeroU64>,
597 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
598 : pub eviction_policy: FieldPatch<EvictionPolicy>,
599 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
600 : pub min_resident_size_override: FieldPatch<u64>,
601 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
602 : pub evictions_low_residence_duration_metric_threshold: FieldPatch<String>,
603 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
604 : pub heatmap_period: FieldPatch<String>,
605 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
606 : pub lazy_slru_download: FieldPatch<bool>,
607 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
608 : pub timeline_get_throttle: FieldPatch<ThrottleConfig>,
609 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
610 : pub image_layer_creation_check_threshold: FieldPatch<u8>,
611 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
612 : pub image_creation_preempt_threshold: FieldPatch<usize>,
613 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
614 : pub lsn_lease_length: FieldPatch<String>,
615 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
616 : pub lsn_lease_length_for_ts: FieldPatch<String>,
617 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
618 : pub timeline_offloading: FieldPatch<bool>,
619 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
620 : pub wal_receiver_protocol_override: FieldPatch<PostgresClientProtocol>,
621 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
622 : pub rel_size_v2_enabled: FieldPatch<bool>,
623 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
624 : pub gc_compaction_enabled: FieldPatch<bool>,
625 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
626 : pub gc_compaction_verification: FieldPatch<bool>,
627 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
628 : pub gc_compaction_initial_threshold_kb: FieldPatch<u64>,
629 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
630 : pub gc_compaction_ratio_percent: FieldPatch<u64>,
631 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
632 : pub sampling_ratio: FieldPatch<Option<Ratio>>,
633 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
634 : pub relsize_snapshot_cache_capacity: FieldPatch<usize>,
635 : #[serde(skip_serializing_if = "FieldPatch::is_noop")]
636 : pub basebackup_cache_enabled: FieldPatch<bool>,
637 : }
638 :
639 : /// Like [`crate::config::TenantConfigToml`], but preserves the information
640 : /// about which parameters are set and which are not.
641 : ///
642 : /// Used in many places, including durably stored ones.
643 2 : #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
644 : #[serde(default)] // this maps omitted fields in deserialization to None
645 : pub struct TenantConfig {
646 : #[serde(skip_serializing_if = "Option::is_none")]
647 : pub checkpoint_distance: Option<u64>,
648 :
649 : #[serde(skip_serializing_if = "Option::is_none")]
650 : #[serde(with = "humantime_serde")]
651 : pub checkpoint_timeout: Option<Duration>,
652 :
653 : #[serde(skip_serializing_if = "Option::is_none")]
654 : pub compaction_target_size: Option<u64>,
655 :
656 : #[serde(skip_serializing_if = "Option::is_none")]
657 : #[serde(with = "humantime_serde")]
658 : pub compaction_period: Option<Duration>,
659 :
660 : #[serde(skip_serializing_if = "Option::is_none")]
661 : pub compaction_threshold: Option<usize>,
662 :
663 : #[serde(skip_serializing_if = "Option::is_none")]
664 : pub compaction_upper_limit: Option<usize>,
665 :
666 : #[serde(skip_serializing_if = "Option::is_none")]
667 : pub compaction_algorithm: Option<CompactionAlgorithmSettings>,
668 :
669 : #[serde(skip_serializing_if = "Option::is_none")]
670 : pub compaction_shard_ancestor: Option<bool>,
671 :
672 : #[serde(skip_serializing_if = "Option::is_none")]
673 : pub compaction_l0_first: Option<bool>,
674 :
675 : #[serde(skip_serializing_if = "Option::is_none")]
676 : pub compaction_l0_semaphore: Option<bool>,
677 :
678 : #[serde(skip_serializing_if = "Option::is_none")]
679 : pub l0_flush_delay_threshold: Option<usize>,
680 :
681 : #[serde(skip_serializing_if = "Option::is_none")]
682 : pub l0_flush_stall_threshold: Option<usize>,
683 :
684 : #[serde(skip_serializing_if = "Option::is_none")]
685 : pub gc_horizon: Option<u64>,
686 :
687 : #[serde(skip_serializing_if = "Option::is_none")]
688 : #[serde(with = "humantime_serde")]
689 : pub gc_period: Option<Duration>,
690 :
691 : #[serde(skip_serializing_if = "Option::is_none")]
692 : pub image_creation_threshold: Option<usize>,
693 :
694 : #[serde(skip_serializing_if = "Option::is_none")]
695 : #[serde(with = "humantime_serde")]
696 : pub pitr_interval: Option<Duration>,
697 :
698 : #[serde(skip_serializing_if = "Option::is_none")]
699 : #[serde(with = "humantime_serde")]
700 : pub walreceiver_connect_timeout: Option<Duration>,
701 :
702 : #[serde(skip_serializing_if = "Option::is_none")]
703 : #[serde(with = "humantime_serde")]
704 : pub lagging_wal_timeout: Option<Duration>,
705 :
706 : #[serde(skip_serializing_if = "Option::is_none")]
707 : pub max_lsn_wal_lag: Option<NonZeroU64>,
708 :
709 : #[serde(skip_serializing_if = "Option::is_none")]
710 : pub eviction_policy: Option<EvictionPolicy>,
711 :
712 : #[serde(skip_serializing_if = "Option::is_none")]
713 : pub min_resident_size_override: Option<u64>,
714 :
715 : #[serde(skip_serializing_if = "Option::is_none")]
716 : #[serde(with = "humantime_serde")]
717 : pub evictions_low_residence_duration_metric_threshold: Option<Duration>,
718 :
719 : #[serde(skip_serializing_if = "Option::is_none")]
720 : #[serde(with = "humantime_serde")]
721 : pub heatmap_period: Option<Duration>,
722 :
723 : #[serde(skip_serializing_if = "Option::is_none")]
724 : pub lazy_slru_download: Option<bool>,
725 :
726 : #[serde(skip_serializing_if = "Option::is_none")]
727 : pub timeline_get_throttle: Option<ThrottleConfig>,
728 :
729 : #[serde(skip_serializing_if = "Option::is_none")]
730 : pub image_layer_creation_check_threshold: Option<u8>,
731 :
732 : #[serde(skip_serializing_if = "Option::is_none")]
733 : pub image_creation_preempt_threshold: Option<usize>,
734 :
735 : #[serde(skip_serializing_if = "Option::is_none")]
736 : #[serde(with = "humantime_serde")]
737 : pub lsn_lease_length: Option<Duration>,
738 :
739 : #[serde(skip_serializing_if = "Option::is_none")]
740 : #[serde(with = "humantime_serde")]
741 : pub lsn_lease_length_for_ts: Option<Duration>,
742 :
743 : #[serde(skip_serializing_if = "Option::is_none")]
744 : pub timeline_offloading: Option<bool>,
745 :
746 : #[serde(skip_serializing_if = "Option::is_none")]
747 : pub wal_receiver_protocol_override: Option<PostgresClientProtocol>,
748 :
749 : #[serde(skip_serializing_if = "Option::is_none")]
750 : pub rel_size_v2_enabled: Option<bool>,
751 :
752 : #[serde(skip_serializing_if = "Option::is_none")]
753 : pub gc_compaction_enabled: Option<bool>,
754 :
755 : #[serde(skip_serializing_if = "Option::is_none")]
756 : pub gc_compaction_verification: Option<bool>,
757 :
758 : #[serde(skip_serializing_if = "Option::is_none")]
759 : pub gc_compaction_initial_threshold_kb: Option<u64>,
760 :
761 : #[serde(skip_serializing_if = "Option::is_none")]
762 : pub gc_compaction_ratio_percent: Option<u64>,
763 :
764 : #[serde(skip_serializing_if = "Option::is_none")]
765 : pub sampling_ratio: Option<Option<Ratio>>,
766 :
767 : #[serde(skip_serializing_if = "Option::is_none")]
768 : pub relsize_snapshot_cache_capacity: Option<usize>,
769 :
770 : #[serde(skip_serializing_if = "Option::is_none")]
771 : pub basebackup_cache_enabled: Option<bool>,
772 : }
773 :
774 : impl TenantConfig {
775 1 : pub fn apply_patch(
776 1 : self,
777 1 : patch: TenantConfigPatch,
778 1 : ) -> Result<TenantConfig, humantime::DurationError> {
779 1 : let Self {
780 1 : mut checkpoint_distance,
781 1 : mut checkpoint_timeout,
782 1 : mut compaction_target_size,
783 1 : mut compaction_period,
784 1 : mut compaction_threshold,
785 1 : mut compaction_upper_limit,
786 1 : mut compaction_algorithm,
787 1 : mut compaction_shard_ancestor,
788 1 : mut compaction_l0_first,
789 1 : mut compaction_l0_semaphore,
790 1 : mut l0_flush_delay_threshold,
791 1 : mut l0_flush_stall_threshold,
792 1 : mut gc_horizon,
793 1 : mut gc_period,
794 1 : mut image_creation_threshold,
795 1 : mut pitr_interval,
796 1 : mut walreceiver_connect_timeout,
797 1 : mut lagging_wal_timeout,
798 1 : mut max_lsn_wal_lag,
799 1 : mut eviction_policy,
800 1 : mut min_resident_size_override,
801 1 : mut evictions_low_residence_duration_metric_threshold,
802 1 : mut heatmap_period,
803 1 : mut lazy_slru_download,
804 1 : mut timeline_get_throttle,
805 1 : mut image_layer_creation_check_threshold,
806 1 : mut image_creation_preempt_threshold,
807 1 : mut lsn_lease_length,
808 1 : mut lsn_lease_length_for_ts,
809 1 : mut timeline_offloading,
810 1 : mut wal_receiver_protocol_override,
811 1 : mut rel_size_v2_enabled,
812 1 : mut gc_compaction_enabled,
813 1 : mut gc_compaction_verification,
814 1 : mut gc_compaction_initial_threshold_kb,
815 1 : mut gc_compaction_ratio_percent,
816 1 : mut sampling_ratio,
817 1 : mut relsize_snapshot_cache_capacity,
818 1 : mut basebackup_cache_enabled,
819 1 : } = self;
820 1 :
821 1 : patch.checkpoint_distance.apply(&mut checkpoint_distance);
822 1 : patch
823 1 : .checkpoint_timeout
824 1 : .map(|v| humantime::parse_duration(&v))?
825 1 : .apply(&mut checkpoint_timeout);
826 1 : patch
827 1 : .compaction_target_size
828 1 : .apply(&mut compaction_target_size);
829 1 : patch
830 1 : .compaction_period
831 1 : .map(|v| humantime::parse_duration(&v))?
832 1 : .apply(&mut compaction_period);
833 1 : patch.compaction_threshold.apply(&mut compaction_threshold);
834 1 : patch
835 1 : .compaction_upper_limit
836 1 : .apply(&mut compaction_upper_limit);
837 1 : patch.compaction_algorithm.apply(&mut compaction_algorithm);
838 1 : patch
839 1 : .compaction_shard_ancestor
840 1 : .apply(&mut compaction_shard_ancestor);
841 1 : patch.compaction_l0_first.apply(&mut compaction_l0_first);
842 1 : patch
843 1 : .compaction_l0_semaphore
844 1 : .apply(&mut compaction_l0_semaphore);
845 1 : patch
846 1 : .l0_flush_delay_threshold
847 1 : .apply(&mut l0_flush_delay_threshold);
848 1 : patch
849 1 : .l0_flush_stall_threshold
850 1 : .apply(&mut l0_flush_stall_threshold);
851 1 : patch.gc_horizon.apply(&mut gc_horizon);
852 1 : patch
853 1 : .gc_period
854 1 : .map(|v| humantime::parse_duration(&v))?
855 1 : .apply(&mut gc_period);
856 1 : patch
857 1 : .image_creation_threshold
858 1 : .apply(&mut image_creation_threshold);
859 1 : patch
860 1 : .pitr_interval
861 1 : .map(|v| humantime::parse_duration(&v))?
862 1 : .apply(&mut pitr_interval);
863 1 : patch
864 1 : .walreceiver_connect_timeout
865 1 : .map(|v| humantime::parse_duration(&v))?
866 1 : .apply(&mut walreceiver_connect_timeout);
867 1 : patch
868 1 : .lagging_wal_timeout
869 1 : .map(|v| humantime::parse_duration(&v))?
870 1 : .apply(&mut lagging_wal_timeout);
871 1 : patch.max_lsn_wal_lag.apply(&mut max_lsn_wal_lag);
872 1 : patch.eviction_policy.apply(&mut eviction_policy);
873 1 : patch
874 1 : .min_resident_size_override
875 1 : .apply(&mut min_resident_size_override);
876 1 : patch
877 1 : .evictions_low_residence_duration_metric_threshold
878 1 : .map(|v| humantime::parse_duration(&v))?
879 1 : .apply(&mut evictions_low_residence_duration_metric_threshold);
880 1 : patch
881 1 : .heatmap_period
882 1 : .map(|v| humantime::parse_duration(&v))?
883 1 : .apply(&mut heatmap_period);
884 1 : patch.lazy_slru_download.apply(&mut lazy_slru_download);
885 1 : patch
886 1 : .timeline_get_throttle
887 1 : .apply(&mut timeline_get_throttle);
888 1 : patch
889 1 : .image_layer_creation_check_threshold
890 1 : .apply(&mut image_layer_creation_check_threshold);
891 1 : patch
892 1 : .image_creation_preempt_threshold
893 1 : .apply(&mut image_creation_preempt_threshold);
894 1 : patch
895 1 : .lsn_lease_length
896 1 : .map(|v| humantime::parse_duration(&v))?
897 1 : .apply(&mut lsn_lease_length);
898 1 : patch
899 1 : .lsn_lease_length_for_ts
900 1 : .map(|v| humantime::parse_duration(&v))?
901 1 : .apply(&mut lsn_lease_length_for_ts);
902 1 : patch.timeline_offloading.apply(&mut timeline_offloading);
903 1 : patch
904 1 : .wal_receiver_protocol_override
905 1 : .apply(&mut wal_receiver_protocol_override);
906 1 : patch.rel_size_v2_enabled.apply(&mut rel_size_v2_enabled);
907 1 : patch
908 1 : .gc_compaction_enabled
909 1 : .apply(&mut gc_compaction_enabled);
910 1 : patch
911 1 : .gc_compaction_verification
912 1 : .apply(&mut gc_compaction_verification);
913 1 : patch
914 1 : .gc_compaction_initial_threshold_kb
915 1 : .apply(&mut gc_compaction_initial_threshold_kb);
916 1 : patch
917 1 : .gc_compaction_ratio_percent
918 1 : .apply(&mut gc_compaction_ratio_percent);
919 1 : patch.sampling_ratio.apply(&mut sampling_ratio);
920 1 : patch
921 1 : .relsize_snapshot_cache_capacity
922 1 : .apply(&mut relsize_snapshot_cache_capacity);
923 1 : patch
924 1 : .basebackup_cache_enabled
925 1 : .apply(&mut basebackup_cache_enabled);
926 1 :
927 1 : Ok(Self {
928 1 : checkpoint_distance,
929 1 : checkpoint_timeout,
930 1 : compaction_target_size,
931 1 : compaction_period,
932 1 : compaction_threshold,
933 1 : compaction_upper_limit,
934 1 : compaction_algorithm,
935 1 : compaction_shard_ancestor,
936 1 : compaction_l0_first,
937 1 : compaction_l0_semaphore,
938 1 : l0_flush_delay_threshold,
939 1 : l0_flush_stall_threshold,
940 1 : gc_horizon,
941 1 : gc_period,
942 1 : image_creation_threshold,
943 1 : pitr_interval,
944 1 : walreceiver_connect_timeout,
945 1 : lagging_wal_timeout,
946 1 : max_lsn_wal_lag,
947 1 : eviction_policy,
948 1 : min_resident_size_override,
949 1 : evictions_low_residence_duration_metric_threshold,
950 1 : heatmap_period,
951 1 : lazy_slru_download,
952 1 : timeline_get_throttle,
953 1 : image_layer_creation_check_threshold,
954 1 : image_creation_preempt_threshold,
955 1 : lsn_lease_length,
956 1 : lsn_lease_length_for_ts,
957 1 : timeline_offloading,
958 1 : wal_receiver_protocol_override,
959 1 : rel_size_v2_enabled,
960 1 : gc_compaction_enabled,
961 1 : gc_compaction_verification,
962 1 : gc_compaction_initial_threshold_kb,
963 1 : gc_compaction_ratio_percent,
964 1 : sampling_ratio,
965 1 : relsize_snapshot_cache_capacity,
966 1 : basebackup_cache_enabled,
967 1 : })
968 1 : }
969 :
970 0 : pub fn merge(
971 0 : &self,
972 0 : global_conf: crate::config::TenantConfigToml,
973 0 : ) -> crate::config::TenantConfigToml {
974 0 : crate::config::TenantConfigToml {
975 0 : checkpoint_distance: self
976 0 : .checkpoint_distance
977 0 : .unwrap_or(global_conf.checkpoint_distance),
978 0 : checkpoint_timeout: self
979 0 : .checkpoint_timeout
980 0 : .unwrap_or(global_conf.checkpoint_timeout),
981 0 : compaction_target_size: self
982 0 : .compaction_target_size
983 0 : .unwrap_or(global_conf.compaction_target_size),
984 0 : compaction_period: self
985 0 : .compaction_period
986 0 : .unwrap_or(global_conf.compaction_period),
987 0 : compaction_threshold: self
988 0 : .compaction_threshold
989 0 : .unwrap_or(global_conf.compaction_threshold),
990 0 : compaction_upper_limit: self
991 0 : .compaction_upper_limit
992 0 : .unwrap_or(global_conf.compaction_upper_limit),
993 0 : compaction_algorithm: self
994 0 : .compaction_algorithm
995 0 : .as_ref()
996 0 : .unwrap_or(&global_conf.compaction_algorithm)
997 0 : .clone(),
998 0 : compaction_shard_ancestor: self
999 0 : .compaction_shard_ancestor
1000 0 : .unwrap_or(global_conf.compaction_shard_ancestor),
1001 0 : compaction_l0_first: self
1002 0 : .compaction_l0_first
1003 0 : .unwrap_or(global_conf.compaction_l0_first),
1004 0 : compaction_l0_semaphore: self
1005 0 : .compaction_l0_semaphore
1006 0 : .unwrap_or(global_conf.compaction_l0_semaphore),
1007 0 : l0_flush_delay_threshold: self
1008 0 : .l0_flush_delay_threshold
1009 0 : .or(global_conf.l0_flush_delay_threshold),
1010 0 : l0_flush_stall_threshold: self
1011 0 : .l0_flush_stall_threshold
1012 0 : .or(global_conf.l0_flush_stall_threshold),
1013 0 : gc_horizon: self.gc_horizon.unwrap_or(global_conf.gc_horizon),
1014 0 : gc_period: self.gc_period.unwrap_or(global_conf.gc_period),
1015 0 : image_creation_threshold: self
1016 0 : .image_creation_threshold
1017 0 : .unwrap_or(global_conf.image_creation_threshold),
1018 0 : pitr_interval: self.pitr_interval.unwrap_or(global_conf.pitr_interval),
1019 0 : walreceiver_connect_timeout: self
1020 0 : .walreceiver_connect_timeout
1021 0 : .unwrap_or(global_conf.walreceiver_connect_timeout),
1022 0 : lagging_wal_timeout: self
1023 0 : .lagging_wal_timeout
1024 0 : .unwrap_or(global_conf.lagging_wal_timeout),
1025 0 : max_lsn_wal_lag: self.max_lsn_wal_lag.unwrap_or(global_conf.max_lsn_wal_lag),
1026 0 : eviction_policy: self.eviction_policy.unwrap_or(global_conf.eviction_policy),
1027 0 : min_resident_size_override: self
1028 0 : .min_resident_size_override
1029 0 : .or(global_conf.min_resident_size_override),
1030 0 : evictions_low_residence_duration_metric_threshold: self
1031 0 : .evictions_low_residence_duration_metric_threshold
1032 0 : .unwrap_or(global_conf.evictions_low_residence_duration_metric_threshold),
1033 0 : heatmap_period: self.heatmap_period.unwrap_or(global_conf.heatmap_period),
1034 0 : lazy_slru_download: self
1035 0 : .lazy_slru_download
1036 0 : .unwrap_or(global_conf.lazy_slru_download),
1037 0 : timeline_get_throttle: self
1038 0 : .timeline_get_throttle
1039 0 : .clone()
1040 0 : .unwrap_or(global_conf.timeline_get_throttle),
1041 0 : image_layer_creation_check_threshold: self
1042 0 : .image_layer_creation_check_threshold
1043 0 : .unwrap_or(global_conf.image_layer_creation_check_threshold),
1044 0 : image_creation_preempt_threshold: self
1045 0 : .image_creation_preempt_threshold
1046 0 : .unwrap_or(global_conf.image_creation_preempt_threshold),
1047 0 : lsn_lease_length: self
1048 0 : .lsn_lease_length
1049 0 : .unwrap_or(global_conf.lsn_lease_length),
1050 0 : lsn_lease_length_for_ts: self
1051 0 : .lsn_lease_length_for_ts
1052 0 : .unwrap_or(global_conf.lsn_lease_length_for_ts),
1053 0 : timeline_offloading: self
1054 0 : .timeline_offloading
1055 0 : .unwrap_or(global_conf.timeline_offloading),
1056 0 : wal_receiver_protocol_override: self
1057 0 : .wal_receiver_protocol_override
1058 0 : .or(global_conf.wal_receiver_protocol_override),
1059 0 : rel_size_v2_enabled: self
1060 0 : .rel_size_v2_enabled
1061 0 : .unwrap_or(global_conf.rel_size_v2_enabled),
1062 0 : gc_compaction_enabled: self
1063 0 : .gc_compaction_enabled
1064 0 : .unwrap_or(global_conf.gc_compaction_enabled),
1065 0 : gc_compaction_verification: self
1066 0 : .gc_compaction_verification
1067 0 : .unwrap_or(global_conf.gc_compaction_verification),
1068 0 : gc_compaction_initial_threshold_kb: self
1069 0 : .gc_compaction_initial_threshold_kb
1070 0 : .unwrap_or(global_conf.gc_compaction_initial_threshold_kb),
1071 0 : gc_compaction_ratio_percent: self
1072 0 : .gc_compaction_ratio_percent
1073 0 : .unwrap_or(global_conf.gc_compaction_ratio_percent),
1074 0 : sampling_ratio: self.sampling_ratio.unwrap_or(global_conf.sampling_ratio),
1075 0 : relsize_snapshot_cache_capacity: self
1076 0 : .relsize_snapshot_cache_capacity
1077 0 : .unwrap_or(global_conf.relsize_snapshot_cache_capacity),
1078 0 : basebackup_cache_enabled: self
1079 0 : .basebackup_cache_enabled
1080 0 : .unwrap_or(global_conf.basebackup_cache_enabled),
1081 0 : }
1082 0 : }
1083 : }
1084 :
1085 : /// The policy for the aux file storage.
1086 : ///
1087 : /// It can be switched through `switch_aux_file_policy` tenant config.
1088 : /// When the first aux file written, the policy will be persisted in the
1089 : /// `index_part.json` file and has a limited migration path.
1090 : ///
1091 : /// Currently, we only allow the following migration path:
1092 : ///
1093 : /// Unset -> V1
1094 : /// -> V2
1095 : /// -> CrossValidation -> V2
1096 : #[derive(
1097 : Eq,
1098 : PartialEq,
1099 : Debug,
1100 : Copy,
1101 : Clone,
1102 0 : strum_macros::EnumString,
1103 : strum_macros::Display,
1104 1 : serde_with::DeserializeFromStr,
1105 : serde_with::SerializeDisplay,
1106 : )]
1107 : #[strum(serialize_all = "kebab-case")]
1108 : pub enum AuxFilePolicy {
1109 : /// V1 aux file policy: store everything in AUX_FILE_KEY
1110 : #[strum(ascii_case_insensitive)]
1111 : V1,
1112 : /// V2 aux file policy: store in the AUX_FILE keyspace
1113 : #[strum(ascii_case_insensitive)]
1114 : V2,
1115 : /// Cross validation runs both formats on the write path and does validation
1116 : /// on the read path.
1117 : #[strum(ascii_case_insensitive)]
1118 : CrossValidation,
1119 : }
1120 :
1121 0 : #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
1122 : #[serde(tag = "kind")]
1123 : pub enum EvictionPolicy {
1124 : NoEviction,
1125 : LayerAccessThreshold(EvictionPolicyLayerAccessThreshold),
1126 : OnlyImitiate(EvictionPolicyLayerAccessThreshold),
1127 : }
1128 :
1129 : impl EvictionPolicy {
1130 0 : pub fn discriminant_str(&self) -> &'static str {
1131 0 : match self {
1132 0 : EvictionPolicy::NoEviction => "NoEviction",
1133 0 : EvictionPolicy::LayerAccessThreshold(_) => "LayerAccessThreshold",
1134 0 : EvictionPolicy::OnlyImitiate(_) => "OnlyImitiate",
1135 : }
1136 0 : }
1137 : }
1138 :
1139 : #[derive(
1140 : Eq,
1141 : PartialEq,
1142 : Debug,
1143 : Copy,
1144 : Clone,
1145 0 : strum_macros::EnumString,
1146 : strum_macros::Display,
1147 0 : serde_with::DeserializeFromStr,
1148 : serde_with::SerializeDisplay,
1149 : )]
1150 : #[strum(serialize_all = "kebab-case")]
1151 : pub enum CompactionAlgorithm {
1152 : Legacy,
1153 : Tiered,
1154 : }
1155 :
1156 : #[derive(
1157 4 : Debug, Clone, Copy, PartialEq, Eq, serde_with::DeserializeFromStr, serde_with::SerializeDisplay,
1158 : )]
1159 : pub enum ImageCompressionAlgorithm {
1160 : // Disabled for writes, support decompressing during read path
1161 : Disabled,
1162 : /// Zstandard compression. Level 0 means and None mean the same (default level). Levels can be negative as well.
1163 : /// For details, see the [manual](http://facebook.github.io/zstd/zstd_manual.html).
1164 : Zstd {
1165 : level: Option<i8>,
1166 : },
1167 : }
1168 :
1169 : impl FromStr for ImageCompressionAlgorithm {
1170 : type Err = anyhow::Error;
1171 8 : fn from_str(s: &str) -> Result<Self, Self::Err> {
1172 8 : let mut components = s.split(['(', ')']);
1173 8 : let first = components
1174 8 : .next()
1175 8 : .ok_or_else(|| anyhow::anyhow!("empty string"))?;
1176 8 : match first {
1177 8 : "disabled" => Ok(ImageCompressionAlgorithm::Disabled),
1178 6 : "zstd" => {
1179 6 : let level = if let Some(v) = components.next() {
1180 4 : let v: i8 = v.parse()?;
1181 4 : Some(v)
1182 : } else {
1183 2 : None
1184 : };
1185 :
1186 6 : Ok(ImageCompressionAlgorithm::Zstd { level })
1187 : }
1188 0 : _ => anyhow::bail!("invalid specifier '{first}'"),
1189 : }
1190 8 : }
1191 : }
1192 :
1193 : impl Display for ImageCompressionAlgorithm {
1194 12 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
1195 12 : match self {
1196 3 : ImageCompressionAlgorithm::Disabled => write!(f, "disabled"),
1197 9 : ImageCompressionAlgorithm::Zstd { level } => {
1198 9 : if let Some(level) = level {
1199 6 : write!(f, "zstd({})", level)
1200 : } else {
1201 3 : write!(f, "zstd")
1202 : }
1203 : }
1204 : }
1205 12 : }
1206 : }
1207 :
1208 0 : #[derive(Eq, PartialEq, Debug, Clone, Serialize, Deserialize)]
1209 : pub struct CompactionAlgorithmSettings {
1210 : pub kind: CompactionAlgorithm,
1211 : }
1212 :
1213 0 : #[derive(Debug, PartialEq, Eq, Clone, Deserialize, Serialize)]
1214 : #[serde(tag = "mode", rename_all = "kebab-case")]
1215 : pub enum L0FlushConfig {
1216 : #[serde(rename_all = "snake_case")]
1217 : Direct { max_concurrency: NonZeroUsize },
1218 : }
1219 :
1220 0 : #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
1221 : pub struct EvictionPolicyLayerAccessThreshold {
1222 : #[serde(with = "humantime_serde")]
1223 : pub period: Duration,
1224 : #[serde(with = "humantime_serde")]
1225 : pub threshold: Duration,
1226 : }
1227 :
1228 6 : #[derive(Debug, Serialize, Deserialize, Clone, PartialEq, Eq)]
1229 : pub struct ThrottleConfig {
1230 : /// See [`ThrottleConfigTaskKinds`] for why we do the serde `rename`.
1231 : #[serde(rename = "task_kinds")]
1232 : pub enabled: ThrottleConfigTaskKinds,
1233 : pub initial: u32,
1234 : #[serde(with = "humantime_serde")]
1235 : pub refill_interval: Duration,
1236 : pub refill_amount: NonZeroU32,
1237 : pub max: u32,
1238 : }
1239 :
1240 : /// Before <https://github.com/neondatabase/neon/pull/9962>
1241 : /// the throttle was a per `Timeline::get`/`Timeline::get_vectored` call.
1242 : /// The `task_kinds` field controlled which Pageserver "Task Kind"s
1243 : /// were subject to the throttle.
1244 : ///
1245 : /// After that PR, the throttle is applied at pagestream request level
1246 : /// and the `task_kinds` field does not apply since the only task kind
1247 : /// that us subject to the throttle is that of the page service.
1248 : ///
1249 : /// However, we don't want to make a breaking config change right now
1250 : /// because it means we have to migrate all the tenant configs.
1251 : /// This will be done in a future PR.
1252 : ///
1253 : /// In the meantime, we use emptiness / non-emptsiness of the `task_kinds`
1254 : /// field to determine if the throttle is enabled or not.
1255 1 : #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq)]
1256 : #[serde(transparent)]
1257 : pub struct ThrottleConfigTaskKinds(Vec<String>);
1258 :
1259 : impl ThrottleConfigTaskKinds {
1260 128 : pub fn disabled() -> Self {
1261 128 : Self(vec![])
1262 128 : }
1263 119 : pub fn is_enabled(&self) -> bool {
1264 119 : !self.0.is_empty()
1265 119 : }
1266 : }
1267 :
1268 : impl ThrottleConfig {
1269 128 : pub fn disabled() -> Self {
1270 128 : Self {
1271 128 : enabled: ThrottleConfigTaskKinds::disabled(),
1272 128 : // other values don't matter with emtpy `task_kinds`.
1273 128 : initial: 0,
1274 128 : refill_interval: Duration::from_millis(1),
1275 128 : refill_amount: NonZeroU32::new(1).unwrap(),
1276 128 : max: 1,
1277 128 : }
1278 128 : }
1279 : /// The requests per second allowed by the given config.
1280 0 : pub fn steady_rps(&self) -> f64 {
1281 0 : (self.refill_amount.get() as f64) / (self.refill_interval.as_secs_f64())
1282 0 : }
1283 : }
1284 :
1285 : #[cfg(test)]
1286 : mod throttle_config_tests {
1287 : use super::*;
1288 :
1289 : #[test]
1290 1 : fn test_disabled_is_disabled() {
1291 1 : let config = ThrottleConfig::disabled();
1292 1 : assert!(!config.enabled.is_enabled());
1293 1 : }
1294 : #[test]
1295 1 : fn test_enabled_backwards_compat() {
1296 1 : let input = serde_json::json!({
1297 1 : "task_kinds": ["PageRequestHandler"],
1298 1 : "initial": 40000,
1299 1 : "refill_interval": "50ms",
1300 1 : "refill_amount": 1000,
1301 1 : "max": 40000,
1302 1 : "fair": true
1303 1 : });
1304 1 : let config: ThrottleConfig = serde_json::from_value(input).unwrap();
1305 1 : assert!(config.enabled.is_enabled());
1306 1 : }
1307 : }
1308 :
1309 : /// A flattened analog of a `pagesever::tenant::LocationMode`, which
1310 : /// lists out all possible states (and the virtual "Detached" state)
1311 : /// in a flat form rather than using rust-style enums.
1312 0 : #[derive(Serialize, Deserialize, Debug, Clone, Copy, Eq, PartialEq)]
1313 : pub enum LocationConfigMode {
1314 : AttachedSingle,
1315 : AttachedMulti,
1316 : AttachedStale,
1317 : Secondary,
1318 : Detached,
1319 : }
1320 :
1321 0 : #[derive(Serialize, Deserialize, Debug, Clone, Eq, PartialEq)]
1322 : pub struct LocationConfigSecondary {
1323 : pub warm: bool,
1324 : }
1325 :
1326 : /// An alternative representation of `pageserver::tenant::LocationConf`,
1327 : /// for use in external-facing APIs.
1328 0 : #[derive(Serialize, Deserialize, Debug, Clone, Eq, PartialEq)]
1329 : pub struct LocationConfig {
1330 : pub mode: LocationConfigMode,
1331 : /// If attaching, in what generation?
1332 : #[serde(default)]
1333 : pub generation: Option<u32>,
1334 :
1335 : // If requesting mode `Secondary`, configuration for that.
1336 : #[serde(default)]
1337 : pub secondary_conf: Option<LocationConfigSecondary>,
1338 :
1339 : // Shard parameters: if shard_count is nonzero, then other shard_* fields
1340 : // must be set accurately.
1341 : #[serde(default)]
1342 : pub shard_number: u8,
1343 : #[serde(default)]
1344 : pub shard_count: u8,
1345 : #[serde(default)]
1346 : pub shard_stripe_size: u32,
1347 :
1348 : // This configuration only affects attached mode, but should be provided irrespective
1349 : // of the mode, as a secondary location might transition on startup if the response
1350 : // to the `/re-attach` control plane API requests it.
1351 : pub tenant_conf: TenantConfig,
1352 : }
1353 :
1354 0 : #[derive(Serialize, Deserialize)]
1355 : pub struct LocationConfigListResponse {
1356 : pub tenant_shards: Vec<(TenantShardId, Option<LocationConfig>)>,
1357 : }
1358 :
1359 : #[derive(Serialize)]
1360 : pub struct StatusResponse {
1361 : pub id: NodeId,
1362 : }
1363 :
1364 0 : #[derive(Serialize, Deserialize, Debug)]
1365 : #[serde(deny_unknown_fields)]
1366 : pub struct TenantLocationConfigRequest {
1367 : #[serde(flatten)]
1368 : pub config: LocationConfig, // as we have a flattened field, we should reject all unknown fields in it
1369 : }
1370 :
1371 0 : #[derive(Serialize, Deserialize, Debug)]
1372 : #[serde(deny_unknown_fields)]
1373 : pub struct TenantTimeTravelRequest {
1374 : pub shard_counts: Vec<ShardCount>,
1375 : }
1376 :
1377 0 : #[derive(Serialize, Deserialize, Debug)]
1378 : #[serde(deny_unknown_fields)]
1379 : pub struct TenantShardLocation {
1380 : pub shard_id: TenantShardId,
1381 : pub node_id: NodeId,
1382 : }
1383 :
1384 0 : #[derive(Serialize, Deserialize, Debug)]
1385 : #[serde(deny_unknown_fields)]
1386 : pub struct TenantLocationConfigResponse {
1387 : pub shards: Vec<TenantShardLocation>,
1388 : // If the shards' ShardCount count is >1, stripe_size will be set.
1389 : pub stripe_size: Option<ShardStripeSize>,
1390 : }
1391 :
1392 2 : #[derive(Serialize, Deserialize, Debug)]
1393 : #[serde(deny_unknown_fields)]
1394 : pub struct TenantConfigRequest {
1395 : pub tenant_id: TenantId,
1396 : #[serde(flatten)]
1397 : pub config: TenantConfig, // as we have a flattened field, we should reject all unknown fields in it
1398 : }
1399 :
1400 : impl std::ops::Deref for TenantConfigRequest {
1401 : type Target = TenantConfig;
1402 :
1403 0 : fn deref(&self) -> &Self::Target {
1404 0 : &self.config
1405 0 : }
1406 : }
1407 :
1408 : impl TenantConfigRequest {
1409 0 : pub fn new(tenant_id: TenantId) -> TenantConfigRequest {
1410 0 : let config = TenantConfig::default();
1411 0 : TenantConfigRequest { tenant_id, config }
1412 0 : }
1413 : }
1414 :
1415 3 : #[derive(Serialize, Deserialize, Debug)]
1416 : #[serde(deny_unknown_fields)]
1417 : pub struct TenantConfigPatchRequest {
1418 : pub tenant_id: TenantId,
1419 : #[serde(flatten)]
1420 : pub config: TenantConfigPatch, // as we have a flattened field, we should reject all unknown fields in it
1421 : }
1422 :
1423 0 : #[derive(Serialize, Deserialize, Debug)]
1424 : pub struct TenantWaitLsnRequest {
1425 : #[serde(flatten)]
1426 : pub timelines: HashMap<TimelineId, Lsn>,
1427 : pub timeout: Duration,
1428 : }
1429 :
1430 : /// See [`TenantState::attachment_status`] and the OpenAPI docs for context.
1431 0 : #[derive(Serialize, Deserialize, Clone)]
1432 : #[serde(tag = "slug", content = "data", rename_all = "snake_case")]
1433 : pub enum TenantAttachmentStatus {
1434 : Maybe,
1435 : Attached,
1436 : Failed { reason: String },
1437 : }
1438 :
1439 0 : #[derive(Serialize, Deserialize, Clone)]
1440 : pub struct TenantInfo {
1441 : pub id: TenantShardId,
1442 : // NB: intentionally not part of OpenAPI, we don't want to commit to a specific set of TenantState's
1443 : pub state: TenantState,
1444 : /// Sum of the size of all layer files.
1445 : /// If a layer is present in both local FS and S3, it counts only once.
1446 : pub current_physical_size: Option<u64>, // physical size is only included in `tenant_status` endpoint
1447 : pub attachment_status: TenantAttachmentStatus,
1448 : pub generation: u32,
1449 :
1450 : /// Opaque explanation if gc is being blocked.
1451 : ///
1452 : /// Only looked up for the individual tenant detail, not the listing.
1453 : #[serde(skip_serializing_if = "Option::is_none")]
1454 : pub gc_blocking: Option<String>,
1455 : }
1456 :
1457 0 : #[derive(Serialize, Deserialize, Clone)]
1458 : pub struct TenantDetails {
1459 : #[serde(flatten)]
1460 : pub tenant_info: TenantInfo,
1461 :
1462 : pub walredo: Option<WalRedoManagerStatus>,
1463 :
1464 : pub timelines: Vec<TimelineId>,
1465 : }
1466 :
1467 0 : #[derive(Serialize, Deserialize, PartialEq, Eq, Clone, Copy, Debug)]
1468 : pub enum TimelineArchivalState {
1469 : Archived,
1470 : Unarchived,
1471 : }
1472 :
1473 0 : #[derive(Serialize, Deserialize, PartialEq, Eq, Clone)]
1474 : pub enum TimelineVisibilityState {
1475 : Visible,
1476 : Invisible,
1477 : }
1478 :
1479 0 : #[derive(Serialize, Deserialize, PartialEq, Eq, Clone)]
1480 : pub struct TimelineArchivalConfigRequest {
1481 : pub state: TimelineArchivalState,
1482 : }
1483 :
1484 0 : #[derive(Serialize, Deserialize, PartialEq, Eq, Clone)]
1485 : pub struct TimelinePatchIndexPartRequest {
1486 : pub rel_size_migration: Option<RelSizeMigration>,
1487 : pub gc_compaction_last_completed_lsn: Option<Lsn>,
1488 : pub applied_gc_cutoff_lsn: Option<Lsn>,
1489 : #[serde(default)]
1490 : pub force_index_update: bool,
1491 : }
1492 :
1493 0 : #[derive(Debug, Serialize, Deserialize, Clone)]
1494 : pub struct TimelinesInfoAndOffloaded {
1495 : pub timelines: Vec<TimelineInfo>,
1496 : pub offloaded: Vec<OffloadedTimelineInfo>,
1497 : }
1498 :
1499 : /// Analog of [`TimelineInfo`] for offloaded timelines.
1500 0 : #[derive(Debug, Serialize, Deserialize, Clone)]
1501 : pub struct OffloadedTimelineInfo {
1502 : pub tenant_id: TenantShardId,
1503 : pub timeline_id: TimelineId,
1504 : /// Whether the timeline has a parent it has been branched off from or not
1505 : pub ancestor_timeline_id: Option<TimelineId>,
1506 : /// Whether to retain the branch lsn at the ancestor or not
1507 : pub ancestor_retain_lsn: Option<Lsn>,
1508 : /// The time point when the timeline was archived
1509 : pub archived_at: chrono::DateTime<chrono::Utc>,
1510 : }
1511 :
1512 4 : #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
1513 : #[serde(rename_all = "camelCase")]
1514 : pub enum RelSizeMigration {
1515 : /// The tenant is using the old rel_size format.
1516 : /// Note that this enum is persisted as `Option<RelSizeMigration>` in the index part, so
1517 : /// `None` is the same as `Some(RelSizeMigration::Legacy)`.
1518 : Legacy,
1519 : /// The tenant is migrating to the new rel_size format. Both old and new rel_size format are
1520 : /// persisted in the index part. The read path will read both formats and merge them.
1521 : Migrating,
1522 : /// The tenant has migrated to the new rel_size format. Only the new rel_size format is persisted
1523 : /// in the index part, and the read path will not read the old format.
1524 : Migrated,
1525 : }
1526 :
1527 : /// This represents the output of the "timeline_detail" and "timeline_list" API calls.
1528 0 : #[derive(Debug, Serialize, Deserialize, Clone)]
1529 : pub struct TimelineInfo {
1530 : pub tenant_id: TenantShardId,
1531 : pub timeline_id: TimelineId,
1532 :
1533 : pub ancestor_timeline_id: Option<TimelineId>,
1534 : pub ancestor_lsn: Option<Lsn>,
1535 : pub last_record_lsn: Lsn,
1536 : pub prev_record_lsn: Option<Lsn>,
1537 :
1538 : /// The LSN up to which GC has advanced: older data may still exist but it is not available for clients.
1539 : /// This LSN is not suitable for deciding where to create branches etc: use [`TimelineInfo::min_readable_lsn`] instead,
1540 : /// as it is easier to reason about.
1541 : #[serde(default)]
1542 : pub applied_gc_cutoff_lsn: Lsn,
1543 :
1544 : /// The upper bound of data which is either already GC'ed, or elegible to be GC'ed at any time based on PITR interval.
1545 : /// This LSN represents the "end of history" for this timeline, and callers should use it to figure out the oldest
1546 : /// LSN at which it is legal to create a branch or ephemeral endpoint.
1547 : ///
1548 : /// Note that holders of valid LSN leases may be able to create branches and read pages earlier
1549 : /// than this LSN, but new leases may not be taken out earlier than this LSN.
1550 : #[serde(default)]
1551 : pub min_readable_lsn: Lsn,
1552 :
1553 : pub disk_consistent_lsn: Lsn,
1554 :
1555 : /// The LSN that we have succesfully uploaded to remote storage
1556 : pub remote_consistent_lsn: Lsn,
1557 :
1558 : /// The LSN that we are advertizing to safekeepers
1559 : pub remote_consistent_lsn_visible: Lsn,
1560 :
1561 : /// The LSN from the start of the root timeline (never changes)
1562 : pub initdb_lsn: Lsn,
1563 :
1564 : pub current_logical_size: u64,
1565 : pub current_logical_size_is_accurate: bool,
1566 :
1567 : pub directory_entries_counts: Vec<u64>,
1568 :
1569 : /// Sum of the size of all layer files.
1570 : /// If a layer is present in both local FS and S3, it counts only once.
1571 : pub current_physical_size: Option<u64>, // is None when timeline is Unloaded
1572 : pub current_logical_size_non_incremental: Option<u64>,
1573 :
1574 : /// How many bytes of WAL are within this branch's pitr_interval. If the pitr_interval goes
1575 : /// beyond the branch's branch point, we only count up to the branch point.
1576 : pub pitr_history_size: u64,
1577 :
1578 : /// Whether this branch's branch point is within its ancestor's PITR interval (i.e. any
1579 : /// ancestor data used by this branch would have been retained anyway). If this is false, then
1580 : /// this branch may be imposing a cost on the ancestor by causing it to retain layers that it would
1581 : /// otherwise be able to GC.
1582 : pub within_ancestor_pitr: bool,
1583 :
1584 : pub timeline_dir_layer_file_size_sum: Option<u64>,
1585 :
1586 : pub wal_source_connstr: Option<String>,
1587 : pub last_received_msg_lsn: Option<Lsn>,
1588 : /// the timestamp (in microseconds) of the last received message
1589 : pub last_received_msg_ts: Option<u128>,
1590 : pub pg_version: u32,
1591 :
1592 : pub state: TimelineState,
1593 :
1594 : pub walreceiver_status: String,
1595 :
1596 : // ALWAYS add new fields at the end of the struct with `Option` to ensure forward/backward compatibility.
1597 : // Backward compatibility: you will get a JSON not containing the newly-added field.
1598 : // Forward compatibility: a previous version of the pageserver will receive a JSON. serde::Deserialize does
1599 : // not deny unknown fields by default so it's safe to set the field to some value, though it won't be
1600 : // read.
1601 : /// Whether the timeline is archived.
1602 : pub is_archived: Option<bool>,
1603 :
1604 : /// The status of the rel_size migration.
1605 : pub rel_size_migration: Option<RelSizeMigration>,
1606 :
1607 : /// Whether the timeline is invisible in synthetic size calculations.
1608 : pub is_invisible: Option<bool>,
1609 : }
1610 :
1611 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1612 : pub struct LayerMapInfo {
1613 : pub in_memory_layers: Vec<InMemoryLayerInfo>,
1614 : pub historic_layers: Vec<HistoricLayerInfo>,
1615 : }
1616 :
1617 : /// The residence status of a layer
1618 0 : #[derive(Debug, Clone, Copy, Serialize, Deserialize)]
1619 : pub enum LayerResidenceStatus {
1620 : /// Residence status for a layer file that exists locally.
1621 : /// It may also exist on the remote, we don't care here.
1622 : Resident,
1623 : /// Residence status for a layer file that only exists on the remote.
1624 : Evicted,
1625 : }
1626 :
1627 : #[serde_as]
1628 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1629 : pub struct LayerAccessStats {
1630 : #[serde_as(as = "serde_with::TimestampMilliSeconds")]
1631 : pub access_time: SystemTime,
1632 :
1633 : #[serde_as(as = "serde_with::TimestampMilliSeconds")]
1634 : pub residence_time: SystemTime,
1635 :
1636 : pub visible: bool,
1637 : }
1638 :
1639 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1640 : #[serde(tag = "kind")]
1641 : pub enum InMemoryLayerInfo {
1642 : Open { lsn_start: Lsn },
1643 : Frozen { lsn_start: Lsn, lsn_end: Lsn },
1644 : }
1645 :
1646 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1647 : #[serde(tag = "kind")]
1648 : pub enum HistoricLayerInfo {
1649 : Delta {
1650 : layer_file_name: String,
1651 : layer_file_size: u64,
1652 :
1653 : lsn_start: Lsn,
1654 : lsn_end: Lsn,
1655 : remote: bool,
1656 : access_stats: LayerAccessStats,
1657 :
1658 : l0: bool,
1659 : },
1660 : Image {
1661 : layer_file_name: String,
1662 : layer_file_size: u64,
1663 :
1664 : lsn_start: Lsn,
1665 : remote: bool,
1666 : access_stats: LayerAccessStats,
1667 : },
1668 : }
1669 :
1670 : impl HistoricLayerInfo {
1671 0 : pub fn layer_file_name(&self) -> &str {
1672 0 : match self {
1673 : HistoricLayerInfo::Delta {
1674 0 : layer_file_name, ..
1675 0 : } => layer_file_name,
1676 : HistoricLayerInfo::Image {
1677 0 : layer_file_name, ..
1678 0 : } => layer_file_name,
1679 : }
1680 0 : }
1681 0 : pub fn is_remote(&self) -> bool {
1682 0 : match self {
1683 0 : HistoricLayerInfo::Delta { remote, .. } => *remote,
1684 0 : HistoricLayerInfo::Image { remote, .. } => *remote,
1685 : }
1686 0 : }
1687 0 : pub fn set_remote(&mut self, value: bool) {
1688 0 : let field = match self {
1689 0 : HistoricLayerInfo::Delta { remote, .. } => remote,
1690 0 : HistoricLayerInfo::Image { remote, .. } => remote,
1691 : };
1692 0 : *field = value;
1693 0 : }
1694 0 : pub fn layer_file_size(&self) -> u64 {
1695 0 : match self {
1696 : HistoricLayerInfo::Delta {
1697 0 : layer_file_size, ..
1698 0 : } => *layer_file_size,
1699 : HistoricLayerInfo::Image {
1700 0 : layer_file_size, ..
1701 0 : } => *layer_file_size,
1702 : }
1703 0 : }
1704 : }
1705 :
1706 0 : #[derive(Debug, Serialize, Deserialize)]
1707 : pub struct DownloadRemoteLayersTaskSpawnRequest {
1708 : pub max_concurrent_downloads: NonZeroUsize,
1709 : }
1710 :
1711 0 : #[derive(Debug, Serialize, Deserialize)]
1712 : pub struct IngestAuxFilesRequest {
1713 : pub aux_files: HashMap<String, String>,
1714 : }
1715 :
1716 0 : #[derive(Debug, Serialize, Deserialize)]
1717 : pub struct ListAuxFilesRequest {
1718 : pub lsn: Lsn,
1719 : }
1720 :
1721 0 : #[derive(Debug, Serialize, Deserialize, Clone)]
1722 : pub struct DownloadRemoteLayersTaskInfo {
1723 : pub task_id: String,
1724 : pub state: DownloadRemoteLayersTaskState,
1725 : pub total_layer_count: u64, // stable once `completed`
1726 : pub successful_download_count: u64, // stable once `completed`
1727 : pub failed_download_count: u64, // stable once `completed`
1728 : }
1729 :
1730 0 : #[derive(Debug, Serialize, Deserialize, Clone)]
1731 : pub enum DownloadRemoteLayersTaskState {
1732 : Running,
1733 : Completed,
1734 : ShutDown,
1735 : }
1736 :
1737 0 : #[derive(Debug, Serialize, Deserialize)]
1738 : pub struct TimelineGcRequest {
1739 : pub gc_horizon: Option<u64>,
1740 : }
1741 :
1742 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1743 : pub struct WalRedoManagerProcessStatus {
1744 : pub pid: u32,
1745 : }
1746 :
1747 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1748 : pub struct WalRedoManagerStatus {
1749 : pub last_redo_at: Option<chrono::DateTime<chrono::Utc>>,
1750 : pub process: Option<WalRedoManagerProcessStatus>,
1751 : }
1752 :
1753 : /// The progress of a secondary tenant.
1754 : ///
1755 : /// It is mostly useful when doing a long running download: e.g. initiating
1756 : /// a download job, timing out while waiting for it to run, and then inspecting this status to understand
1757 : /// what's happening.
1758 0 : #[derive(Default, Debug, Serialize, Deserialize, Clone)]
1759 : pub struct SecondaryProgress {
1760 : /// The remote storage LastModified time of the heatmap object we last downloaded.
1761 : pub heatmap_mtime: Option<serde_system_time::SystemTime>,
1762 :
1763 : /// The number of layers currently on-disk
1764 : pub layers_downloaded: usize,
1765 : /// The number of layers in the most recently seen heatmap
1766 : pub layers_total: usize,
1767 :
1768 : /// The number of layer bytes currently on-disk
1769 : pub bytes_downloaded: u64,
1770 : /// The number of layer bytes in the most recently seen heatmap
1771 : pub bytes_total: u64,
1772 : }
1773 :
1774 0 : #[derive(Serialize, Deserialize, Debug)]
1775 : pub struct TenantScanRemoteStorageShard {
1776 : pub tenant_shard_id: TenantShardId,
1777 : pub generation: Option<u32>,
1778 : pub stripe_size: Option<ShardStripeSize>,
1779 : }
1780 :
1781 0 : #[derive(Serialize, Deserialize, Debug, Default)]
1782 : pub struct TenantScanRemoteStorageResponse {
1783 : pub shards: Vec<TenantScanRemoteStorageShard>,
1784 : }
1785 :
1786 0 : #[derive(Serialize, Deserialize, Debug, Clone)]
1787 : #[serde(rename_all = "snake_case")]
1788 : pub enum TenantSorting {
1789 : /// Total size of layers on local disk for all timelines in a shard.
1790 : ResidentSize,
1791 : /// The logical size of the largest timeline within a _tenant_ (not shard). Only tracked on
1792 : /// shard 0, contains the sum across all shards.
1793 : MaxLogicalSize,
1794 : /// The logical size of the largest timeline within a _tenant_ (not shard), divided by number of
1795 : /// shards. Only tracked on shard 0, and estimates the per-shard logical size.
1796 : MaxLogicalSizePerShard,
1797 : }
1798 :
1799 : impl Default for TenantSorting {
1800 0 : fn default() -> Self {
1801 0 : Self::ResidentSize
1802 0 : }
1803 : }
1804 :
1805 0 : #[derive(Serialize, Deserialize, Debug, Clone)]
1806 : pub struct TopTenantShardsRequest {
1807 : // How would you like to sort the tenants?
1808 : pub order_by: TenantSorting,
1809 :
1810 : // How many results?
1811 : pub limit: usize,
1812 :
1813 : // Omit tenants with more than this many shards (e.g. if this is the max number of shards
1814 : // that the caller would ever split to)
1815 : pub where_shards_lt: Option<ShardCount>,
1816 :
1817 : // Omit tenants where the ordering metric is less than this (this is an optimization to
1818 : // let us quickly exclude numerous tiny shards)
1819 : pub where_gt: Option<u64>,
1820 : }
1821 :
1822 0 : #[derive(Serialize, Deserialize, Debug, PartialEq, Eq)]
1823 : pub struct TopTenantShardItem {
1824 : pub id: TenantShardId,
1825 :
1826 : /// Total size of layers on local disk for all timelines in this shard.
1827 : pub resident_size: u64,
1828 :
1829 : /// Total size of layers in remote storage for all timelines in this shard.
1830 : pub physical_size: u64,
1831 :
1832 : /// The largest logical size of a timeline within this _tenant_ (not shard). This is only
1833 : /// tracked on shard 0, and contains the sum of the logical size across all shards.
1834 : pub max_logical_size: u64,
1835 :
1836 : /// The largest logical size of a timeline within this _tenant_ (not shard) divided by number of
1837 : /// shards. This is only tracked on shard 0, and is only an estimate as we divide it evenly by
1838 : /// shard count, rounded up.
1839 : pub max_logical_size_per_shard: u64,
1840 : }
1841 :
1842 0 : #[derive(Serialize, Deserialize, Debug, Default)]
1843 : pub struct TopTenantShardsResponse {
1844 : pub shards: Vec<TopTenantShardItem>,
1845 : }
1846 :
1847 : pub mod virtual_file {
1848 :
1849 : #[derive(
1850 : Copy,
1851 : Clone,
1852 : PartialEq,
1853 : Eq,
1854 : Hash,
1855 0 : strum_macros::EnumString,
1856 : strum_macros::Display,
1857 0 : serde_with::DeserializeFromStr,
1858 : serde_with::SerializeDisplay,
1859 : Debug,
1860 : )]
1861 : #[strum(serialize_all = "kebab-case")]
1862 : pub enum IoEngineKind {
1863 : StdFs,
1864 : #[cfg(target_os = "linux")]
1865 : TokioEpollUring,
1866 : }
1867 :
1868 : /// Direct IO modes for a pageserver.
1869 : #[derive(
1870 : Copy,
1871 : Clone,
1872 : PartialEq,
1873 : Eq,
1874 : Hash,
1875 0 : strum_macros::EnumString,
1876 0 : strum_macros::EnumIter,
1877 : strum_macros::Display,
1878 0 : serde_with::DeserializeFromStr,
1879 : serde_with::SerializeDisplay,
1880 : Debug,
1881 : )]
1882 : #[strum(serialize_all = "kebab-case")]
1883 : #[repr(u8)]
1884 : pub enum IoMode {
1885 : /// Uses buffered IO.
1886 : Buffered,
1887 : /// Uses direct IO for reads only.
1888 : Direct,
1889 : /// Use direct IO for reads and writes.
1890 : DirectRw,
1891 : }
1892 :
1893 : impl IoMode {
1894 247 : pub fn preferred() -> Self {
1895 247 : IoMode::DirectRw
1896 247 : }
1897 : }
1898 :
1899 : impl TryFrom<u8> for IoMode {
1900 : type Error = u8;
1901 :
1902 2576 : fn try_from(value: u8) -> Result<Self, Self::Error> {
1903 2576 : Ok(match value {
1904 2576 : v if v == (IoMode::Buffered as u8) => IoMode::Buffered,
1905 2576 : v if v == (IoMode::Direct as u8) => IoMode::Direct,
1906 2576 : v if v == (IoMode::DirectRw as u8) => IoMode::DirectRw,
1907 0 : x => return Err(x),
1908 : })
1909 2576 : }
1910 : }
1911 : }
1912 :
1913 0 : #[derive(Debug, Clone, Serialize, Deserialize)]
1914 : pub struct ScanDisposableKeysResponse {
1915 : pub disposable_count: usize,
1916 : pub not_disposable_count: usize,
1917 : }
1918 :
1919 : // Wrapped in libpq CopyData
1920 : #[derive(PartialEq, Eq, Debug)]
1921 : pub enum PagestreamFeMessage {
1922 : Exists(PagestreamExistsRequest),
1923 : Nblocks(PagestreamNblocksRequest),
1924 : GetPage(PagestreamGetPageRequest),
1925 : DbSize(PagestreamDbSizeRequest),
1926 : GetSlruSegment(PagestreamGetSlruSegmentRequest),
1927 : #[cfg(feature = "testing")]
1928 : Test(PagestreamTestRequest),
1929 : }
1930 :
1931 : // Wrapped in libpq CopyData
1932 : #[derive(strum_macros::EnumProperty)]
1933 : pub enum PagestreamBeMessage {
1934 : Exists(PagestreamExistsResponse),
1935 : Nblocks(PagestreamNblocksResponse),
1936 : GetPage(PagestreamGetPageResponse),
1937 : Error(PagestreamErrorResponse),
1938 : DbSize(PagestreamDbSizeResponse),
1939 : GetSlruSegment(PagestreamGetSlruSegmentResponse),
1940 : #[cfg(feature = "testing")]
1941 : Test(PagestreamTestResponse),
1942 : }
1943 :
1944 : // Keep in sync with `pagestore_client.h`
1945 : #[repr(u8)]
1946 : enum PagestreamFeMessageTag {
1947 : Exists = 0,
1948 : Nblocks = 1,
1949 : GetPage = 2,
1950 : DbSize = 3,
1951 : GetSlruSegment = 4,
1952 : /* future tags above this line */
1953 : /// For testing purposes, not available in production.
1954 : #[cfg(feature = "testing")]
1955 : Test = 99,
1956 : }
1957 :
1958 : // Keep in sync with `pagestore_client.h`
1959 : #[repr(u8)]
1960 : enum PagestreamBeMessageTag {
1961 : Exists = 100,
1962 : Nblocks = 101,
1963 : GetPage = 102,
1964 : Error = 103,
1965 : DbSize = 104,
1966 : GetSlruSegment = 105,
1967 : /* future tags above this line */
1968 : /// For testing purposes, not available in production.
1969 : #[cfg(feature = "testing")]
1970 : Test = 199,
1971 : }
1972 :
1973 : impl TryFrom<u8> for PagestreamFeMessageTag {
1974 : type Error = u8;
1975 4 : fn try_from(value: u8) -> Result<Self, u8> {
1976 4 : match value {
1977 1 : 0 => Ok(PagestreamFeMessageTag::Exists),
1978 1 : 1 => Ok(PagestreamFeMessageTag::Nblocks),
1979 1 : 2 => Ok(PagestreamFeMessageTag::GetPage),
1980 1 : 3 => Ok(PagestreamFeMessageTag::DbSize),
1981 0 : 4 => Ok(PagestreamFeMessageTag::GetSlruSegment),
1982 : #[cfg(feature = "testing")]
1983 0 : 99 => Ok(PagestreamFeMessageTag::Test),
1984 0 : _ => Err(value),
1985 : }
1986 4 : }
1987 : }
1988 :
1989 : impl TryFrom<u8> for PagestreamBeMessageTag {
1990 : type Error = u8;
1991 0 : fn try_from(value: u8) -> Result<Self, u8> {
1992 0 : match value {
1993 0 : 100 => Ok(PagestreamBeMessageTag::Exists),
1994 0 : 101 => Ok(PagestreamBeMessageTag::Nblocks),
1995 0 : 102 => Ok(PagestreamBeMessageTag::GetPage),
1996 0 : 103 => Ok(PagestreamBeMessageTag::Error),
1997 0 : 104 => Ok(PagestreamBeMessageTag::DbSize),
1998 0 : 105 => Ok(PagestreamBeMessageTag::GetSlruSegment),
1999 : #[cfg(feature = "testing")]
2000 0 : 199 => Ok(PagestreamBeMessageTag::Test),
2001 0 : _ => Err(value),
2002 : }
2003 0 : }
2004 : }
2005 :
2006 : // A GetPage request contains two LSN values:
2007 : //
2008 : // request_lsn: Get the page version at this point in time. Lsn::Max is a special value that means
2009 : // "get the latest version present". It's used by the primary server, which knows that no one else
2010 : // is writing WAL. 'not_modified_since' must be set to a proper value even if request_lsn is
2011 : // Lsn::Max. Standby servers use the current replay LSN as the request LSN.
2012 : //
2013 : // not_modified_since: Hint to the pageserver that the client knows that the page has not been
2014 : // modified between 'not_modified_since' and the request LSN. It's always correct to set
2015 : // 'not_modified_since equal' to 'request_lsn' (unless Lsn::Max is used as the 'request_lsn'), but
2016 : // passing an earlier LSN can speed up the request, by allowing the pageserver to process the
2017 : // request without waiting for 'request_lsn' to arrive.
2018 : //
2019 : // The now-defunct V1 interface contained only one LSN, and a boolean 'latest' flag. The V1 interface was
2020 : // sufficient for the primary; the 'lsn' was equivalent to the 'not_modified_since' value, and
2021 : // 'latest' was set to true. The V2 interface was added because there was no correct way for a
2022 : // standby to request a page at a particular non-latest LSN, and also include the
2023 : // 'not_modified_since' hint. That led to an awkward choice of either using an old LSN in the
2024 : // request, if the standby knows that the page hasn't been modified since, and risk getting an error
2025 : // if that LSN has fallen behind the GC horizon, or requesting the current replay LSN, which could
2026 : // require the pageserver unnecessarily to wait for the WAL to arrive up to that point. The new V2
2027 : // interface allows sending both LSNs, and let the pageserver do the right thing. There was no
2028 : // difference in the responses between V1 and V2.
2029 : //
2030 : // V3 version of protocol adds request ID to all requests. This request ID is also included in response
2031 : // as well as other fields from requests, which allows to verify that we receive response for our request.
2032 : // We copy fields from request to response to make checking more reliable: request ID is formed from process ID
2033 : // and local counter, so in principle there can be duplicated requests IDs if process PID is reused.
2034 : //
2035 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2036 : pub enum PagestreamProtocolVersion {
2037 : V2,
2038 : V3,
2039 : }
2040 :
2041 : pub type RequestId = u64;
2042 :
2043 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2044 : pub struct PagestreamRequest {
2045 : pub reqid: RequestId,
2046 : pub request_lsn: Lsn,
2047 : pub not_modified_since: Lsn,
2048 : }
2049 :
2050 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2051 : pub struct PagestreamExistsRequest {
2052 : pub hdr: PagestreamRequest,
2053 : pub rel: RelTag,
2054 : }
2055 :
2056 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2057 : pub struct PagestreamNblocksRequest {
2058 : pub hdr: PagestreamRequest,
2059 : pub rel: RelTag,
2060 : }
2061 :
2062 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2063 : pub struct PagestreamGetPageRequest {
2064 : pub hdr: PagestreamRequest,
2065 : pub rel: RelTag,
2066 : pub blkno: u32,
2067 : }
2068 :
2069 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2070 : pub struct PagestreamDbSizeRequest {
2071 : pub hdr: PagestreamRequest,
2072 : pub dbnode: u32,
2073 : }
2074 :
2075 : #[derive(Debug, PartialEq, Eq, Clone, Copy)]
2076 : pub struct PagestreamGetSlruSegmentRequest {
2077 : pub hdr: PagestreamRequest,
2078 : pub kind: u8,
2079 : pub segno: u32,
2080 : }
2081 :
2082 : #[derive(Debug)]
2083 : pub struct PagestreamExistsResponse {
2084 : pub req: PagestreamExistsRequest,
2085 : pub exists: bool,
2086 : }
2087 :
2088 : #[derive(Debug)]
2089 : pub struct PagestreamNblocksResponse {
2090 : pub req: PagestreamNblocksRequest,
2091 : pub n_blocks: u32,
2092 : }
2093 :
2094 : #[derive(Debug)]
2095 : pub struct PagestreamGetPageResponse {
2096 : pub req: PagestreamGetPageRequest,
2097 : pub page: Bytes,
2098 : }
2099 :
2100 : #[derive(Debug)]
2101 : pub struct PagestreamGetSlruSegmentResponse {
2102 : pub req: PagestreamGetSlruSegmentRequest,
2103 : pub segment: Bytes,
2104 : }
2105 :
2106 : #[derive(Debug)]
2107 : pub struct PagestreamErrorResponse {
2108 : pub req: PagestreamRequest,
2109 : pub message: String,
2110 : }
2111 :
2112 : #[derive(Debug)]
2113 : pub struct PagestreamDbSizeResponse {
2114 : pub req: PagestreamDbSizeRequest,
2115 : pub db_size: i64,
2116 : }
2117 :
2118 : #[cfg(feature = "testing")]
2119 : #[derive(Debug, PartialEq, Eq, Clone)]
2120 : pub struct PagestreamTestRequest {
2121 : pub hdr: PagestreamRequest,
2122 : pub batch_key: u64,
2123 : pub message: String,
2124 : }
2125 :
2126 : #[cfg(feature = "testing")]
2127 : #[derive(Debug)]
2128 : pub struct PagestreamTestResponse {
2129 : pub req: PagestreamTestRequest,
2130 : }
2131 :
2132 : // This is a cut-down version of TenantHistorySize from the pageserver crate, omitting fields
2133 : // that require pageserver-internal types. It is sufficient to get the total size.
2134 0 : #[derive(Serialize, Deserialize, Debug)]
2135 : pub struct TenantHistorySize {
2136 : pub id: TenantId,
2137 : /// Size is a mixture of WAL and logical size, so the unit is bytes.
2138 : ///
2139 : /// Will be none if `?inputs_only=true` was given.
2140 : pub size: Option<u64>,
2141 : }
2142 :
2143 : impl PagestreamFeMessage {
2144 : /// Serialize a compute -> pageserver message. This is currently only used in testing
2145 : /// tools. Always uses protocol version 3.
2146 4 : pub fn serialize(&self) -> Bytes {
2147 4 : let mut bytes = BytesMut::new();
2148 4 :
2149 4 : match self {
2150 1 : Self::Exists(req) => {
2151 1 : bytes.put_u8(PagestreamFeMessageTag::Exists as u8);
2152 1 : bytes.put_u64(req.hdr.reqid);
2153 1 : bytes.put_u64(req.hdr.request_lsn.0);
2154 1 : bytes.put_u64(req.hdr.not_modified_since.0);
2155 1 : bytes.put_u32(req.rel.spcnode);
2156 1 : bytes.put_u32(req.rel.dbnode);
2157 1 : bytes.put_u32(req.rel.relnode);
2158 1 : bytes.put_u8(req.rel.forknum);
2159 1 : }
2160 :
2161 1 : Self::Nblocks(req) => {
2162 1 : bytes.put_u8(PagestreamFeMessageTag::Nblocks as u8);
2163 1 : bytes.put_u64(req.hdr.reqid);
2164 1 : bytes.put_u64(req.hdr.request_lsn.0);
2165 1 : bytes.put_u64(req.hdr.not_modified_since.0);
2166 1 : bytes.put_u32(req.rel.spcnode);
2167 1 : bytes.put_u32(req.rel.dbnode);
2168 1 : bytes.put_u32(req.rel.relnode);
2169 1 : bytes.put_u8(req.rel.forknum);
2170 1 : }
2171 :
2172 1 : Self::GetPage(req) => {
2173 1 : bytes.put_u8(PagestreamFeMessageTag::GetPage as u8);
2174 1 : bytes.put_u64(req.hdr.reqid);
2175 1 : bytes.put_u64(req.hdr.request_lsn.0);
2176 1 : bytes.put_u64(req.hdr.not_modified_since.0);
2177 1 : bytes.put_u32(req.rel.spcnode);
2178 1 : bytes.put_u32(req.rel.dbnode);
2179 1 : bytes.put_u32(req.rel.relnode);
2180 1 : bytes.put_u8(req.rel.forknum);
2181 1 : bytes.put_u32(req.blkno);
2182 1 : }
2183 :
2184 1 : Self::DbSize(req) => {
2185 1 : bytes.put_u8(PagestreamFeMessageTag::DbSize as u8);
2186 1 : bytes.put_u64(req.hdr.reqid);
2187 1 : bytes.put_u64(req.hdr.request_lsn.0);
2188 1 : bytes.put_u64(req.hdr.not_modified_since.0);
2189 1 : bytes.put_u32(req.dbnode);
2190 1 : }
2191 :
2192 0 : Self::GetSlruSegment(req) => {
2193 0 : bytes.put_u8(PagestreamFeMessageTag::GetSlruSegment as u8);
2194 0 : bytes.put_u64(req.hdr.reqid);
2195 0 : bytes.put_u64(req.hdr.request_lsn.0);
2196 0 : bytes.put_u64(req.hdr.not_modified_since.0);
2197 0 : bytes.put_u8(req.kind);
2198 0 : bytes.put_u32(req.segno);
2199 0 : }
2200 : #[cfg(feature = "testing")]
2201 0 : Self::Test(req) => {
2202 0 : bytes.put_u8(PagestreamFeMessageTag::Test as u8);
2203 0 : bytes.put_u64(req.hdr.reqid);
2204 0 : bytes.put_u64(req.hdr.request_lsn.0);
2205 0 : bytes.put_u64(req.hdr.not_modified_since.0);
2206 0 : bytes.put_u64(req.batch_key);
2207 0 : let message = req.message.as_bytes();
2208 0 : bytes.put_u64(message.len() as u64);
2209 0 : bytes.put_slice(message);
2210 0 : }
2211 : }
2212 :
2213 4 : bytes.into()
2214 4 : }
2215 :
2216 4 : pub fn parse<R: std::io::Read>(
2217 4 : body: &mut R,
2218 4 : protocol_version: PagestreamProtocolVersion,
2219 4 : ) -> anyhow::Result<PagestreamFeMessage> {
2220 : // these correspond to the NeonMessageTag enum in pagestore_client.h
2221 : //
2222 : // TODO: consider using protobuf or serde bincode for less error prone
2223 : // serialization.
2224 4 : let msg_tag = body.read_u8()?;
2225 4 : let (reqid, request_lsn, not_modified_since) = match protocol_version {
2226 : PagestreamProtocolVersion::V2 => (
2227 : 0,
2228 0 : Lsn::from(body.read_u64::<BigEndian>()?),
2229 0 : Lsn::from(body.read_u64::<BigEndian>()?),
2230 : ),
2231 : PagestreamProtocolVersion::V3 => (
2232 4 : body.read_u64::<BigEndian>()?,
2233 4 : Lsn::from(body.read_u64::<BigEndian>()?),
2234 4 : Lsn::from(body.read_u64::<BigEndian>()?),
2235 : ),
2236 : };
2237 :
2238 4 : match PagestreamFeMessageTag::try_from(msg_tag)
2239 4 : .map_err(|tag: u8| anyhow::anyhow!("invalid tag {tag}"))?
2240 : {
2241 : PagestreamFeMessageTag::Exists => {
2242 : Ok(PagestreamFeMessage::Exists(PagestreamExistsRequest {
2243 1 : hdr: PagestreamRequest {
2244 1 : reqid,
2245 1 : request_lsn,
2246 1 : not_modified_since,
2247 1 : },
2248 1 : rel: RelTag {
2249 1 : spcnode: body.read_u32::<BigEndian>()?,
2250 1 : dbnode: body.read_u32::<BigEndian>()?,
2251 1 : relnode: body.read_u32::<BigEndian>()?,
2252 1 : forknum: body.read_u8()?,
2253 : },
2254 : }))
2255 : }
2256 : PagestreamFeMessageTag::Nblocks => {
2257 : Ok(PagestreamFeMessage::Nblocks(PagestreamNblocksRequest {
2258 1 : hdr: PagestreamRequest {
2259 1 : reqid,
2260 1 : request_lsn,
2261 1 : not_modified_since,
2262 1 : },
2263 1 : rel: RelTag {
2264 1 : spcnode: body.read_u32::<BigEndian>()?,
2265 1 : dbnode: body.read_u32::<BigEndian>()?,
2266 1 : relnode: body.read_u32::<BigEndian>()?,
2267 1 : forknum: body.read_u8()?,
2268 : },
2269 : }))
2270 : }
2271 : PagestreamFeMessageTag::GetPage => {
2272 : Ok(PagestreamFeMessage::GetPage(PagestreamGetPageRequest {
2273 1 : hdr: PagestreamRequest {
2274 1 : reqid,
2275 1 : request_lsn,
2276 1 : not_modified_since,
2277 1 : },
2278 1 : rel: RelTag {
2279 1 : spcnode: body.read_u32::<BigEndian>()?,
2280 1 : dbnode: body.read_u32::<BigEndian>()?,
2281 1 : relnode: body.read_u32::<BigEndian>()?,
2282 1 : forknum: body.read_u8()?,
2283 : },
2284 1 : blkno: body.read_u32::<BigEndian>()?,
2285 : }))
2286 : }
2287 : PagestreamFeMessageTag::DbSize => {
2288 : Ok(PagestreamFeMessage::DbSize(PagestreamDbSizeRequest {
2289 1 : hdr: PagestreamRequest {
2290 1 : reqid,
2291 1 : request_lsn,
2292 1 : not_modified_since,
2293 1 : },
2294 1 : dbnode: body.read_u32::<BigEndian>()?,
2295 : }))
2296 : }
2297 : PagestreamFeMessageTag::GetSlruSegment => Ok(PagestreamFeMessage::GetSlruSegment(
2298 : PagestreamGetSlruSegmentRequest {
2299 0 : hdr: PagestreamRequest {
2300 0 : reqid,
2301 0 : request_lsn,
2302 0 : not_modified_since,
2303 0 : },
2304 0 : kind: body.read_u8()?,
2305 0 : segno: body.read_u32::<BigEndian>()?,
2306 : },
2307 : )),
2308 : #[cfg(feature = "testing")]
2309 : PagestreamFeMessageTag::Test => Ok(PagestreamFeMessage::Test(PagestreamTestRequest {
2310 0 : hdr: PagestreamRequest {
2311 0 : reqid,
2312 0 : request_lsn,
2313 0 : not_modified_since,
2314 0 : },
2315 0 : batch_key: body.read_u64::<BigEndian>()?,
2316 : message: {
2317 0 : let len = body.read_u64::<BigEndian>()?;
2318 0 : let mut buf = vec![0; len as usize];
2319 0 : body.read_exact(&mut buf)?;
2320 0 : String::from_utf8(buf)?
2321 : },
2322 : })),
2323 : }
2324 4 : }
2325 : }
2326 :
2327 : impl PagestreamBeMessage {
2328 0 : pub fn serialize(&self, protocol_version: PagestreamProtocolVersion) -> Bytes {
2329 0 : let mut bytes = BytesMut::new();
2330 :
2331 : use PagestreamBeMessageTag as Tag;
2332 0 : match protocol_version {
2333 : PagestreamProtocolVersion::V2 => {
2334 0 : match self {
2335 0 : Self::Exists(resp) => {
2336 0 : bytes.put_u8(Tag::Exists as u8);
2337 0 : bytes.put_u8(resp.exists as u8);
2338 0 : }
2339 :
2340 0 : Self::Nblocks(resp) => {
2341 0 : bytes.put_u8(Tag::Nblocks as u8);
2342 0 : bytes.put_u32(resp.n_blocks);
2343 0 : }
2344 :
2345 0 : Self::GetPage(resp) => {
2346 0 : bytes.put_u8(Tag::GetPage as u8);
2347 0 : bytes.put(&resp.page[..])
2348 : }
2349 :
2350 0 : Self::Error(resp) => {
2351 0 : bytes.put_u8(Tag::Error as u8);
2352 0 : bytes.put(resp.message.as_bytes());
2353 0 : bytes.put_u8(0); // null terminator
2354 0 : }
2355 0 : Self::DbSize(resp) => {
2356 0 : bytes.put_u8(Tag::DbSize as u8);
2357 0 : bytes.put_i64(resp.db_size);
2358 0 : }
2359 :
2360 0 : Self::GetSlruSegment(resp) => {
2361 0 : bytes.put_u8(Tag::GetSlruSegment as u8);
2362 0 : bytes.put_u32((resp.segment.len() / BLCKSZ as usize) as u32);
2363 0 : bytes.put(&resp.segment[..]);
2364 0 : }
2365 :
2366 : #[cfg(feature = "testing")]
2367 0 : Self::Test(resp) => {
2368 0 : bytes.put_u8(Tag::Test as u8);
2369 0 : bytes.put_u64(resp.req.batch_key);
2370 0 : let message = resp.req.message.as_bytes();
2371 0 : bytes.put_u64(message.len() as u64);
2372 0 : bytes.put_slice(message);
2373 0 : }
2374 : }
2375 : }
2376 : PagestreamProtocolVersion::V3 => {
2377 0 : match self {
2378 0 : Self::Exists(resp) => {
2379 0 : bytes.put_u8(Tag::Exists as u8);
2380 0 : bytes.put_u64(resp.req.hdr.reqid);
2381 0 : bytes.put_u64(resp.req.hdr.request_lsn.0);
2382 0 : bytes.put_u64(resp.req.hdr.not_modified_since.0);
2383 0 : bytes.put_u32(resp.req.rel.spcnode);
2384 0 : bytes.put_u32(resp.req.rel.dbnode);
2385 0 : bytes.put_u32(resp.req.rel.relnode);
2386 0 : bytes.put_u8(resp.req.rel.forknum);
2387 0 : bytes.put_u8(resp.exists as u8);
2388 0 : }
2389 :
2390 0 : Self::Nblocks(resp) => {
2391 0 : bytes.put_u8(Tag::Nblocks as u8);
2392 0 : bytes.put_u64(resp.req.hdr.reqid);
2393 0 : bytes.put_u64(resp.req.hdr.request_lsn.0);
2394 0 : bytes.put_u64(resp.req.hdr.not_modified_since.0);
2395 0 : bytes.put_u32(resp.req.rel.spcnode);
2396 0 : bytes.put_u32(resp.req.rel.dbnode);
2397 0 : bytes.put_u32(resp.req.rel.relnode);
2398 0 : bytes.put_u8(resp.req.rel.forknum);
2399 0 : bytes.put_u32(resp.n_blocks);
2400 0 : }
2401 :
2402 0 : Self::GetPage(resp) => {
2403 0 : bytes.put_u8(Tag::GetPage as u8);
2404 0 : bytes.put_u64(resp.req.hdr.reqid);
2405 0 : bytes.put_u64(resp.req.hdr.request_lsn.0);
2406 0 : bytes.put_u64(resp.req.hdr.not_modified_since.0);
2407 0 : bytes.put_u32(resp.req.rel.spcnode);
2408 0 : bytes.put_u32(resp.req.rel.dbnode);
2409 0 : bytes.put_u32(resp.req.rel.relnode);
2410 0 : bytes.put_u8(resp.req.rel.forknum);
2411 0 : bytes.put_u32(resp.req.blkno);
2412 0 : bytes.put(&resp.page[..])
2413 : }
2414 :
2415 0 : Self::Error(resp) => {
2416 0 : bytes.put_u8(Tag::Error as u8);
2417 0 : bytes.put_u64(resp.req.reqid);
2418 0 : bytes.put_u64(resp.req.request_lsn.0);
2419 0 : bytes.put_u64(resp.req.not_modified_since.0);
2420 0 : bytes.put(resp.message.as_bytes());
2421 0 : bytes.put_u8(0); // null terminator
2422 0 : }
2423 0 : Self::DbSize(resp) => {
2424 0 : bytes.put_u8(Tag::DbSize as u8);
2425 0 : bytes.put_u64(resp.req.hdr.reqid);
2426 0 : bytes.put_u64(resp.req.hdr.request_lsn.0);
2427 0 : bytes.put_u64(resp.req.hdr.not_modified_since.0);
2428 0 : bytes.put_u32(resp.req.dbnode);
2429 0 : bytes.put_i64(resp.db_size);
2430 0 : }
2431 :
2432 0 : Self::GetSlruSegment(resp) => {
2433 0 : bytes.put_u8(Tag::GetSlruSegment as u8);
2434 0 : bytes.put_u64(resp.req.hdr.reqid);
2435 0 : bytes.put_u64(resp.req.hdr.request_lsn.0);
2436 0 : bytes.put_u64(resp.req.hdr.not_modified_since.0);
2437 0 : bytes.put_u8(resp.req.kind);
2438 0 : bytes.put_u32(resp.req.segno);
2439 0 : bytes.put_u32((resp.segment.len() / BLCKSZ as usize) as u32);
2440 0 : bytes.put(&resp.segment[..]);
2441 0 : }
2442 :
2443 : #[cfg(feature = "testing")]
2444 0 : Self::Test(resp) => {
2445 0 : bytes.put_u8(Tag::Test as u8);
2446 0 : bytes.put_u64(resp.req.hdr.reqid);
2447 0 : bytes.put_u64(resp.req.hdr.request_lsn.0);
2448 0 : bytes.put_u64(resp.req.hdr.not_modified_since.0);
2449 0 : bytes.put_u64(resp.req.batch_key);
2450 0 : let message = resp.req.message.as_bytes();
2451 0 : bytes.put_u64(message.len() as u64);
2452 0 : bytes.put_slice(message);
2453 0 : }
2454 : }
2455 : }
2456 : }
2457 0 : bytes.into()
2458 0 : }
2459 :
2460 0 : pub fn deserialize(buf: Bytes) -> anyhow::Result<Self> {
2461 0 : let mut buf = buf.reader();
2462 0 : let msg_tag = buf.read_u8()?;
2463 :
2464 : use PagestreamBeMessageTag as Tag;
2465 0 : let ok =
2466 0 : match Tag::try_from(msg_tag).map_err(|tag: u8| anyhow::anyhow!("invalid tag {tag}"))? {
2467 : Tag::Exists => {
2468 0 : let reqid = buf.read_u64::<BigEndian>()?;
2469 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2470 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2471 0 : let rel = RelTag {
2472 0 : spcnode: buf.read_u32::<BigEndian>()?,
2473 0 : dbnode: buf.read_u32::<BigEndian>()?,
2474 0 : relnode: buf.read_u32::<BigEndian>()?,
2475 0 : forknum: buf.read_u8()?,
2476 : };
2477 0 : let exists = buf.read_u8()? != 0;
2478 0 : Self::Exists(PagestreamExistsResponse {
2479 0 : req: PagestreamExistsRequest {
2480 0 : hdr: PagestreamRequest {
2481 0 : reqid,
2482 0 : request_lsn,
2483 0 : not_modified_since,
2484 0 : },
2485 0 : rel,
2486 0 : },
2487 0 : exists,
2488 0 : })
2489 : }
2490 : Tag::Nblocks => {
2491 0 : let reqid = buf.read_u64::<BigEndian>()?;
2492 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2493 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2494 0 : let rel = RelTag {
2495 0 : spcnode: buf.read_u32::<BigEndian>()?,
2496 0 : dbnode: buf.read_u32::<BigEndian>()?,
2497 0 : relnode: buf.read_u32::<BigEndian>()?,
2498 0 : forknum: buf.read_u8()?,
2499 : };
2500 0 : let n_blocks = buf.read_u32::<BigEndian>()?;
2501 0 : Self::Nblocks(PagestreamNblocksResponse {
2502 0 : req: PagestreamNblocksRequest {
2503 0 : hdr: PagestreamRequest {
2504 0 : reqid,
2505 0 : request_lsn,
2506 0 : not_modified_since,
2507 0 : },
2508 0 : rel,
2509 0 : },
2510 0 : n_blocks,
2511 0 : })
2512 : }
2513 : Tag::GetPage => {
2514 0 : let reqid = buf.read_u64::<BigEndian>()?;
2515 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2516 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2517 0 : let rel = RelTag {
2518 0 : spcnode: buf.read_u32::<BigEndian>()?,
2519 0 : dbnode: buf.read_u32::<BigEndian>()?,
2520 0 : relnode: buf.read_u32::<BigEndian>()?,
2521 0 : forknum: buf.read_u8()?,
2522 : };
2523 0 : let blkno = buf.read_u32::<BigEndian>()?;
2524 0 : let mut page = vec![0; 8192]; // TODO: use MaybeUninit
2525 0 : buf.read_exact(&mut page)?;
2526 0 : Self::GetPage(PagestreamGetPageResponse {
2527 0 : req: PagestreamGetPageRequest {
2528 0 : hdr: PagestreamRequest {
2529 0 : reqid,
2530 0 : request_lsn,
2531 0 : not_modified_since,
2532 0 : },
2533 0 : rel,
2534 0 : blkno,
2535 0 : },
2536 0 : page: page.into(),
2537 0 : })
2538 : }
2539 : Tag::Error => {
2540 0 : let reqid = buf.read_u64::<BigEndian>()?;
2541 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2542 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2543 0 : let mut msg = Vec::new();
2544 0 : buf.read_until(0, &mut msg)?;
2545 0 : let cstring = std::ffi::CString::from_vec_with_nul(msg)?;
2546 0 : let rust_str = cstring.to_str()?;
2547 0 : Self::Error(PagestreamErrorResponse {
2548 0 : req: PagestreamRequest {
2549 0 : reqid,
2550 0 : request_lsn,
2551 0 : not_modified_since,
2552 0 : },
2553 0 : message: rust_str.to_owned(),
2554 0 : })
2555 : }
2556 : Tag::DbSize => {
2557 0 : let reqid = buf.read_u64::<BigEndian>()?;
2558 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2559 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2560 0 : let dbnode = buf.read_u32::<BigEndian>()?;
2561 0 : let db_size = buf.read_i64::<BigEndian>()?;
2562 0 : Self::DbSize(PagestreamDbSizeResponse {
2563 0 : req: PagestreamDbSizeRequest {
2564 0 : hdr: PagestreamRequest {
2565 0 : reqid,
2566 0 : request_lsn,
2567 0 : not_modified_since,
2568 0 : },
2569 0 : dbnode,
2570 0 : },
2571 0 : db_size,
2572 0 : })
2573 : }
2574 : Tag::GetSlruSegment => {
2575 0 : let reqid = buf.read_u64::<BigEndian>()?;
2576 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2577 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2578 0 : let kind = buf.read_u8()?;
2579 0 : let segno = buf.read_u32::<BigEndian>()?;
2580 0 : let n_blocks = buf.read_u32::<BigEndian>()?;
2581 0 : let mut segment = vec![0; n_blocks as usize * BLCKSZ as usize];
2582 0 : buf.read_exact(&mut segment)?;
2583 0 : Self::GetSlruSegment(PagestreamGetSlruSegmentResponse {
2584 0 : req: PagestreamGetSlruSegmentRequest {
2585 0 : hdr: PagestreamRequest {
2586 0 : reqid,
2587 0 : request_lsn,
2588 0 : not_modified_since,
2589 0 : },
2590 0 : kind,
2591 0 : segno,
2592 0 : },
2593 0 : segment: segment.into(),
2594 0 : })
2595 : }
2596 : #[cfg(feature = "testing")]
2597 : Tag::Test => {
2598 0 : let reqid = buf.read_u64::<BigEndian>()?;
2599 0 : let request_lsn = Lsn(buf.read_u64::<BigEndian>()?);
2600 0 : let not_modified_since = Lsn(buf.read_u64::<BigEndian>()?);
2601 0 : let batch_key = buf.read_u64::<BigEndian>()?;
2602 0 : let len = buf.read_u64::<BigEndian>()?;
2603 0 : let mut msg = vec![0; len as usize];
2604 0 : buf.read_exact(&mut msg)?;
2605 0 : let message = String::from_utf8(msg)?;
2606 0 : Self::Test(PagestreamTestResponse {
2607 0 : req: PagestreamTestRequest {
2608 0 : hdr: PagestreamRequest {
2609 0 : reqid,
2610 0 : request_lsn,
2611 0 : not_modified_since,
2612 0 : },
2613 0 : batch_key,
2614 0 : message,
2615 0 : },
2616 0 : })
2617 : }
2618 : };
2619 0 : let remaining = buf.into_inner();
2620 0 : if !remaining.is_empty() {
2621 0 : anyhow::bail!(
2622 0 : "remaining bytes in msg with tag={msg_tag}: {}",
2623 0 : remaining.len()
2624 0 : );
2625 0 : }
2626 0 : Ok(ok)
2627 0 : }
2628 :
2629 0 : pub fn kind(&self) -> &'static str {
2630 0 : match self {
2631 0 : Self::Exists(_) => "Exists",
2632 0 : Self::Nblocks(_) => "Nblocks",
2633 0 : Self::GetPage(_) => "GetPage",
2634 0 : Self::Error(_) => "Error",
2635 0 : Self::DbSize(_) => "DbSize",
2636 0 : Self::GetSlruSegment(_) => "GetSlruSegment",
2637 : #[cfg(feature = "testing")]
2638 0 : Self::Test(_) => "Test",
2639 : }
2640 0 : }
2641 : }
2642 :
2643 0 : #[derive(Debug, Serialize, Deserialize)]
2644 : pub struct PageTraceEvent {
2645 : pub key: CompactKey,
2646 : pub effective_lsn: Lsn,
2647 : pub time: SystemTime,
2648 : }
2649 :
2650 : impl Default for PageTraceEvent {
2651 0 : fn default() -> Self {
2652 0 : Self {
2653 0 : key: Default::default(),
2654 0 : effective_lsn: Default::default(),
2655 0 : time: std::time::UNIX_EPOCH,
2656 0 : }
2657 0 : }
2658 : }
2659 :
2660 : #[cfg(test)]
2661 : mod tests {
2662 : use std::str::FromStr;
2663 :
2664 : use serde_json::json;
2665 :
2666 : use super::*;
2667 :
2668 : #[test]
2669 1 : fn test_pagestream() {
2670 1 : // Test serialization/deserialization of PagestreamFeMessage
2671 1 : let messages = vec![
2672 1 : PagestreamFeMessage::Exists(PagestreamExistsRequest {
2673 1 : hdr: PagestreamRequest {
2674 1 : reqid: 0,
2675 1 : request_lsn: Lsn(4),
2676 1 : not_modified_since: Lsn(3),
2677 1 : },
2678 1 : rel: RelTag {
2679 1 : forknum: 1,
2680 1 : spcnode: 2,
2681 1 : dbnode: 3,
2682 1 : relnode: 4,
2683 1 : },
2684 1 : }),
2685 1 : PagestreamFeMessage::Nblocks(PagestreamNblocksRequest {
2686 1 : hdr: PagestreamRequest {
2687 1 : reqid: 0,
2688 1 : request_lsn: Lsn(4),
2689 1 : not_modified_since: Lsn(4),
2690 1 : },
2691 1 : rel: RelTag {
2692 1 : forknum: 1,
2693 1 : spcnode: 2,
2694 1 : dbnode: 3,
2695 1 : relnode: 4,
2696 1 : },
2697 1 : }),
2698 1 : PagestreamFeMessage::GetPage(PagestreamGetPageRequest {
2699 1 : hdr: PagestreamRequest {
2700 1 : reqid: 0,
2701 1 : request_lsn: Lsn(4),
2702 1 : not_modified_since: Lsn(3),
2703 1 : },
2704 1 : rel: RelTag {
2705 1 : forknum: 1,
2706 1 : spcnode: 2,
2707 1 : dbnode: 3,
2708 1 : relnode: 4,
2709 1 : },
2710 1 : blkno: 7,
2711 1 : }),
2712 1 : PagestreamFeMessage::DbSize(PagestreamDbSizeRequest {
2713 1 : hdr: PagestreamRequest {
2714 1 : reqid: 0,
2715 1 : request_lsn: Lsn(4),
2716 1 : not_modified_since: Lsn(3),
2717 1 : },
2718 1 : dbnode: 7,
2719 1 : }),
2720 1 : ];
2721 5 : for msg in messages {
2722 4 : let bytes = msg.serialize();
2723 4 : let reconstructed =
2724 4 : PagestreamFeMessage::parse(&mut bytes.reader(), PagestreamProtocolVersion::V3)
2725 4 : .unwrap();
2726 4 : assert!(msg == reconstructed);
2727 : }
2728 1 : }
2729 :
2730 : #[test]
2731 1 : fn test_tenantinfo_serde() {
2732 1 : // Test serialization/deserialization of TenantInfo
2733 1 : let original_active = TenantInfo {
2734 1 : id: TenantShardId::unsharded(TenantId::generate()),
2735 1 : state: TenantState::Active,
2736 1 : current_physical_size: Some(42),
2737 1 : attachment_status: TenantAttachmentStatus::Attached,
2738 1 : generation: 1,
2739 1 : gc_blocking: None,
2740 1 : };
2741 1 : let expected_active = json!({
2742 1 : "id": original_active.id.to_string(),
2743 1 : "state": {
2744 1 : "slug": "Active",
2745 1 : },
2746 1 : "current_physical_size": 42,
2747 1 : "attachment_status": {
2748 1 : "slug":"attached",
2749 1 : },
2750 1 : "generation" : 1
2751 1 : });
2752 1 :
2753 1 : let original_broken = TenantInfo {
2754 1 : id: TenantShardId::unsharded(TenantId::generate()),
2755 1 : state: TenantState::Broken {
2756 1 : reason: "reason".into(),
2757 1 : backtrace: "backtrace info".into(),
2758 1 : },
2759 1 : current_physical_size: Some(42),
2760 1 : attachment_status: TenantAttachmentStatus::Attached,
2761 1 : generation: 1,
2762 1 : gc_blocking: None,
2763 1 : };
2764 1 : let expected_broken = json!({
2765 1 : "id": original_broken.id.to_string(),
2766 1 : "state": {
2767 1 : "slug": "Broken",
2768 1 : "data": {
2769 1 : "backtrace": "backtrace info",
2770 1 : "reason": "reason",
2771 1 : }
2772 1 : },
2773 1 : "current_physical_size": 42,
2774 1 : "attachment_status": {
2775 1 : "slug":"attached",
2776 1 : },
2777 1 : "generation" : 1
2778 1 : });
2779 1 :
2780 1 : assert_eq!(
2781 1 : serde_json::to_value(&original_active).unwrap(),
2782 1 : expected_active
2783 1 : );
2784 :
2785 1 : assert_eq!(
2786 1 : serde_json::to_value(&original_broken).unwrap(),
2787 1 : expected_broken
2788 1 : );
2789 1 : assert!(format!("{:?}", &original_broken.state).contains("reason"));
2790 1 : assert!(format!("{:?}", &original_broken.state).contains("backtrace info"));
2791 1 : }
2792 :
2793 : #[test]
2794 1 : fn test_reject_unknown_field() {
2795 1 : let id = TenantId::generate();
2796 1 : let config_request = json!({
2797 1 : "tenant_id": id.to_string(),
2798 1 : "unknown_field": "unknown_value".to_string(),
2799 1 : });
2800 1 : let err = serde_json::from_value::<TenantConfigRequest>(config_request).unwrap_err();
2801 1 : assert!(
2802 1 : err.to_string().contains("unknown field `unknown_field`"),
2803 0 : "expect unknown field `unknown_field` error, got: {}",
2804 : err
2805 : );
2806 1 : }
2807 :
2808 : #[test]
2809 1 : fn tenantstatus_activating_serde() {
2810 1 : let states = [TenantState::Activating(ActivatingFrom::Attaching)];
2811 1 : let expected = "[{\"slug\":\"Activating\",\"data\":\"Attaching\"}]";
2812 1 :
2813 1 : let actual = serde_json::to_string(&states).unwrap();
2814 1 :
2815 1 : assert_eq!(actual, expected);
2816 :
2817 1 : let parsed = serde_json::from_str::<Vec<TenantState>>(&actual).unwrap();
2818 1 :
2819 1 : assert_eq!(states.as_slice(), &parsed);
2820 1 : }
2821 :
2822 : #[test]
2823 1 : fn tenantstatus_activating_strum() {
2824 1 : // tests added, because we use these for metrics
2825 1 : let examples = [
2826 1 : (line!(), TenantState::Attaching, "Attaching"),
2827 1 : (
2828 1 : line!(),
2829 1 : TenantState::Activating(ActivatingFrom::Attaching),
2830 1 : "Activating",
2831 1 : ),
2832 1 : (line!(), TenantState::Active, "Active"),
2833 1 : (
2834 1 : line!(),
2835 1 : TenantState::Stopping { progress: None },
2836 1 : "Stopping",
2837 1 : ),
2838 1 : (
2839 1 : line!(),
2840 1 : TenantState::Stopping {
2841 1 : progress: Some(completion::Barrier::default()),
2842 1 : },
2843 1 : "Stopping",
2844 1 : ),
2845 1 : (
2846 1 : line!(),
2847 1 : TenantState::Broken {
2848 1 : reason: "Example".into(),
2849 1 : backtrace: "Looooong backtrace".into(),
2850 1 : },
2851 1 : "Broken",
2852 1 : ),
2853 1 : ];
2854 :
2855 7 : for (line, rendered, expected) in examples {
2856 6 : let actual: &'static str = rendered.into();
2857 6 : assert_eq!(actual, expected, "example on {line}");
2858 : }
2859 1 : }
2860 :
2861 : #[test]
2862 1 : fn test_image_compression_algorithm_parsing() {
2863 : use ImageCompressionAlgorithm::*;
2864 1 : let cases = [
2865 1 : ("disabled", Disabled),
2866 1 : ("zstd", Zstd { level: None }),
2867 1 : ("zstd(18)", Zstd { level: Some(18) }),
2868 1 : ("zstd(-3)", Zstd { level: Some(-3) }),
2869 1 : ];
2870 :
2871 5 : for (display, expected) in cases {
2872 4 : assert_eq!(
2873 4 : ImageCompressionAlgorithm::from_str(display).unwrap(),
2874 : expected,
2875 0 : "parsing works"
2876 : );
2877 4 : assert_eq!(format!("{expected}"), display, "Display FromStr roundtrip");
2878 :
2879 4 : let ser = serde_json::to_string(&expected).expect("serialization");
2880 4 : assert_eq!(
2881 4 : serde_json::from_str::<ImageCompressionAlgorithm>(&ser).unwrap(),
2882 : expected,
2883 0 : "serde roundtrip"
2884 : );
2885 :
2886 4 : assert_eq!(
2887 4 : serde_json::Value::String(display.to_string()),
2888 4 : serde_json::to_value(expected).unwrap(),
2889 0 : "Display is the serde serialization"
2890 : );
2891 : }
2892 1 : }
2893 :
2894 : #[test]
2895 1 : fn test_tenant_config_patch_request_serde() {
2896 1 : let patch_request = TenantConfigPatchRequest {
2897 1 : tenant_id: TenantId::from_str("17c6d121946a61e5ab0fe5a2fd4d8215").unwrap(),
2898 1 : config: TenantConfigPatch {
2899 1 : checkpoint_distance: FieldPatch::Upsert(42),
2900 1 : gc_horizon: FieldPatch::Remove,
2901 1 : compaction_threshold: FieldPatch::Noop,
2902 1 : ..TenantConfigPatch::default()
2903 1 : },
2904 1 : };
2905 1 :
2906 1 : let json = serde_json::to_string(&patch_request).unwrap();
2907 1 :
2908 1 : let expected = r#"{"tenant_id":"17c6d121946a61e5ab0fe5a2fd4d8215","checkpoint_distance":42,"gc_horizon":null}"#;
2909 1 : assert_eq!(json, expected);
2910 :
2911 1 : let decoded: TenantConfigPatchRequest = serde_json::from_str(&json).unwrap();
2912 1 : assert_eq!(decoded.tenant_id, patch_request.tenant_id);
2913 1 : assert_eq!(decoded.config, patch_request.config);
2914 :
2915 : // Now apply the patch to a config to demonstrate semantics
2916 :
2917 1 : let base = TenantConfig {
2918 1 : checkpoint_distance: Some(28),
2919 1 : gc_horizon: Some(100),
2920 1 : compaction_target_size: Some(1024),
2921 1 : ..Default::default()
2922 1 : };
2923 1 :
2924 1 : let expected = TenantConfig {
2925 1 : checkpoint_distance: Some(42),
2926 1 : gc_horizon: None,
2927 1 : ..base.clone()
2928 1 : };
2929 1 :
2930 1 : let patched = base.apply_patch(decoded.config).unwrap();
2931 1 :
2932 1 : assert_eq!(patched, expected);
2933 1 : }
2934 : }
|
