Line data Source code
1 : //!
2 : //! Management HTTP API
3 : //!
4 : use std::cmp::Reverse;
5 : use std::collections::BinaryHeap;
6 : use std::collections::HashMap;
7 : use std::str::FromStr;
8 : use std::sync::Arc;
9 : use std::time::Duration;
10 :
11 : use anyhow::{anyhow, Context, Result};
12 : use enumset::EnumSet;
13 : use futures::StreamExt;
14 : use futures::TryFutureExt;
15 : use humantime::format_rfc3339;
16 : use hyper::header;
17 : use hyper::StatusCode;
18 : use hyper::{Body, Request, Response, Uri};
19 : use metrics::launch_timestamp::LaunchTimestamp;
20 : use pageserver_api::models::virtual_file::IoMode;
21 : use pageserver_api::models::DownloadRemoteLayersTaskSpawnRequest;
22 : use pageserver_api::models::IngestAuxFilesRequest;
23 : use pageserver_api::models::ListAuxFilesRequest;
24 : use pageserver_api::models::LocationConfig;
25 : use pageserver_api::models::LocationConfigListResponse;
26 : use pageserver_api::models::LocationConfigMode;
27 : use pageserver_api::models::LsnLease;
28 : use pageserver_api::models::LsnLeaseRequest;
29 : use pageserver_api::models::OffloadedTimelineInfo;
30 : use pageserver_api::models::ShardParameters;
31 : use pageserver_api::models::TenantDetails;
32 : use pageserver_api::models::TenantLocationConfigRequest;
33 : use pageserver_api::models::TenantLocationConfigResponse;
34 : use pageserver_api::models::TenantScanRemoteStorageResponse;
35 : use pageserver_api::models::TenantScanRemoteStorageShard;
36 : use pageserver_api::models::TenantShardLocation;
37 : use pageserver_api::models::TenantShardSplitRequest;
38 : use pageserver_api::models::TenantShardSplitResponse;
39 : use pageserver_api::models::TenantSorting;
40 : use pageserver_api::models::TenantState;
41 : use pageserver_api::models::TimelineArchivalConfigRequest;
42 : use pageserver_api::models::TimelineCreateRequestMode;
43 : use pageserver_api::models::TimelinesInfoAndOffloaded;
44 : use pageserver_api::models::TopTenantShardItem;
45 : use pageserver_api::models::TopTenantShardsRequest;
46 : use pageserver_api::models::TopTenantShardsResponse;
47 : use pageserver_api::shard::ShardCount;
48 : use pageserver_api::shard::TenantShardId;
49 : use remote_storage::DownloadError;
50 : use remote_storage::GenericRemoteStorage;
51 : use remote_storage::TimeTravelError;
52 : use tenant_size_model::{svg::SvgBranchKind, SizeResult, StorageModel};
53 : use tokio_util::io::StreamReader;
54 : use tokio_util::sync::CancellationToken;
55 : use tracing::*;
56 : use utils::auth::JwtAuth;
57 : use utils::failpoint_support::failpoints_handler;
58 : use utils::http::endpoint::prometheus_metrics_handler;
59 : use utils::http::endpoint::request_span;
60 : use utils::http::request::must_parse_query_param;
61 : use utils::http::request::{get_request_param, must_get_query_param, parse_query_param};
62 :
63 : use crate::config::PageServerConf;
64 : use crate::context::{DownloadBehavior, RequestContext};
65 : use crate::deletion_queue::DeletionQueueClient;
66 : use crate::pgdatadir_mapping::LsnForTimestamp;
67 : use crate::task_mgr::TaskKind;
68 : use crate::tenant::config::{LocationConf, TenantConfOpt};
69 : use crate::tenant::mgr::GetActiveTenantError;
70 : use crate::tenant::mgr::{
71 : GetTenantError, TenantManager, TenantMapError, TenantMapInsertError, TenantSlotError,
72 : TenantSlotUpsertError, TenantStateError,
73 : };
74 : use crate::tenant::mgr::{TenantSlot, UpsertLocationError};
75 : use crate::tenant::remote_timeline_client;
76 : use crate::tenant::remote_timeline_client::download_index_part;
77 : use crate::tenant::remote_timeline_client::list_remote_tenant_shards;
78 : use crate::tenant::remote_timeline_client::list_remote_timelines;
79 : use crate::tenant::secondary::SecondaryController;
80 : use crate::tenant::size::ModelInputs;
81 : use crate::tenant::storage_layer::LayerAccessStatsReset;
82 : use crate::tenant::storage_layer::LayerName;
83 : use crate::tenant::timeline::offload::offload_timeline;
84 : use crate::tenant::timeline::offload::OffloadError;
85 : use crate::tenant::timeline::CompactFlags;
86 : use crate::tenant::timeline::CompactOptions;
87 : use crate::tenant::timeline::CompactRange;
88 : use crate::tenant::timeline::CompactionError;
89 : use crate::tenant::timeline::Timeline;
90 : use crate::tenant::GetTimelineError;
91 : use crate::tenant::OffloadedTimeline;
92 : use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError};
93 : use crate::DEFAULT_PG_VERSION;
94 : use crate::{disk_usage_eviction_task, tenant};
95 : use pageserver_api::models::{
96 : StatusResponse, TenantConfigRequest, TenantInfo, TimelineCreateRequest, TimelineGcRequest,
97 : TimelineInfo,
98 : };
99 : use utils::{
100 : auth::SwappableJwtAuth,
101 : generation::Generation,
102 : http::{
103 : endpoint::{self, attach_openapi_ui, auth_middleware, check_permission_with},
104 : error::{ApiError, HttpErrorBody},
105 : json::{json_request, json_request_maybe, json_response},
106 : request::parse_request_param,
107 : RequestExt, RouterBuilder,
108 : },
109 : id::{TenantId, TimelineId},
110 : lsn::Lsn,
111 : };
112 :
113 : // For APIs that require an Active tenant, how long should we block waiting for that state?
114 : // This is not functionally necessary (clients will retry), but avoids generating a lot of
115 : // failed API calls while tenants are activating.
116 : #[cfg(not(feature = "testing"))]
117 : pub(crate) const ACTIVE_TENANT_TIMEOUT: Duration = Duration::from_millis(5000);
118 :
119 : // Tests run on slow/oversubscribed nodes, and may need to wait much longer for tenants to
120 : // finish attaching, if calls to remote storage are slow.
121 : #[cfg(feature = "testing")]
122 : pub(crate) const ACTIVE_TENANT_TIMEOUT: Duration = Duration::from_millis(30000);
123 :
124 : pub struct State {
125 : conf: &'static PageServerConf,
126 : tenant_manager: Arc<TenantManager>,
127 : auth: Option<Arc<SwappableJwtAuth>>,
128 : allowlist_routes: Vec<Uri>,
129 : remote_storage: GenericRemoteStorage,
130 : broker_client: storage_broker::BrokerClientChannel,
131 : disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
132 : deletion_queue_client: DeletionQueueClient,
133 : secondary_controller: SecondaryController,
134 : latest_utilization: tokio::sync::Mutex<Option<(std::time::Instant, bytes::Bytes)>>,
135 : }
136 :
137 : impl State {
138 : #[allow(clippy::too_many_arguments)]
139 0 : pub fn new(
140 0 : conf: &'static PageServerConf,
141 0 : tenant_manager: Arc<TenantManager>,
142 0 : auth: Option<Arc<SwappableJwtAuth>>,
143 0 : remote_storage: GenericRemoteStorage,
144 0 : broker_client: storage_broker::BrokerClientChannel,
145 0 : disk_usage_eviction_state: Arc<disk_usage_eviction_task::State>,
146 0 : deletion_queue_client: DeletionQueueClient,
147 0 : secondary_controller: SecondaryController,
148 0 : ) -> anyhow::Result<Self> {
149 0 : let allowlist_routes = ["/v1/status", "/v1/doc", "/swagger.yml", "/metrics"]
150 0 : .iter()
151 0 : .map(|v| v.parse().unwrap())
152 0 : .collect::<Vec<_>>();
153 0 : Ok(Self {
154 0 : conf,
155 0 : tenant_manager,
156 0 : auth,
157 0 : allowlist_routes,
158 0 : remote_storage,
159 0 : broker_client,
160 0 : disk_usage_eviction_state,
161 0 : deletion_queue_client,
162 0 : secondary_controller,
163 0 : latest_utilization: Default::default(),
164 0 : })
165 0 : }
166 : }
167 :
168 : #[inline(always)]
169 0 : fn get_state(request: &Request<Body>) -> &State {
170 0 : request
171 0 : .data::<Arc<State>>()
172 0 : .expect("unknown state type")
173 0 : .as_ref()
174 0 : }
175 :
176 : #[inline(always)]
177 0 : fn get_config(request: &Request<Body>) -> &'static PageServerConf {
178 0 : get_state(request).conf
179 0 : }
180 :
181 : /// Check that the requester is authorized to operate on given tenant
182 0 : fn check_permission(request: &Request<Body>, tenant_id: Option<TenantId>) -> Result<(), ApiError> {
183 0 : check_permission_with(request, |claims| {
184 0 : crate::auth::check_permission(claims, tenant_id)
185 0 : })
186 0 : }
187 :
188 : impl From<PageReconstructError> for ApiError {
189 0 : fn from(pre: PageReconstructError) -> ApiError {
190 0 : match pre {
191 0 : PageReconstructError::Other(other) => ApiError::InternalServerError(other),
192 0 : PageReconstructError::MissingKey(e) => ApiError::InternalServerError(e.into()),
193 0 : PageReconstructError::Cancelled => ApiError::Cancelled,
194 0 : PageReconstructError::AncestorLsnTimeout(e) => ApiError::Timeout(format!("{e}").into()),
195 0 : PageReconstructError::WalRedo(pre) => ApiError::InternalServerError(pre),
196 : }
197 0 : }
198 : }
199 :
200 : impl From<TenantMapInsertError> for ApiError {
201 0 : fn from(tmie: TenantMapInsertError) -> ApiError {
202 0 : match tmie {
203 0 : TenantMapInsertError::SlotError(e) => e.into(),
204 0 : TenantMapInsertError::SlotUpsertError(e) => e.into(),
205 0 : TenantMapInsertError::Other(e) => ApiError::InternalServerError(e),
206 : }
207 0 : }
208 : }
209 :
210 : impl From<TenantSlotError> for ApiError {
211 0 : fn from(e: TenantSlotError) -> ApiError {
212 : use TenantSlotError::*;
213 0 : match e {
214 0 : NotFound(tenant_id) => {
215 0 : ApiError::NotFound(anyhow::anyhow!("NotFound: tenant {tenant_id}").into())
216 : }
217 : InProgress => {
218 0 : ApiError::ResourceUnavailable("Tenant is being modified concurrently".into())
219 : }
220 0 : MapState(e) => e.into(),
221 : }
222 0 : }
223 : }
224 :
225 : impl From<TenantSlotUpsertError> for ApiError {
226 0 : fn from(e: TenantSlotUpsertError) -> ApiError {
227 : use TenantSlotUpsertError::*;
228 0 : match e {
229 0 : InternalError(e) => ApiError::InternalServerError(anyhow::anyhow!("{e}")),
230 0 : MapState(e) => e.into(),
231 0 : ShuttingDown(_) => ApiError::ShuttingDown,
232 : }
233 0 : }
234 : }
235 :
236 : impl From<UpsertLocationError> for ApiError {
237 0 : fn from(e: UpsertLocationError) -> ApiError {
238 : use UpsertLocationError::*;
239 0 : match e {
240 0 : BadRequest(e) => ApiError::BadRequest(e),
241 0 : Unavailable(_) => ApiError::ShuttingDown,
242 0 : e @ InProgress => ApiError::Conflict(format!("{e}")),
243 0 : Flush(e) | InternalError(e) => ApiError::InternalServerError(e),
244 : }
245 0 : }
246 : }
247 :
248 : impl From<TenantMapError> for ApiError {
249 0 : fn from(e: TenantMapError) -> ApiError {
250 : use TenantMapError::*;
251 0 : match e {
252 : StillInitializing | ShuttingDown => {
253 0 : ApiError::ResourceUnavailable(format!("{e}").into())
254 0 : }
255 0 : }
256 0 : }
257 : }
258 :
259 : impl From<TenantStateError> for ApiError {
260 0 : fn from(tse: TenantStateError) -> ApiError {
261 0 : match tse {
262 : TenantStateError::IsStopping(_) => {
263 0 : ApiError::ResourceUnavailable("Tenant is stopping".into())
264 : }
265 0 : TenantStateError::SlotError(e) => e.into(),
266 0 : TenantStateError::SlotUpsertError(e) => e.into(),
267 0 : TenantStateError::Other(e) => ApiError::InternalServerError(anyhow!(e)),
268 : }
269 0 : }
270 : }
271 :
272 : impl From<GetTenantError> for ApiError {
273 0 : fn from(tse: GetTenantError) -> ApiError {
274 0 : match tse {
275 0 : GetTenantError::NotFound(tid) => ApiError::NotFound(anyhow!("tenant {}", tid).into()),
276 : GetTenantError::NotActive(_) => {
277 : // Why is this not `ApiError::NotFound`?
278 : // Because we must be careful to never return 404 for a tenant if it does
279 : // in fact exist locally. If we did, the caller could draw the conclusion
280 : // that it can attach the tenant to another PS and we'd be in split-brain.
281 0 : ApiError::ResourceUnavailable("Tenant not yet active".into())
282 : }
283 0 : GetTenantError::MapState(e) => ApiError::ResourceUnavailable(format!("{e}").into()),
284 : }
285 0 : }
286 : }
287 :
288 : impl From<GetTimelineError> for ApiError {
289 0 : fn from(gte: GetTimelineError) -> Self {
290 0 : // Rationale: tenant is activated only after eligble timelines activate
291 0 : ApiError::NotFound(gte.into())
292 0 : }
293 : }
294 :
295 : impl From<GetActiveTenantError> for ApiError {
296 0 : fn from(e: GetActiveTenantError) -> ApiError {
297 0 : match e {
298 0 : GetActiveTenantError::Broken(reason) => {
299 0 : ApiError::InternalServerError(anyhow!("tenant is broken: {}", reason))
300 : }
301 : GetActiveTenantError::WillNotBecomeActive(TenantState::Stopping { .. }) => {
302 0 : ApiError::ShuttingDown
303 : }
304 0 : GetActiveTenantError::WillNotBecomeActive(_) => ApiError::Conflict(format!("{}", e)),
305 0 : GetActiveTenantError::Cancelled => ApiError::ShuttingDown,
306 0 : GetActiveTenantError::NotFound(gte) => gte.into(),
307 : GetActiveTenantError::WaitForActiveTimeout { .. } => {
308 0 : ApiError::ResourceUnavailable(format!("{}", e).into())
309 : }
310 : GetActiveTenantError::SwitchedTenant => {
311 : // in our HTTP handlers, this error doesn't happen
312 : // TODO: separate error types
313 0 : ApiError::ResourceUnavailable("switched tenant".into())
314 : }
315 : }
316 0 : }
317 : }
318 :
319 : impl From<crate::tenant::DeleteTimelineError> for ApiError {
320 0 : fn from(value: crate::tenant::DeleteTimelineError) -> Self {
321 : use crate::tenant::DeleteTimelineError::*;
322 0 : match value {
323 0 : NotFound => ApiError::NotFound(anyhow::anyhow!("timeline not found").into()),
324 0 : HasChildren(children) => ApiError::PreconditionFailed(
325 0 : format!("Cannot delete timeline which has child timelines: {children:?}")
326 0 : .into_boxed_str(),
327 0 : ),
328 0 : a @ AlreadyInProgress(_) => ApiError::Conflict(a.to_string()),
329 0 : Cancelled => ApiError::ResourceUnavailable("shutting down".into()),
330 0 : Other(e) => ApiError::InternalServerError(e),
331 : }
332 0 : }
333 : }
334 :
335 : impl From<crate::tenant::TimelineArchivalError> for ApiError {
336 0 : fn from(value: crate::tenant::TimelineArchivalError) -> Self {
337 : use crate::tenant::TimelineArchivalError::*;
338 0 : match value {
339 0 : NotFound => ApiError::NotFound(anyhow::anyhow!("timeline not found").into()),
340 0 : Timeout => ApiError::Timeout("hit pageserver internal timeout".into()),
341 0 : Cancelled => ApiError::ShuttingDown,
342 0 : e @ HasArchivedParent(_) => {
343 0 : ApiError::PreconditionFailed(e.to_string().into_boxed_str())
344 : }
345 0 : HasUnarchivedChildren(children) => ApiError::PreconditionFailed(
346 0 : format!(
347 0 : "Cannot archive timeline which has non-archived child timelines: {children:?}"
348 0 : )
349 0 : .into_boxed_str(),
350 0 : ),
351 0 : a @ AlreadyInProgress => ApiError::Conflict(a.to_string()),
352 0 : Other(e) => ApiError::InternalServerError(e),
353 : }
354 0 : }
355 : }
356 :
357 : impl From<crate::tenant::mgr::DeleteTimelineError> for ApiError {
358 0 : fn from(value: crate::tenant::mgr::DeleteTimelineError) -> Self {
359 : use crate::tenant::mgr::DeleteTimelineError::*;
360 0 : match value {
361 : // Report Precondition failed so client can distinguish between
362 : // "tenant is missing" case from "timeline is missing"
363 0 : Tenant(GetTenantError::NotFound(..)) => ApiError::PreconditionFailed(
364 0 : "Requested tenant is missing".to_owned().into_boxed_str(),
365 0 : ),
366 0 : Tenant(t) => ApiError::from(t),
367 0 : Timeline(t) => ApiError::from(t),
368 : }
369 0 : }
370 : }
371 :
372 : impl From<crate::tenant::mgr::DeleteTenantError> for ApiError {
373 0 : fn from(value: crate::tenant::mgr::DeleteTenantError) -> Self {
374 : use crate::tenant::mgr::DeleteTenantError::*;
375 0 : match value {
376 0 : SlotError(e) => e.into(),
377 0 : Other(o) => ApiError::InternalServerError(o),
378 0 : Cancelled => ApiError::ShuttingDown,
379 : }
380 0 : }
381 : }
382 :
383 : // Helper function to construct a TimelineInfo struct for a timeline
384 0 : async fn build_timeline_info(
385 0 : timeline: &Arc<Timeline>,
386 0 : include_non_incremental_logical_size: bool,
387 0 : force_await_initial_logical_size: bool,
388 0 : ctx: &RequestContext,
389 0 : ) -> anyhow::Result<TimelineInfo> {
390 0 : crate::tenant::debug_assert_current_span_has_tenant_and_timeline_id();
391 0 :
392 0 : if force_await_initial_logical_size {
393 0 : timeline.clone().await_initial_logical_size().await
394 0 : }
395 :
396 0 : let mut info = build_timeline_info_common(
397 0 : timeline,
398 0 : ctx,
399 0 : tenant::timeline::GetLogicalSizePriority::Background,
400 0 : )
401 0 : .await?;
402 0 : if include_non_incremental_logical_size {
403 : // XXX we should be using spawn_ondemand_logical_size_calculation here.
404 : // Otherwise, if someone deletes the timeline / detaches the tenant while
405 : // we're executing this function, we will outlive the timeline on-disk state.
406 : info.current_logical_size_non_incremental = Some(
407 0 : timeline
408 0 : .get_current_logical_size_non_incremental(info.last_record_lsn, ctx)
409 0 : .await?,
410 : );
411 0 : }
412 0 : Ok(info)
413 0 : }
414 :
415 0 : async fn build_timeline_info_common(
416 0 : timeline: &Arc<Timeline>,
417 0 : ctx: &RequestContext,
418 0 : logical_size_task_priority: tenant::timeline::GetLogicalSizePriority,
419 0 : ) -> anyhow::Result<TimelineInfo> {
420 0 : crate::tenant::debug_assert_current_span_has_tenant_and_timeline_id();
421 0 : let initdb_lsn = timeline.initdb_lsn;
422 0 : let last_record_lsn = timeline.get_last_record_lsn();
423 0 : let (wal_source_connstr, last_received_msg_lsn, last_received_msg_ts) = {
424 0 : let guard = timeline.last_received_wal.lock().unwrap();
425 0 : if let Some(info) = guard.as_ref() {
426 0 : (
427 0 : Some(format!("{}", info.wal_source_connconf)), // Password is hidden, but it's for statistics only.
428 0 : Some(info.last_received_msg_lsn),
429 0 : Some(info.last_received_msg_ts),
430 0 : )
431 : } else {
432 0 : (None, None, None)
433 : }
434 : };
435 :
436 0 : let ancestor_timeline_id = timeline.get_ancestor_timeline_id();
437 0 : let ancestor_lsn = match timeline.get_ancestor_lsn() {
438 0 : Lsn(0) => None,
439 0 : lsn @ Lsn(_) => Some(lsn),
440 : };
441 0 : let current_logical_size = timeline.get_current_logical_size(logical_size_task_priority, ctx);
442 0 : let current_physical_size = Some(timeline.layer_size_sum().await);
443 0 : let state = timeline.current_state();
444 0 : // Report is_archived = false if the timeline is still loading
445 0 : let is_archived = timeline.is_archived().unwrap_or(false);
446 0 : let remote_consistent_lsn_projected = timeline
447 0 : .get_remote_consistent_lsn_projected()
448 0 : .unwrap_or(Lsn(0));
449 0 : let remote_consistent_lsn_visible = timeline
450 0 : .get_remote_consistent_lsn_visible()
451 0 : .unwrap_or(Lsn(0));
452 0 :
453 0 : let walreceiver_status = timeline.walreceiver_status();
454 0 :
455 0 : let (pitr_history_size, within_ancestor_pitr) = timeline.get_pitr_history_stats();
456 :
457 0 : let info = TimelineInfo {
458 0 : tenant_id: timeline.tenant_shard_id,
459 0 : timeline_id: timeline.timeline_id,
460 0 : ancestor_timeline_id,
461 0 : ancestor_lsn,
462 0 : disk_consistent_lsn: timeline.get_disk_consistent_lsn(),
463 0 : remote_consistent_lsn: remote_consistent_lsn_projected,
464 0 : remote_consistent_lsn_visible,
465 0 : initdb_lsn,
466 0 : last_record_lsn,
467 0 : prev_record_lsn: Some(timeline.get_prev_record_lsn()),
468 0 : latest_gc_cutoff_lsn: *timeline.get_latest_gc_cutoff_lsn(),
469 0 : current_logical_size: current_logical_size.size_dont_care_about_accuracy(),
470 0 : current_logical_size_is_accurate: match current_logical_size.accuracy() {
471 0 : tenant::timeline::logical_size::Accuracy::Approximate => false,
472 0 : tenant::timeline::logical_size::Accuracy::Exact => true,
473 : },
474 0 : directory_entries_counts: timeline.get_directory_metrics().to_vec(),
475 0 : current_physical_size,
476 0 : current_logical_size_non_incremental: None,
477 0 : pitr_history_size,
478 0 : within_ancestor_pitr,
479 0 : timeline_dir_layer_file_size_sum: None,
480 0 : wal_source_connstr,
481 0 : last_received_msg_lsn,
482 0 : last_received_msg_ts,
483 0 : pg_version: timeline.pg_version,
484 0 :
485 0 : state,
486 0 : is_archived: Some(is_archived),
487 0 :
488 0 : walreceiver_status,
489 0 : };
490 0 : Ok(info)
491 0 : }
492 :
493 0 : fn build_timeline_offloaded_info(offloaded: &Arc<OffloadedTimeline>) -> OffloadedTimelineInfo {
494 0 : let &OffloadedTimeline {
495 0 : tenant_shard_id,
496 0 : timeline_id,
497 0 : ancestor_retain_lsn,
498 0 : ancestor_timeline_id,
499 0 : archived_at,
500 0 : ..
501 0 : } = offloaded.as_ref();
502 0 : OffloadedTimelineInfo {
503 0 : tenant_id: tenant_shard_id,
504 0 : timeline_id,
505 0 : ancestor_retain_lsn,
506 0 : ancestor_timeline_id,
507 0 : archived_at: archived_at.and_utc(),
508 0 : }
509 0 : }
510 :
511 : // healthcheck handler
512 0 : async fn status_handler(
513 0 : request: Request<Body>,
514 0 : _cancel: CancellationToken,
515 0 : ) -> Result<Response<Body>, ApiError> {
516 0 : check_permission(&request, None)?;
517 0 : let config = get_config(&request);
518 0 : json_response(StatusCode::OK, StatusResponse { id: config.id })
519 0 : }
520 :
521 0 : async fn reload_auth_validation_keys_handler(
522 0 : request: Request<Body>,
523 0 : _cancel: CancellationToken,
524 0 : ) -> Result<Response<Body>, ApiError> {
525 0 : check_permission(&request, None)?;
526 0 : let config = get_config(&request);
527 0 : let state = get_state(&request);
528 0 : let Some(shared_auth) = &state.auth else {
529 0 : return json_response(StatusCode::BAD_REQUEST, ());
530 : };
531 : // unwrap is ok because check is performed when creating config, so path is set and exists
532 0 : let key_path = config.auth_validation_public_key_path.as_ref().unwrap();
533 0 : info!("Reloading public key(s) for verifying JWT tokens from {key_path:?}");
534 :
535 0 : match JwtAuth::from_key_path(key_path) {
536 0 : Ok(new_auth) => {
537 0 : shared_auth.swap(new_auth);
538 0 : json_response(StatusCode::OK, ())
539 : }
540 0 : Err(e) => {
541 0 : let err_msg = "Error reloading public keys";
542 0 : warn!("Error reloading public keys from {key_path:?}: {e:}");
543 0 : json_response(
544 0 : StatusCode::INTERNAL_SERVER_ERROR,
545 0 : HttpErrorBody::from_msg(err_msg.to_string()),
546 0 : )
547 : }
548 : }
549 0 : }
550 :
551 0 : async fn timeline_create_handler(
552 0 : mut request: Request<Body>,
553 0 : _cancel: CancellationToken,
554 0 : ) -> Result<Response<Body>, ApiError> {
555 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
556 0 : let request_data: TimelineCreateRequest = json_request(&mut request).await?;
557 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
558 :
559 0 : let new_timeline_id = request_data.new_timeline_id;
560 : // fill in the default pg_version if not provided & convert request into domain model
561 0 : let params: tenant::CreateTimelineParams = match request_data.mode {
562 : TimelineCreateRequestMode::Bootstrap {
563 0 : existing_initdb_timeline_id,
564 0 : pg_version,
565 0 : } => tenant::CreateTimelineParams::Bootstrap(tenant::CreateTimelineParamsBootstrap {
566 0 : new_timeline_id,
567 0 : existing_initdb_timeline_id,
568 0 : pg_version: pg_version.unwrap_or(DEFAULT_PG_VERSION),
569 0 : }),
570 : TimelineCreateRequestMode::Branch {
571 0 : ancestor_timeline_id,
572 0 : ancestor_start_lsn,
573 0 : pg_version: _,
574 0 : } => tenant::CreateTimelineParams::Branch(tenant::CreateTimelineParamsBranch {
575 0 : new_timeline_id,
576 0 : ancestor_timeline_id,
577 0 : ancestor_start_lsn,
578 0 : }),
579 : };
580 :
581 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Error);
582 0 :
583 0 : let state = get_state(&request);
584 :
585 0 : async {
586 0 : let tenant = state
587 0 : .tenant_manager
588 0 : .get_attached_tenant_shard(tenant_shard_id)?;
589 :
590 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
591 :
592 : // earlier versions of the code had pg_version and ancestor_lsn in the span
593 : // => continue to provide that information, but, through a log message that doesn't require us to destructure
594 0 : tracing::info!(?params, "creating timeline");
595 :
596 0 : match tenant
597 0 : .create_timeline(params, state.broker_client.clone(), &ctx)
598 0 : .await
599 : {
600 0 : Ok(new_timeline) => {
601 : // Created. Construct a TimelineInfo for it.
602 0 : let timeline_info = build_timeline_info_common(
603 0 : &new_timeline,
604 0 : &ctx,
605 0 : tenant::timeline::GetLogicalSizePriority::User,
606 0 : )
607 0 : .await
608 0 : .map_err(ApiError::InternalServerError)?;
609 0 : json_response(StatusCode::CREATED, timeline_info)
610 : }
611 0 : Err(_) if tenant.cancel.is_cancelled() => {
612 0 : // In case we get some ugly error type during shutdown, cast it into a clean 503.
613 0 : json_response(
614 0 : StatusCode::SERVICE_UNAVAILABLE,
615 0 : HttpErrorBody::from_msg("Tenant shutting down".to_string()),
616 0 : )
617 : }
618 0 : Err(e @ tenant::CreateTimelineError::Conflict) => {
619 0 : json_response(StatusCode::CONFLICT, HttpErrorBody::from_msg(e.to_string()))
620 : }
621 0 : Err(e @ tenant::CreateTimelineError::AlreadyCreating) => json_response(
622 0 : StatusCode::TOO_MANY_REQUESTS,
623 0 : HttpErrorBody::from_msg(e.to_string()),
624 0 : ),
625 0 : Err(tenant::CreateTimelineError::AncestorLsn(err)) => json_response(
626 0 : StatusCode::NOT_ACCEPTABLE,
627 0 : HttpErrorBody::from_msg(format!("{err:#}")),
628 0 : ),
629 0 : Err(e @ tenant::CreateTimelineError::AncestorNotActive) => json_response(
630 0 : StatusCode::SERVICE_UNAVAILABLE,
631 0 : HttpErrorBody::from_msg(e.to_string()),
632 0 : ),
633 0 : Err(e @ tenant::CreateTimelineError::AncestorArchived) => json_response(
634 0 : StatusCode::NOT_ACCEPTABLE,
635 0 : HttpErrorBody::from_msg(e.to_string()),
636 0 : ),
637 0 : Err(tenant::CreateTimelineError::ShuttingDown) => json_response(
638 0 : StatusCode::SERVICE_UNAVAILABLE,
639 0 : HttpErrorBody::from_msg("tenant shutting down".to_string()),
640 0 : ),
641 0 : Err(tenant::CreateTimelineError::Other(err)) => Err(ApiError::InternalServerError(err)),
642 : }
643 0 : }
644 0 : .instrument(info_span!("timeline_create",
645 : tenant_id = %tenant_shard_id.tenant_id,
646 0 : shard_id = %tenant_shard_id.shard_slug(),
647 : timeline_id = %new_timeline_id,
648 : ))
649 0 : .await
650 0 : }
651 :
652 0 : async fn timeline_list_handler(
653 0 : request: Request<Body>,
654 0 : _cancel: CancellationToken,
655 0 : ) -> Result<Response<Body>, ApiError> {
656 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
657 0 : let include_non_incremental_logical_size: Option<bool> =
658 0 : parse_query_param(&request, "include-non-incremental-logical-size")?;
659 0 : let force_await_initial_logical_size: Option<bool> =
660 0 : parse_query_param(&request, "force-await-initial-logical-size")?;
661 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
662 :
663 0 : let state = get_state(&request);
664 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
665 :
666 0 : let response_data = async {
667 0 : let tenant = state
668 0 : .tenant_manager
669 0 : .get_attached_tenant_shard(tenant_shard_id)?;
670 :
671 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
672 :
673 0 : let timelines = tenant.list_timelines();
674 0 :
675 0 : let mut response_data = Vec::with_capacity(timelines.len());
676 0 : for timeline in timelines {
677 0 : let timeline_info = build_timeline_info(
678 0 : &timeline,
679 0 : include_non_incremental_logical_size.unwrap_or(false),
680 0 : force_await_initial_logical_size.unwrap_or(false),
681 0 : &ctx,
682 0 : )
683 0 : .instrument(info_span!("build_timeline_info", timeline_id = %timeline.timeline_id))
684 0 : .await
685 0 : .context("Failed to build timeline info")
686 0 : .map_err(ApiError::InternalServerError)?;
687 :
688 0 : response_data.push(timeline_info);
689 : }
690 0 : Ok::<Vec<TimelineInfo>, ApiError>(response_data)
691 0 : }
692 0 : .instrument(info_span!("timeline_list",
693 : tenant_id = %tenant_shard_id.tenant_id,
694 0 : shard_id = %tenant_shard_id.shard_slug()))
695 0 : .await?;
696 :
697 0 : json_response(StatusCode::OK, response_data)
698 0 : }
699 :
700 0 : async fn timeline_and_offloaded_list_handler(
701 0 : request: Request<Body>,
702 0 : _cancel: CancellationToken,
703 0 : ) -> Result<Response<Body>, ApiError> {
704 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
705 0 : let include_non_incremental_logical_size: Option<bool> =
706 0 : parse_query_param(&request, "include-non-incremental-logical-size")?;
707 0 : let force_await_initial_logical_size: Option<bool> =
708 0 : parse_query_param(&request, "force-await-initial-logical-size")?;
709 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
710 :
711 0 : let state = get_state(&request);
712 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
713 :
714 0 : let response_data = async {
715 0 : let tenant = state
716 0 : .tenant_manager
717 0 : .get_attached_tenant_shard(tenant_shard_id)?;
718 :
719 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
720 :
721 0 : let (timelines, offloadeds) = tenant.list_timelines_and_offloaded();
722 0 :
723 0 : let mut timeline_infos = Vec::with_capacity(timelines.len());
724 0 : for timeline in timelines {
725 0 : let timeline_info = build_timeline_info(
726 0 : &timeline,
727 0 : include_non_incremental_logical_size.unwrap_or(false),
728 0 : force_await_initial_logical_size.unwrap_or(false),
729 0 : &ctx,
730 0 : )
731 0 : .instrument(info_span!("build_timeline_info", timeline_id = %timeline.timeline_id))
732 0 : .await
733 0 : .context("Failed to build timeline info")
734 0 : .map_err(ApiError::InternalServerError)?;
735 :
736 0 : timeline_infos.push(timeline_info);
737 : }
738 0 : let offloaded_infos = offloadeds
739 0 : .into_iter()
740 0 : .map(|offloaded| build_timeline_offloaded_info(&offloaded))
741 0 : .collect::<Vec<_>>();
742 0 : let res = TimelinesInfoAndOffloaded {
743 0 : timelines: timeline_infos,
744 0 : offloaded: offloaded_infos,
745 0 : };
746 0 : Ok::<TimelinesInfoAndOffloaded, ApiError>(res)
747 0 : }
748 0 : .instrument(info_span!("timeline_and_offloaded_list",
749 : tenant_id = %tenant_shard_id.tenant_id,
750 0 : shard_id = %tenant_shard_id.shard_slug()))
751 0 : .await?;
752 :
753 0 : json_response(StatusCode::OK, response_data)
754 0 : }
755 :
756 0 : async fn timeline_preserve_initdb_handler(
757 0 : request: Request<Body>,
758 0 : _cancel: CancellationToken,
759 0 : ) -> Result<Response<Body>, ApiError> {
760 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
761 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
762 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
763 0 : let state = get_state(&request);
764 :
765 : // Part of the process for disaster recovery from safekeeper-stored WAL:
766 : // If we don't recover into a new timeline but want to keep the timeline ID,
767 : // then the initdb archive is deleted. This endpoint copies it to a different
768 : // location where timeline recreation cand find it.
769 :
770 0 : async {
771 0 : let tenant = state
772 0 : .tenant_manager
773 0 : .get_attached_tenant_shard(tenant_shard_id)?;
774 :
775 0 : let timeline = tenant.get_timeline(timeline_id, false)?;
776 :
777 0 : timeline
778 0 : .preserve_initdb_archive()
779 0 : .await
780 0 : .context("preserving initdb archive")
781 0 : .map_err(ApiError::InternalServerError)?;
782 :
783 0 : Ok::<_, ApiError>(())
784 0 : }
785 0 : .instrument(info_span!("timeline_preserve_initdb_archive",
786 : tenant_id = %tenant_shard_id.tenant_id,
787 0 : shard_id = %tenant_shard_id.shard_slug(),
788 : %timeline_id))
789 0 : .await?;
790 :
791 0 : json_response(StatusCode::OK, ())
792 0 : }
793 :
794 0 : async fn timeline_archival_config_handler(
795 0 : mut request: Request<Body>,
796 0 : _cancel: CancellationToken,
797 0 : ) -> Result<Response<Body>, ApiError> {
798 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
799 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
800 :
801 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
802 :
803 0 : let request_data: TimelineArchivalConfigRequest = json_request(&mut request).await?;
804 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
805 0 : let state = get_state(&request);
806 :
807 0 : async {
808 0 : let tenant = state
809 0 : .tenant_manager
810 0 : .get_attached_tenant_shard(tenant_shard_id)?;
811 :
812 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
813 :
814 0 : tenant
815 0 : .apply_timeline_archival_config(
816 0 : timeline_id,
817 0 : request_data.state,
818 0 : state.broker_client.clone(),
819 0 : ctx,
820 0 : )
821 0 : .await?;
822 0 : Ok::<_, ApiError>(())
823 0 : }
824 0 : .instrument(info_span!("timeline_archival_config",
825 : tenant_id = %tenant_shard_id.tenant_id,
826 0 : shard_id = %tenant_shard_id.shard_slug(),
827 : state = ?request_data.state,
828 : %timeline_id))
829 0 : .await?;
830 :
831 0 : json_response(StatusCode::OK, ())
832 0 : }
833 :
834 0 : async fn timeline_detail_handler(
835 0 : request: Request<Body>,
836 0 : _cancel: CancellationToken,
837 0 : ) -> Result<Response<Body>, ApiError> {
838 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
839 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
840 0 : let include_non_incremental_logical_size: Option<bool> =
841 0 : parse_query_param(&request, "include-non-incremental-logical-size")?;
842 0 : let force_await_initial_logical_size: Option<bool> =
843 0 : parse_query_param(&request, "force-await-initial-logical-size")?;
844 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
845 :
846 : // Logical size calculation needs downloading.
847 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
848 0 : let state = get_state(&request);
849 :
850 0 : let timeline_info = async {
851 0 : let tenant = state
852 0 : .tenant_manager
853 0 : .get_attached_tenant_shard(tenant_shard_id)?;
854 :
855 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
856 :
857 0 : let timeline = tenant.get_timeline(timeline_id, false)?;
858 :
859 0 : let timeline_info = build_timeline_info(
860 0 : &timeline,
861 0 : include_non_incremental_logical_size.unwrap_or(false),
862 0 : force_await_initial_logical_size.unwrap_or(false),
863 0 : &ctx,
864 0 : )
865 0 : .await
866 0 : .context("get local timeline info")
867 0 : .map_err(ApiError::InternalServerError)?;
868 :
869 0 : Ok::<_, ApiError>(timeline_info)
870 0 : }
871 0 : .instrument(info_span!("timeline_detail",
872 : tenant_id = %tenant_shard_id.tenant_id,
873 0 : shard_id = %tenant_shard_id.shard_slug(),
874 : %timeline_id))
875 0 : .await?;
876 :
877 0 : json_response(StatusCode::OK, timeline_info)
878 0 : }
879 :
880 0 : async fn get_lsn_by_timestamp_handler(
881 0 : request: Request<Body>,
882 0 : cancel: CancellationToken,
883 0 : ) -> Result<Response<Body>, ApiError> {
884 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
885 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
886 0 : let state = get_state(&request);
887 0 :
888 0 : if !tenant_shard_id.is_shard_zero() {
889 : // Requires SLRU contents, which are only stored on shard zero
890 0 : return Err(ApiError::BadRequest(anyhow!(
891 0 : "Size calculations are only available on shard zero"
892 0 : )));
893 0 : }
894 :
895 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
896 0 : let timestamp_raw = must_get_query_param(&request, "timestamp")?;
897 0 : let timestamp = humantime::parse_rfc3339(×tamp_raw)
898 0 : .with_context(|| format!("Invalid time: {:?}", timestamp_raw))
899 0 : .map_err(ApiError::BadRequest)?;
900 0 : let timestamp_pg = postgres_ffi::to_pg_timestamp(timestamp);
901 :
902 0 : let with_lease = parse_query_param(&request, "with_lease")?.unwrap_or(false);
903 0 :
904 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
905 :
906 0 : let timeline =
907 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
908 0 : .await?;
909 0 : let result = timeline
910 0 : .find_lsn_for_timestamp(timestamp_pg, &cancel, &ctx)
911 0 : .await?;
912 :
913 : #[derive(serde::Serialize, Debug)]
914 : struct Result {
915 : lsn: Lsn,
916 : kind: &'static str,
917 : #[serde(default)]
918 : #[serde(skip_serializing_if = "Option::is_none")]
919 : #[serde(flatten)]
920 : lease: Option<LsnLease>,
921 : }
922 0 : let (lsn, kind) = match result {
923 0 : LsnForTimestamp::Present(lsn) => (lsn, "present"),
924 0 : LsnForTimestamp::Future(lsn) => (lsn, "future"),
925 0 : LsnForTimestamp::Past(lsn) => (lsn, "past"),
926 0 : LsnForTimestamp::NoData(lsn) => (lsn, "nodata"),
927 : };
928 :
929 0 : let lease = if with_lease {
930 0 : timeline
931 0 : .init_lsn_lease(lsn, timeline.get_lsn_lease_length_for_ts(), &ctx)
932 0 : .inspect_err(|_| {
933 0 : warn!("fail to grant a lease to {}", lsn);
934 0 : })
935 0 : .ok()
936 : } else {
937 0 : None
938 : };
939 :
940 0 : let result = Result { lsn, kind, lease };
941 0 : let valid_until = result
942 0 : .lease
943 0 : .as_ref()
944 0 : .map(|l| humantime::format_rfc3339_millis(l.valid_until).to_string());
945 0 : tracing::info!(
946 : lsn=?result.lsn,
947 : kind=%result.kind,
948 : timestamp=%timestamp_raw,
949 : valid_until=?valid_until,
950 0 : "lsn_by_timestamp finished"
951 : );
952 0 : json_response(StatusCode::OK, result)
953 0 : }
954 :
955 0 : async fn get_timestamp_of_lsn_handler(
956 0 : request: Request<Body>,
957 0 : _cancel: CancellationToken,
958 0 : ) -> Result<Response<Body>, ApiError> {
959 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
960 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
961 0 : let state = get_state(&request);
962 0 :
963 0 : if !tenant_shard_id.is_shard_zero() {
964 : // Requires SLRU contents, which are only stored on shard zero
965 0 : return Err(ApiError::BadRequest(anyhow!(
966 0 : "Size calculations are only available on shard zero"
967 0 : )));
968 0 : }
969 :
970 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
971 :
972 0 : let lsn_str = must_get_query_param(&request, "lsn")?;
973 0 : let lsn = Lsn::from_str(&lsn_str)
974 0 : .with_context(|| format!("Invalid LSN: {lsn_str:?}"))
975 0 : .map_err(ApiError::BadRequest)?;
976 :
977 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
978 0 : let timeline =
979 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
980 0 : .await?;
981 0 : let result = timeline.get_timestamp_for_lsn(lsn, &ctx).await?;
982 :
983 0 : match result {
984 0 : Some(time) => {
985 0 : let time = format_rfc3339(
986 0 : postgres_ffi::try_from_pg_timestamp(time).map_err(ApiError::InternalServerError)?,
987 : )
988 0 : .to_string();
989 0 : json_response(StatusCode::OK, time)
990 : }
991 0 : None => Err(ApiError::NotFound(
992 0 : anyhow::anyhow!("Timestamp for lsn {} not found", lsn).into(),
993 0 : )),
994 : }
995 0 : }
996 :
997 0 : async fn timeline_delete_handler(
998 0 : request: Request<Body>,
999 0 : _cancel: CancellationToken,
1000 0 : ) -> Result<Response<Body>, ApiError> {
1001 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1002 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1003 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1004 :
1005 0 : let state = get_state(&request);
1006 :
1007 0 : let tenant = state
1008 0 : .tenant_manager
1009 0 : .get_attached_tenant_shard(tenant_shard_id)
1010 0 : .map_err(|e| {
1011 0 : match e {
1012 : // GetTenantError has a built-in conversion to ApiError, but in this context we don't
1013 : // want to treat missing tenants as 404, to avoid ambiguity with successful deletions.
1014 0 : GetTenantError::NotFound(_) => ApiError::PreconditionFailed(
1015 0 : "Requested tenant is missing".to_string().into_boxed_str(),
1016 0 : ),
1017 0 : e => e.into(),
1018 : }
1019 0 : })?;
1020 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
1021 0 : tenant.delete_timeline(timeline_id).instrument(info_span!("timeline_delete", tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id))
1022 0 : .await?;
1023 :
1024 0 : json_response(StatusCode::ACCEPTED, ())
1025 0 : }
1026 :
1027 0 : async fn tenant_reset_handler(
1028 0 : request: Request<Body>,
1029 0 : _cancel: CancellationToken,
1030 0 : ) -> Result<Response<Body>, ApiError> {
1031 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1032 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1033 :
1034 0 : let drop_cache: Option<bool> = parse_query_param(&request, "drop_cache")?;
1035 :
1036 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
1037 0 : let state = get_state(&request);
1038 0 : state
1039 0 : .tenant_manager
1040 0 : .reset_tenant(tenant_shard_id, drop_cache.unwrap_or(false), &ctx)
1041 0 : .await
1042 0 : .map_err(ApiError::InternalServerError)?;
1043 :
1044 0 : json_response(StatusCode::OK, ())
1045 0 : }
1046 :
1047 0 : async fn tenant_list_handler(
1048 0 : request: Request<Body>,
1049 0 : _cancel: CancellationToken,
1050 0 : ) -> Result<Response<Body>, ApiError> {
1051 0 : check_permission(&request, None)?;
1052 0 : let state = get_state(&request);
1053 :
1054 0 : let response_data = state
1055 0 : .tenant_manager
1056 0 : .list_tenants()
1057 0 : .map_err(|_| {
1058 0 : ApiError::ResourceUnavailable("Tenant map is initializing or shutting down".into())
1059 0 : })?
1060 0 : .iter()
1061 0 : .map(|(id, state, gen)| TenantInfo {
1062 0 : id: *id,
1063 0 : state: state.clone(),
1064 0 : current_physical_size: None,
1065 0 : attachment_status: state.attachment_status(),
1066 0 : generation: (*gen)
1067 0 : .into()
1068 0 : .expect("Tenants are always attached with a generation"),
1069 0 : gc_blocking: None,
1070 0 : })
1071 0 : .collect::<Vec<TenantInfo>>();
1072 0 :
1073 0 : json_response(StatusCode::OK, response_data)
1074 0 : }
1075 :
1076 0 : async fn tenant_status(
1077 0 : request: Request<Body>,
1078 0 : _cancel: CancellationToken,
1079 0 : ) -> Result<Response<Body>, ApiError> {
1080 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1081 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1082 0 : let state = get_state(&request);
1083 0 :
1084 0 : // In tests, sometimes we want to query the state of a tenant without auto-activating it if it's currently waiting.
1085 0 : let activate = true;
1086 : #[cfg(feature = "testing")]
1087 0 : let activate = parse_query_param(&request, "activate")?.unwrap_or(activate);
1088 :
1089 0 : let tenant_info = async {
1090 0 : let tenant = state
1091 0 : .tenant_manager
1092 0 : .get_attached_tenant_shard(tenant_shard_id)?;
1093 :
1094 0 : if activate {
1095 : // This is advisory: we prefer to let the tenant activate on-demand when this function is
1096 : // called, but it is still valid to return 200 and describe the current state of the tenant
1097 : // if it doesn't make it into an active state.
1098 0 : tenant
1099 0 : .wait_to_become_active(ACTIVE_TENANT_TIMEOUT)
1100 0 : .await
1101 0 : .ok();
1102 0 : }
1103 :
1104 : // Calculate total physical size of all timelines
1105 0 : let mut current_physical_size = 0;
1106 0 : for timeline in tenant.list_timelines().iter() {
1107 0 : current_physical_size += timeline.layer_size_sum().await;
1108 : }
1109 :
1110 0 : let state = tenant.current_state();
1111 0 : Result::<_, ApiError>::Ok(TenantDetails {
1112 0 : tenant_info: TenantInfo {
1113 0 : id: tenant_shard_id,
1114 0 : state: state.clone(),
1115 0 : current_physical_size: Some(current_physical_size),
1116 0 : attachment_status: state.attachment_status(),
1117 0 : generation: tenant
1118 0 : .generation()
1119 0 : .into()
1120 0 : .expect("Tenants are always attached with a generation"),
1121 0 : gc_blocking: tenant.gc_block.summary().map(|x| format!("{x:?}")),
1122 0 : },
1123 0 : walredo: tenant.wal_redo_manager_status(),
1124 0 : timelines: tenant.list_timeline_ids(),
1125 0 : })
1126 0 : }
1127 0 : .instrument(info_span!("tenant_status_handler",
1128 : tenant_id = %tenant_shard_id.tenant_id,
1129 0 : shard_id = %tenant_shard_id.shard_slug()))
1130 0 : .await?;
1131 :
1132 0 : json_response(StatusCode::OK, tenant_info)
1133 0 : }
1134 :
1135 0 : async fn tenant_delete_handler(
1136 0 : request: Request<Body>,
1137 0 : _cancel: CancellationToken,
1138 0 : ) -> Result<Response<Body>, ApiError> {
1139 : // TODO openapi spec
1140 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1141 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1142 :
1143 0 : let state = get_state(&request);
1144 0 :
1145 0 : state
1146 0 : .tenant_manager
1147 0 : .delete_tenant(tenant_shard_id)
1148 0 : .instrument(info_span!("tenant_delete_handler",
1149 : tenant_id = %tenant_shard_id.tenant_id,
1150 0 : shard_id = %tenant_shard_id.shard_slug()
1151 : ))
1152 0 : .await?;
1153 :
1154 0 : json_response(StatusCode::OK, ())
1155 0 : }
1156 :
1157 : /// HTTP endpoint to query the current tenant_size of a tenant.
1158 : ///
1159 : /// This is not used by consumption metrics under [`crate::consumption_metrics`], but can be used
1160 : /// to debug any of the calculations. Requires `tenant_id` request parameter, supports
1161 : /// `inputs_only=true|false` (default false) which supports debugging failure to calculate model
1162 : /// values.
1163 : ///
1164 : /// 'retention_period' query parameter overrides the cutoff that is used to calculate the size
1165 : /// (only if it is shorter than the real cutoff).
1166 : ///
1167 : /// Note: we don't update the cached size and prometheus metric here.
1168 : /// The retention period might be different, and it's nice to have a method to just calculate it
1169 : /// without modifying anything anyway.
1170 0 : async fn tenant_size_handler(
1171 0 : request: Request<Body>,
1172 0 : cancel: CancellationToken,
1173 0 : ) -> Result<Response<Body>, ApiError> {
1174 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1175 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1176 0 : let inputs_only: Option<bool> = parse_query_param(&request, "inputs_only")?;
1177 0 : let retention_period: Option<u64> = parse_query_param(&request, "retention_period")?;
1178 0 : let headers = request.headers();
1179 0 : let state = get_state(&request);
1180 0 :
1181 0 : if !tenant_shard_id.is_shard_zero() {
1182 0 : return Err(ApiError::BadRequest(anyhow!(
1183 0 : "Size calculations are only available on shard zero"
1184 0 : )));
1185 0 : }
1186 0 :
1187 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
1188 0 : let tenant = state
1189 0 : .tenant_manager
1190 0 : .get_attached_tenant_shard(tenant_shard_id)?;
1191 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
1192 :
1193 : // this can be long operation
1194 0 : let inputs = tenant
1195 0 : .gather_size_inputs(
1196 0 : retention_period,
1197 0 : LogicalSizeCalculationCause::TenantSizeHandler,
1198 0 : &cancel,
1199 0 : &ctx,
1200 0 : )
1201 0 : .await
1202 0 : .map_err(|e| match e {
1203 0 : crate::tenant::size::CalculateSyntheticSizeError::Cancelled => ApiError::ShuttingDown,
1204 0 : other => ApiError::InternalServerError(anyhow::anyhow!(other)),
1205 0 : })?;
1206 :
1207 0 : let mut sizes = None;
1208 0 : let accepts_html = headers
1209 0 : .get(header::ACCEPT)
1210 0 : .map(|v| v == "text/html")
1211 0 : .unwrap_or_default();
1212 0 : if !inputs_only.unwrap_or(false) {
1213 0 : let storage_model = inputs.calculate_model();
1214 0 : let size = storage_model.calculate();
1215 0 :
1216 0 : // If request header expects html, return html
1217 0 : if accepts_html {
1218 0 : return synthetic_size_html_response(inputs, storage_model, size);
1219 0 : }
1220 0 : sizes = Some(size);
1221 0 : } else if accepts_html {
1222 0 : return Err(ApiError::BadRequest(anyhow!(
1223 0 : "inputs_only parameter is incompatible with html output request"
1224 0 : )));
1225 0 : }
1226 :
1227 : /// The type resides in the pageserver not to expose `ModelInputs`.
1228 : #[derive(serde::Serialize)]
1229 : struct TenantHistorySize {
1230 : id: TenantId,
1231 : /// Size is a mixture of WAL and logical size, so the unit is bytes.
1232 : ///
1233 : /// Will be none if `?inputs_only=true` was given.
1234 : size: Option<u64>,
1235 : /// Size of each segment used in the model.
1236 : /// Will be null if `?inputs_only=true` was given.
1237 : segment_sizes: Option<Vec<tenant_size_model::SegmentSizeResult>>,
1238 : inputs: crate::tenant::size::ModelInputs,
1239 : }
1240 :
1241 0 : json_response(
1242 0 : StatusCode::OK,
1243 0 : TenantHistorySize {
1244 0 : id: tenant_shard_id.tenant_id,
1245 0 : size: sizes.as_ref().map(|x| x.total_size),
1246 0 : segment_sizes: sizes.map(|x| x.segments),
1247 0 : inputs,
1248 0 : },
1249 0 : )
1250 0 : }
1251 :
1252 0 : async fn tenant_shard_split_handler(
1253 0 : mut request: Request<Body>,
1254 0 : _cancel: CancellationToken,
1255 0 : ) -> Result<Response<Body>, ApiError> {
1256 0 : let req: TenantShardSplitRequest = json_request(&mut request).await?;
1257 :
1258 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1259 0 : let state = get_state(&request);
1260 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
1261 :
1262 0 : let tenant = state
1263 0 : .tenant_manager
1264 0 : .get_attached_tenant_shard(tenant_shard_id)?;
1265 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
1266 :
1267 0 : let new_shards = state
1268 0 : .tenant_manager
1269 0 : .shard_split(
1270 0 : tenant,
1271 0 : ShardCount::new(req.new_shard_count),
1272 0 : req.new_stripe_size,
1273 0 : &ctx,
1274 0 : )
1275 0 : .await
1276 0 : .map_err(ApiError::InternalServerError)?;
1277 :
1278 0 : json_response(StatusCode::OK, TenantShardSplitResponse { new_shards })
1279 0 : }
1280 :
1281 0 : async fn layer_map_info_handler(
1282 0 : request: Request<Body>,
1283 0 : _cancel: CancellationToken,
1284 0 : ) -> Result<Response<Body>, ApiError> {
1285 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1286 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1287 0 : let reset: LayerAccessStatsReset =
1288 0 : parse_query_param(&request, "reset")?.unwrap_or(LayerAccessStatsReset::NoReset);
1289 0 : let state = get_state(&request);
1290 0 :
1291 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1292 :
1293 0 : let timeline =
1294 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
1295 0 : .await?;
1296 0 : let layer_map_info = timeline
1297 0 : .layer_map_info(reset)
1298 0 : .await
1299 0 : .map_err(|_shutdown| ApiError::ShuttingDown)?;
1300 :
1301 0 : json_response(StatusCode::OK, layer_map_info)
1302 0 : }
1303 :
1304 0 : #[instrument(skip_all, fields(tenant_id, shard_id, timeline_id, layer_name))]
1305 : async fn timeline_layer_scan_disposable_keys(
1306 : request: Request<Body>,
1307 : cancel: CancellationToken,
1308 : ) -> Result<Response<Body>, ApiError> {
1309 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1310 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1311 : let layer_name: LayerName = parse_request_param(&request, "layer_name")?;
1312 :
1313 : tracing::Span::current().record(
1314 : "tenant_id",
1315 : tracing::field::display(&tenant_shard_id.tenant_id),
1316 : );
1317 : tracing::Span::current().record(
1318 : "shard_id",
1319 : tracing::field::display(tenant_shard_id.shard_slug()),
1320 : );
1321 : tracing::Span::current().record("timeline_id", tracing::field::display(&timeline_id));
1322 : tracing::Span::current().record("layer_name", tracing::field::display(&layer_name));
1323 :
1324 : let state = get_state(&request);
1325 :
1326 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1327 :
1328 : // technically the timeline need not be active for this scan to complete
1329 : let timeline =
1330 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
1331 : .await?;
1332 :
1333 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
1334 :
1335 : let guard = timeline.layers.read().await;
1336 : let Some(layer) = guard.try_get_from_key(&layer_name.clone().into()) else {
1337 : return Err(ApiError::NotFound(
1338 : anyhow::anyhow!("Layer {tenant_shard_id}/{timeline_id}/{layer_name} not found").into(),
1339 : ));
1340 : };
1341 :
1342 : let resident_layer = layer
1343 : .download_and_keep_resident()
1344 : .await
1345 0 : .map_err(|err| match err {
1346 : tenant::storage_layer::layer::DownloadError::TimelineShutdown
1347 : | tenant::storage_layer::layer::DownloadError::DownloadCancelled => {
1348 0 : ApiError::ShuttingDown
1349 : }
1350 : tenant::storage_layer::layer::DownloadError::ContextAndConfigReallyDeniesDownloads
1351 : | tenant::storage_layer::layer::DownloadError::DownloadRequired
1352 : | tenant::storage_layer::layer::DownloadError::NotFile(_)
1353 : | tenant::storage_layer::layer::DownloadError::DownloadFailed
1354 : | tenant::storage_layer::layer::DownloadError::PreStatFailed(_) => {
1355 0 : ApiError::InternalServerError(err.into())
1356 : }
1357 : #[cfg(test)]
1358 : tenant::storage_layer::layer::DownloadError::Failpoint(_) => {
1359 0 : ApiError::InternalServerError(err.into())
1360 : }
1361 0 : })?;
1362 :
1363 : let keys = resident_layer
1364 : .load_keys(&ctx)
1365 : .await
1366 : .map_err(ApiError::InternalServerError)?;
1367 :
1368 : let shard_identity = timeline.get_shard_identity();
1369 :
1370 : let mut disposable_count = 0;
1371 : let mut not_disposable_count = 0;
1372 : let cancel = cancel.clone();
1373 : for (i, key) in keys.into_iter().enumerate() {
1374 : if shard_identity.is_key_disposable(&key) {
1375 : disposable_count += 1;
1376 : tracing::debug!(key = %key, key.dbg=?key, "disposable key");
1377 : } else {
1378 : not_disposable_count += 1;
1379 : }
1380 : #[allow(clippy::collapsible_if)]
1381 : if i % 10000 == 0 {
1382 : if cancel.is_cancelled() || timeline.cancel.is_cancelled() || timeline.is_stopping() {
1383 : return Err(ApiError::ShuttingDown);
1384 : }
1385 : }
1386 : }
1387 :
1388 : json_response(
1389 : StatusCode::OK,
1390 : pageserver_api::models::ScanDisposableKeysResponse {
1391 : disposable_count,
1392 : not_disposable_count,
1393 : },
1394 : )
1395 : }
1396 :
1397 0 : async fn layer_download_handler(
1398 0 : request: Request<Body>,
1399 0 : _cancel: CancellationToken,
1400 0 : ) -> Result<Response<Body>, ApiError> {
1401 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1402 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1403 0 : let layer_file_name = get_request_param(&request, "layer_file_name")?;
1404 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1405 0 : let layer_name = LayerName::from_str(layer_file_name)
1406 0 : .map_err(|s| ApiError::BadRequest(anyhow::anyhow!(s)))?;
1407 0 : let state = get_state(&request);
1408 :
1409 0 : let timeline =
1410 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
1411 0 : .await?;
1412 0 : let downloaded = timeline
1413 0 : .download_layer(&layer_name)
1414 0 : .await
1415 0 : .map_err(ApiError::InternalServerError)?;
1416 :
1417 0 : match downloaded {
1418 0 : Some(true) => json_response(StatusCode::OK, ()),
1419 0 : Some(false) => json_response(StatusCode::NOT_MODIFIED, ()),
1420 0 : None => json_response(
1421 0 : StatusCode::BAD_REQUEST,
1422 0 : format!("Layer {tenant_shard_id}/{timeline_id}/{layer_file_name} not found"),
1423 0 : ),
1424 : }
1425 0 : }
1426 :
1427 0 : async fn evict_timeline_layer_handler(
1428 0 : request: Request<Body>,
1429 0 : _cancel: CancellationToken,
1430 0 : ) -> Result<Response<Body>, ApiError> {
1431 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1432 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1433 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1434 0 : let layer_file_name = get_request_param(&request, "layer_file_name")?;
1435 0 : let state = get_state(&request);
1436 :
1437 0 : let layer_name = LayerName::from_str(layer_file_name)
1438 0 : .map_err(|s| ApiError::BadRequest(anyhow::anyhow!(s)))?;
1439 :
1440 0 : let timeline =
1441 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
1442 0 : .await?;
1443 0 : let evicted = timeline
1444 0 : .evict_layer(&layer_name)
1445 0 : .await
1446 0 : .map_err(ApiError::InternalServerError)?;
1447 :
1448 0 : match evicted {
1449 0 : Some(true) => json_response(StatusCode::OK, ()),
1450 0 : Some(false) => json_response(StatusCode::NOT_MODIFIED, ()),
1451 0 : None => json_response(
1452 0 : StatusCode::BAD_REQUEST,
1453 0 : format!("Layer {tenant_shard_id}/{timeline_id}/{layer_file_name} not found"),
1454 0 : ),
1455 : }
1456 0 : }
1457 :
1458 0 : async fn timeline_gc_blocking_handler(
1459 0 : request: Request<Body>,
1460 0 : _cancel: CancellationToken,
1461 0 : ) -> Result<Response<Body>, ApiError> {
1462 0 : block_or_unblock_gc(request, true).await
1463 0 : }
1464 :
1465 0 : async fn timeline_gc_unblocking_handler(
1466 0 : request: Request<Body>,
1467 0 : _cancel: CancellationToken,
1468 0 : ) -> Result<Response<Body>, ApiError> {
1469 0 : block_or_unblock_gc(request, false).await
1470 0 : }
1471 :
1472 : /// Adding a block is `POST ../block_gc`, removing a block is `POST ../unblock_gc`.
1473 : ///
1474 : /// Both are technically unsafe because they might fire off index uploads, thus they are POST.
1475 0 : async fn block_or_unblock_gc(
1476 0 : request: Request<Body>,
1477 0 : block: bool,
1478 0 : ) -> Result<Response<Body>, ApiError> {
1479 : use crate::tenant::{
1480 : remote_timeline_client::WaitCompletionError, upload_queue::NotInitialized,
1481 : };
1482 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1483 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1484 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1485 0 : let state = get_state(&request);
1486 :
1487 0 : let tenant = state
1488 0 : .tenant_manager
1489 0 : .get_attached_tenant_shard(tenant_shard_id)?;
1490 :
1491 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
1492 :
1493 0 : let timeline = tenant.get_timeline(timeline_id, true)?;
1494 :
1495 0 : let fut = async {
1496 0 : if block {
1497 0 : timeline.block_gc(&tenant).await.map(|_| ())
1498 : } else {
1499 0 : timeline.unblock_gc(&tenant).await
1500 : }
1501 0 : };
1502 :
1503 0 : let span = tracing::info_span!(
1504 : "block_or_unblock_gc",
1505 : tenant_id = %tenant_shard_id.tenant_id,
1506 0 : shard_id = %tenant_shard_id.shard_slug(),
1507 : timeline_id = %timeline_id,
1508 : block = block,
1509 : );
1510 :
1511 0 : let res = fut.instrument(span).await;
1512 :
1513 0 : res.map_err(|e| {
1514 0 : if e.is::<NotInitialized>() || e.is::<WaitCompletionError>() {
1515 0 : ApiError::ShuttingDown
1516 : } else {
1517 0 : ApiError::InternalServerError(e)
1518 : }
1519 0 : })?;
1520 :
1521 0 : json_response(StatusCode::OK, ())
1522 0 : }
1523 :
1524 : /// Get tenant_size SVG graph along with the JSON data.
1525 0 : fn synthetic_size_html_response(
1526 0 : inputs: ModelInputs,
1527 0 : storage_model: StorageModel,
1528 0 : sizes: SizeResult,
1529 0 : ) -> Result<Response<Body>, ApiError> {
1530 0 : let mut timeline_ids: Vec<String> = Vec::new();
1531 0 : let mut timeline_map: HashMap<TimelineId, usize> = HashMap::new();
1532 0 : for (index, ti) in inputs.timeline_inputs.iter().enumerate() {
1533 0 : timeline_map.insert(ti.timeline_id, index);
1534 0 : timeline_ids.push(ti.timeline_id.to_string());
1535 0 : }
1536 0 : let seg_to_branch: Vec<(usize, SvgBranchKind)> = inputs
1537 0 : .segments
1538 0 : .iter()
1539 0 : .map(|seg| {
1540 0 : (
1541 0 : *timeline_map.get(&seg.timeline_id).unwrap(),
1542 0 : seg.kind.into(),
1543 0 : )
1544 0 : })
1545 0 : .collect();
1546 :
1547 0 : let svg =
1548 0 : tenant_size_model::svg::draw_svg(&storage_model, &timeline_ids, &seg_to_branch, &sizes)
1549 0 : .map_err(ApiError::InternalServerError)?;
1550 :
1551 0 : let mut response = String::new();
1552 :
1553 : use std::fmt::Write;
1554 0 : write!(response, "<html>\n<body>\n").unwrap();
1555 0 : write!(response, "<div>\n{svg}\n</div>").unwrap();
1556 0 : writeln!(response, "Project size: {}", sizes.total_size).unwrap();
1557 0 : writeln!(response, "<pre>").unwrap();
1558 0 : writeln!(
1559 0 : response,
1560 0 : "{}",
1561 0 : serde_json::to_string_pretty(&inputs).unwrap()
1562 0 : )
1563 0 : .unwrap();
1564 0 : writeln!(
1565 0 : response,
1566 0 : "{}",
1567 0 : serde_json::to_string_pretty(&sizes.segments).unwrap()
1568 0 : )
1569 0 : .unwrap();
1570 0 : writeln!(response, "</pre>").unwrap();
1571 0 : write!(response, "</body>\n</html>\n").unwrap();
1572 0 :
1573 0 : html_response(StatusCode::OK, response)
1574 0 : }
1575 :
1576 0 : pub fn html_response(status: StatusCode, data: String) -> Result<Response<Body>, ApiError> {
1577 0 : let response = Response::builder()
1578 0 : .status(status)
1579 0 : .header(header::CONTENT_TYPE, "text/html")
1580 0 : .body(Body::from(data.as_bytes().to_vec()))
1581 0 : .map_err(|e| ApiError::InternalServerError(e.into()))?;
1582 0 : Ok(response)
1583 0 : }
1584 :
1585 0 : async fn get_tenant_config_handler(
1586 0 : request: Request<Body>,
1587 0 : _cancel: CancellationToken,
1588 0 : ) -> Result<Response<Body>, ApiError> {
1589 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1590 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1591 0 : let state = get_state(&request);
1592 :
1593 0 : let tenant = state
1594 0 : .tenant_manager
1595 0 : .get_attached_tenant_shard(tenant_shard_id)?;
1596 :
1597 0 : let response = HashMap::from([
1598 : (
1599 : "tenant_specific_overrides",
1600 0 : serde_json::to_value(tenant.tenant_specific_overrides())
1601 0 : .context("serializing tenant specific overrides")
1602 0 : .map_err(ApiError::InternalServerError)?,
1603 : ),
1604 : (
1605 0 : "effective_config",
1606 0 : serde_json::to_value(tenant.effective_config())
1607 0 : .context("serializing effective config")
1608 0 : .map_err(ApiError::InternalServerError)?,
1609 : ),
1610 : ]);
1611 :
1612 0 : json_response(StatusCode::OK, response)
1613 0 : }
1614 :
1615 0 : async fn update_tenant_config_handler(
1616 0 : mut request: Request<Body>,
1617 0 : _cancel: CancellationToken,
1618 0 : ) -> Result<Response<Body>, ApiError> {
1619 0 : let request_data: TenantConfigRequest = json_request(&mut request).await?;
1620 0 : let tenant_id = request_data.tenant_id;
1621 0 : check_permission(&request, Some(tenant_id))?;
1622 :
1623 0 : let new_tenant_conf =
1624 0 : TenantConfOpt::try_from(&request_data.config).map_err(ApiError::BadRequest)?;
1625 :
1626 0 : let state = get_state(&request);
1627 0 :
1628 0 : let tenant_shard_id = TenantShardId::unsharded(tenant_id);
1629 :
1630 0 : let tenant = state
1631 0 : .tenant_manager
1632 0 : .get_attached_tenant_shard(tenant_shard_id)?;
1633 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
1634 :
1635 : // This is a legacy API that only operates on attached tenants: the preferred
1636 : // API to use is the location_config/ endpoint, which lets the caller provide
1637 : // the full LocationConf.
1638 0 : let location_conf = LocationConf::attached_single(
1639 0 : new_tenant_conf.clone(),
1640 0 : tenant.get_generation(),
1641 0 : &ShardParameters::default(),
1642 0 : );
1643 0 :
1644 0 : crate::tenant::Tenant::persist_tenant_config(state.conf, &tenant_shard_id, &location_conf)
1645 0 : .await
1646 0 : .map_err(|e| ApiError::InternalServerError(anyhow::anyhow!(e)))?;
1647 0 : tenant.set_new_tenant_config(new_tenant_conf);
1648 0 :
1649 0 : json_response(StatusCode::OK, ())
1650 0 : }
1651 :
1652 0 : async fn put_tenant_location_config_handler(
1653 0 : mut request: Request<Body>,
1654 0 : _cancel: CancellationToken,
1655 0 : ) -> Result<Response<Body>, ApiError> {
1656 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1657 :
1658 0 : let request_data: TenantLocationConfigRequest = json_request(&mut request).await?;
1659 0 : let flush = parse_query_param(&request, "flush_ms")?.map(Duration::from_millis);
1660 0 : let lazy = parse_query_param(&request, "lazy")?.unwrap_or(false);
1661 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1662 :
1663 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
1664 0 : let state = get_state(&request);
1665 0 : let conf = state.conf;
1666 0 :
1667 0 : // The `Detached` state is special, it doesn't upsert a tenant, it removes
1668 0 : // its local disk content and drops it from memory.
1669 0 : if let LocationConfigMode::Detached = request_data.config.mode {
1670 0 : if let Err(e) = state
1671 0 : .tenant_manager
1672 0 : .detach_tenant(conf, tenant_shard_id, &state.deletion_queue_client)
1673 0 : .instrument(info_span!("tenant_detach",
1674 : tenant_id = %tenant_shard_id.tenant_id,
1675 0 : shard_id = %tenant_shard_id.shard_slug()
1676 : ))
1677 0 : .await
1678 : {
1679 0 : match e {
1680 0 : TenantStateError::SlotError(TenantSlotError::NotFound(_)) => {
1681 0 : // This API is idempotent: a NotFound on a detach is fine.
1682 0 : }
1683 0 : _ => return Err(e.into()),
1684 : }
1685 0 : }
1686 0 : return json_response(StatusCode::OK, ());
1687 0 : }
1688 :
1689 0 : let location_conf =
1690 0 : LocationConf::try_from(&request_data.config).map_err(ApiError::BadRequest)?;
1691 :
1692 : // lazy==true queues up for activation or jumps the queue like normal when a compute connects,
1693 : // similar to at startup ordering.
1694 0 : let spawn_mode = if lazy {
1695 0 : tenant::SpawnMode::Lazy
1696 : } else {
1697 0 : tenant::SpawnMode::Eager
1698 : };
1699 :
1700 0 : let tenant = state
1701 0 : .tenant_manager
1702 0 : .upsert_location(tenant_shard_id, location_conf, flush, spawn_mode, &ctx)
1703 0 : .await?;
1704 0 : let stripe_size = tenant.as_ref().map(|t| t.get_shard_stripe_size());
1705 0 : let attached = tenant.is_some();
1706 :
1707 0 : if let Some(_flush_ms) = flush {
1708 0 : match state
1709 0 : .secondary_controller
1710 0 : .upload_tenant(tenant_shard_id)
1711 0 : .await
1712 : {
1713 : Ok(()) => {
1714 0 : tracing::info!("Uploaded heatmap during flush");
1715 : }
1716 0 : Err(e) => {
1717 0 : tracing::warn!("Failed to flush heatmap: {e}");
1718 : }
1719 : }
1720 : } else {
1721 0 : tracing::info!("No flush requested when configuring");
1722 : }
1723 :
1724 : // This API returns a vector of pageservers where the tenant is attached: this is
1725 : // primarily for use in the sharding service. For compatibilty, we also return this
1726 : // when called directly on a pageserver, but the payload is always zero or one shards.
1727 0 : let mut response = TenantLocationConfigResponse {
1728 0 : shards: Vec::new(),
1729 0 : stripe_size: None,
1730 0 : };
1731 0 : if attached {
1732 0 : response.shards.push(TenantShardLocation {
1733 0 : shard_id: tenant_shard_id,
1734 0 : node_id: state.conf.id,
1735 0 : });
1736 0 : if tenant_shard_id.shard_count.count() > 1 {
1737 : // Stripe size should be set if we are attached
1738 0 : debug_assert!(stripe_size.is_some());
1739 0 : response.stripe_size = stripe_size;
1740 0 : }
1741 0 : }
1742 :
1743 0 : json_response(StatusCode::OK, response)
1744 0 : }
1745 :
1746 0 : async fn list_location_config_handler(
1747 0 : request: Request<Body>,
1748 0 : _cancel: CancellationToken,
1749 0 : ) -> Result<Response<Body>, ApiError> {
1750 0 : let state = get_state(&request);
1751 0 : let slots = state.tenant_manager.list();
1752 0 : let result = LocationConfigListResponse {
1753 0 : tenant_shards: slots
1754 0 : .into_iter()
1755 0 : .map(|(tenant_shard_id, slot)| {
1756 0 : let v = match slot {
1757 0 : TenantSlot::Attached(t) => Some(t.get_location_conf()),
1758 0 : TenantSlot::Secondary(s) => Some(s.get_location_conf()),
1759 0 : TenantSlot::InProgress(_) => None,
1760 : };
1761 0 : (tenant_shard_id, v)
1762 0 : })
1763 0 : .collect(),
1764 0 : };
1765 0 : json_response(StatusCode::OK, result)
1766 0 : }
1767 :
1768 0 : async fn get_location_config_handler(
1769 0 : request: Request<Body>,
1770 0 : _cancel: CancellationToken,
1771 0 : ) -> Result<Response<Body>, ApiError> {
1772 0 : let state = get_state(&request);
1773 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1774 0 : let slot = state.tenant_manager.get(tenant_shard_id);
1775 :
1776 0 : let Some(slot) = slot else {
1777 0 : return Err(ApiError::NotFound(
1778 0 : anyhow::anyhow!("Tenant shard not found").into(),
1779 0 : ));
1780 : };
1781 :
1782 0 : let result: Option<LocationConfig> = match slot {
1783 0 : TenantSlot::Attached(t) => Some(t.get_location_conf()),
1784 0 : TenantSlot::Secondary(s) => Some(s.get_location_conf()),
1785 0 : TenantSlot::InProgress(_) => None,
1786 : };
1787 :
1788 0 : json_response(StatusCode::OK, result)
1789 0 : }
1790 :
1791 : // Do a time travel recovery on the given tenant/tenant shard. Tenant needs to be detached
1792 : // (from all pageservers) as it invalidates consistency assumptions.
1793 0 : async fn tenant_time_travel_remote_storage_handler(
1794 0 : request: Request<Body>,
1795 0 : cancel: CancellationToken,
1796 0 : ) -> Result<Response<Body>, ApiError> {
1797 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1798 :
1799 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1800 :
1801 0 : let timestamp_raw = must_get_query_param(&request, "travel_to")?;
1802 0 : let timestamp = humantime::parse_rfc3339(×tamp_raw)
1803 0 : .with_context(|| format!("Invalid time for travel_to: {timestamp_raw:?}"))
1804 0 : .map_err(ApiError::BadRequest)?;
1805 :
1806 0 : let done_if_after_raw = must_get_query_param(&request, "done_if_after")?;
1807 0 : let done_if_after = humantime::parse_rfc3339(&done_if_after_raw)
1808 0 : .with_context(|| format!("Invalid time for done_if_after: {done_if_after_raw:?}"))
1809 0 : .map_err(ApiError::BadRequest)?;
1810 :
1811 : // This is just a sanity check to fend off naive wrong usages of the API:
1812 : // the tenant needs to be detached *everywhere*
1813 0 : let state = get_state(&request);
1814 0 : let we_manage_tenant = state.tenant_manager.manages_tenant_shard(tenant_shard_id);
1815 0 : if we_manage_tenant {
1816 0 : return Err(ApiError::BadRequest(anyhow!(
1817 0 : "Tenant {tenant_shard_id} is already attached at this pageserver"
1818 0 : )));
1819 0 : }
1820 0 :
1821 0 : if timestamp > done_if_after {
1822 0 : return Err(ApiError::BadRequest(anyhow!(
1823 0 : "The done_if_after timestamp comes before the timestamp to recover to"
1824 0 : )));
1825 0 : }
1826 0 :
1827 0 : tracing::info!("Issuing time travel request internally. timestamp={timestamp_raw}, done_if_after={done_if_after_raw}");
1828 :
1829 0 : remote_timeline_client::upload::time_travel_recover_tenant(
1830 0 : &state.remote_storage,
1831 0 : &tenant_shard_id,
1832 0 : timestamp,
1833 0 : done_if_after,
1834 0 : &cancel,
1835 0 : )
1836 0 : .await
1837 0 : .map_err(|e| match e {
1838 0 : TimeTravelError::BadInput(e) => {
1839 0 : warn!("bad input error: {e}");
1840 0 : ApiError::BadRequest(anyhow!("bad input error"))
1841 : }
1842 : TimeTravelError::Unimplemented => {
1843 0 : ApiError::BadRequest(anyhow!("unimplemented for the configured remote storage"))
1844 : }
1845 0 : TimeTravelError::Cancelled => ApiError::InternalServerError(anyhow!("cancelled")),
1846 : TimeTravelError::TooManyVersions => {
1847 0 : ApiError::InternalServerError(anyhow!("too many versions in remote storage"))
1848 : }
1849 0 : TimeTravelError::Other(e) => {
1850 0 : warn!("internal error: {e}");
1851 0 : ApiError::InternalServerError(anyhow!("internal error"))
1852 : }
1853 0 : })?;
1854 :
1855 0 : json_response(StatusCode::OK, ())
1856 0 : }
1857 :
1858 : /// Testing helper to transition a tenant to [`crate::tenant::TenantState::Broken`].
1859 0 : async fn handle_tenant_break(
1860 0 : r: Request<Body>,
1861 0 : _cancel: CancellationToken,
1862 0 : ) -> Result<Response<Body>, ApiError> {
1863 0 : let tenant_shard_id: TenantShardId = parse_request_param(&r, "tenant_shard_id")?;
1864 :
1865 0 : let state = get_state(&r);
1866 0 : state
1867 0 : .tenant_manager
1868 0 : .get_attached_tenant_shard(tenant_shard_id)?
1869 0 : .set_broken("broken from test".to_owned())
1870 0 : .await;
1871 :
1872 0 : json_response(StatusCode::OK, ())
1873 0 : }
1874 :
1875 : // Obtains an lsn lease on the given timeline.
1876 0 : async fn lsn_lease_handler(
1877 0 : mut request: Request<Body>,
1878 0 : _cancel: CancellationToken,
1879 0 : ) -> Result<Response<Body>, ApiError> {
1880 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1881 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1882 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1883 0 : let lsn = json_request::<LsnLeaseRequest>(&mut request).await?.lsn;
1884 :
1885 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
1886 0 :
1887 0 : let state = get_state(&request);
1888 :
1889 0 : let timeline =
1890 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
1891 0 : .await?;
1892 :
1893 0 : let result = async {
1894 0 : timeline
1895 0 : .init_lsn_lease(lsn, timeline.get_lsn_lease_length(), &ctx)
1896 0 : .map_err(|e| {
1897 0 : ApiError::InternalServerError(
1898 0 : e.context(format!("invalid lsn lease request at {lsn}")),
1899 0 : )
1900 0 : })
1901 0 : }
1902 0 : .instrument(info_span!("init_lsn_lease", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug(), %timeline_id))
1903 0 : .await?;
1904 :
1905 0 : json_response(StatusCode::OK, result)
1906 0 : }
1907 :
1908 : // Run GC immediately on given timeline.
1909 0 : async fn timeline_gc_handler(
1910 0 : mut request: Request<Body>,
1911 0 : cancel: CancellationToken,
1912 0 : ) -> Result<Response<Body>, ApiError> {
1913 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1914 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1915 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1916 :
1917 0 : let gc_req: TimelineGcRequest = json_request(&mut request).await?;
1918 :
1919 0 : let state = get_state(&request);
1920 0 :
1921 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
1922 0 : let gc_result = state
1923 0 : .tenant_manager
1924 0 : .immediate_gc(tenant_shard_id, timeline_id, gc_req, cancel, &ctx)
1925 0 : .await?;
1926 :
1927 0 : json_response(StatusCode::OK, gc_result)
1928 0 : }
1929 :
1930 : // Run compaction immediately on given timeline.
1931 0 : async fn timeline_compact_handler(
1932 0 : mut request: Request<Body>,
1933 0 : cancel: CancellationToken,
1934 0 : ) -> Result<Response<Body>, ApiError> {
1935 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1936 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1937 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1938 :
1939 0 : let compact_range = json_request_maybe::<Option<CompactRange>>(&mut request).await?;
1940 :
1941 0 : let state = get_state(&request);
1942 0 :
1943 0 : let mut flags = EnumSet::empty();
1944 0 :
1945 0 : if Some(true) == parse_query_param::<_, bool>(&request, "force_l0_compaction")? {
1946 0 : flags |= CompactFlags::ForceL0Compaction;
1947 0 : }
1948 0 : if Some(true) == parse_query_param::<_, bool>(&request, "force_repartition")? {
1949 0 : flags |= CompactFlags::ForceRepartition;
1950 0 : }
1951 0 : if Some(true) == parse_query_param::<_, bool>(&request, "force_image_layer_creation")? {
1952 0 : flags |= CompactFlags::ForceImageLayerCreation;
1953 0 : }
1954 0 : if Some(true) == parse_query_param::<_, bool>(&request, "enhanced_gc_bottom_most_compaction")? {
1955 0 : flags |= CompactFlags::EnhancedGcBottomMostCompaction;
1956 0 : }
1957 0 : if Some(true) == parse_query_param::<_, bool>(&request, "dry_run")? {
1958 0 : flags |= CompactFlags::DryRun;
1959 0 : }
1960 :
1961 0 : let wait_until_uploaded =
1962 0 : parse_query_param::<_, bool>(&request, "wait_until_uploaded")?.unwrap_or(false);
1963 0 :
1964 0 : let options = CompactOptions {
1965 0 : compact_range,
1966 0 : flags,
1967 0 : };
1968 :
1969 0 : async {
1970 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
1971 0 : let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
1972 0 : timeline
1973 0 : .compact_with_options(&cancel, options, &ctx)
1974 0 : .await
1975 0 : .map_err(|e| ApiError::InternalServerError(e.into()))?;
1976 0 : if wait_until_uploaded {
1977 0 : timeline.remote_client.wait_completion().await
1978 : // XXX map to correct ApiError for the cases where it's due to shutdown
1979 0 : .context("wait completion").map_err(ApiError::InternalServerError)?;
1980 0 : }
1981 0 : json_response(StatusCode::OK, ())
1982 0 : }
1983 0 : .instrument(info_span!("manual_compaction", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug(), %timeline_id))
1984 0 : .await
1985 0 : }
1986 :
1987 : // Run offload immediately on given timeline.
1988 0 : async fn timeline_offload_handler(
1989 0 : request: Request<Body>,
1990 0 : _cancel: CancellationToken,
1991 0 : ) -> Result<Response<Body>, ApiError> {
1992 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
1993 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
1994 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
1995 :
1996 0 : let state = get_state(&request);
1997 :
1998 0 : async {
1999 0 : let tenant = state
2000 0 : .tenant_manager
2001 0 : .get_attached_tenant_shard(tenant_shard_id)?;
2002 :
2003 0 : if tenant.get_offloaded_timeline(timeline_id).is_ok() {
2004 0 : return json_response(StatusCode::OK, ());
2005 0 : }
2006 0 : let timeline =
2007 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
2008 0 : .await?;
2009 :
2010 0 : if !tenant.timeline_has_no_attached_children(timeline_id) {
2011 0 : return Err(ApiError::PreconditionFailed(
2012 0 : "timeline has attached children".into(),
2013 0 : ));
2014 0 : }
2015 0 : if let (false, reason) = timeline.can_offload() {
2016 0 : return Err(ApiError::PreconditionFailed(
2017 0 : format!("Timeline::can_offload() check failed: {}", reason) .into(),
2018 0 : ));
2019 0 : }
2020 0 : offload_timeline(&tenant, &timeline)
2021 0 : .await
2022 0 : .map_err(|e| {
2023 0 : match e {
2024 0 : OffloadError::Cancelled => ApiError::ResourceUnavailable("Timeline shutting down".into()),
2025 0 : _ => ApiError::InternalServerError(anyhow!(e))
2026 : }
2027 0 : })?;
2028 :
2029 0 : json_response(StatusCode::OK, ())
2030 0 : }
2031 0 : .instrument(info_span!("manual_timeline_offload", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug(), %timeline_id))
2032 0 : .await
2033 0 : }
2034 :
2035 : // Run checkpoint immediately on given timeline.
2036 0 : async fn timeline_checkpoint_handler(
2037 0 : request: Request<Body>,
2038 0 : cancel: CancellationToken,
2039 0 : ) -> Result<Response<Body>, ApiError> {
2040 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2041 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2042 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2043 :
2044 0 : let state = get_state(&request);
2045 0 :
2046 0 : let mut flags = EnumSet::empty();
2047 0 : if Some(true) == parse_query_param::<_, bool>(&request, "force_l0_compaction")? {
2048 0 : flags |= CompactFlags::ForceL0Compaction;
2049 0 : }
2050 0 : if Some(true) == parse_query_param::<_, bool>(&request, "force_repartition")? {
2051 0 : flags |= CompactFlags::ForceRepartition;
2052 0 : }
2053 0 : if Some(true) == parse_query_param::<_, bool>(&request, "force_image_layer_creation")? {
2054 0 : flags |= CompactFlags::ForceImageLayerCreation;
2055 0 : }
2056 :
2057 : // By default, checkpoints come with a compaction, but this may be optionally disabled by tests that just want to flush + upload.
2058 0 : let compact = parse_query_param::<_, bool>(&request, "compact")?.unwrap_or(true);
2059 :
2060 0 : let wait_until_uploaded =
2061 0 : parse_query_param::<_, bool>(&request, "wait_until_uploaded")?.unwrap_or(false);
2062 :
2063 0 : async {
2064 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
2065 0 : let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
2066 0 : timeline
2067 0 : .freeze_and_flush()
2068 0 : .await
2069 0 : .map_err(|e| {
2070 0 : match e {
2071 0 : tenant::timeline::FlushLayerError::Cancelled => ApiError::ShuttingDown,
2072 0 : other => ApiError::InternalServerError(other.into()),
2073 :
2074 : }
2075 0 : })?;
2076 0 : if compact {
2077 0 : timeline
2078 0 : .compact(&cancel, flags, &ctx)
2079 0 : .await
2080 0 : .map_err(|e|
2081 0 : match e {
2082 0 : CompactionError::ShuttingDown => ApiError::ShuttingDown,
2083 0 : CompactionError::Offload(e) => ApiError::InternalServerError(anyhow::anyhow!(e)),
2084 0 : CompactionError::Other(e) => ApiError::InternalServerError(e)
2085 0 : }
2086 0 : )?;
2087 0 : }
2088 :
2089 0 : if wait_until_uploaded {
2090 0 : tracing::info!("Waiting for uploads to complete...");
2091 0 : timeline.remote_client.wait_completion().await
2092 : // XXX map to correct ApiError for the cases where it's due to shutdown
2093 0 : .context("wait completion").map_err(ApiError::InternalServerError)?;
2094 0 : tracing::info!("Uploads completed up to {}", timeline.get_remote_consistent_lsn_projected().unwrap_or(Lsn(0)));
2095 0 : }
2096 :
2097 0 : json_response(StatusCode::OK, ())
2098 0 : }
2099 0 : .instrument(info_span!("manual_checkpoint", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug(), %timeline_id))
2100 0 : .await
2101 0 : }
2102 :
2103 0 : async fn timeline_download_remote_layers_handler_post(
2104 0 : mut request: Request<Body>,
2105 0 : _cancel: CancellationToken,
2106 0 : ) -> Result<Response<Body>, ApiError> {
2107 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2108 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2109 0 : let body: DownloadRemoteLayersTaskSpawnRequest = json_request(&mut request).await?;
2110 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2111 :
2112 0 : let state = get_state(&request);
2113 :
2114 0 : let timeline =
2115 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
2116 0 : .await?;
2117 0 : match timeline.spawn_download_all_remote_layers(body).await {
2118 0 : Ok(st) => json_response(StatusCode::ACCEPTED, st),
2119 0 : Err(st) => json_response(StatusCode::CONFLICT, st),
2120 : }
2121 0 : }
2122 :
2123 0 : async fn timeline_download_remote_layers_handler_get(
2124 0 : request: Request<Body>,
2125 0 : _cancel: CancellationToken,
2126 0 : ) -> Result<Response<Body>, ApiError> {
2127 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2128 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2129 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2130 0 : let state = get_state(&request);
2131 :
2132 0 : let timeline =
2133 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
2134 0 : .await?;
2135 0 : let info = timeline
2136 0 : .get_download_all_remote_layers_task_info()
2137 0 : .context("task never started since last pageserver process start")
2138 0 : .map_err(|e| ApiError::NotFound(e.into()))?;
2139 0 : json_response(StatusCode::OK, info)
2140 0 : }
2141 :
2142 0 : async fn timeline_detach_ancestor_handler(
2143 0 : request: Request<Body>,
2144 0 : _cancel: CancellationToken,
2145 0 : ) -> Result<Response<Body>, ApiError> {
2146 : use crate::tenant::timeline::detach_ancestor;
2147 : use pageserver_api::models::detach_ancestor::AncestorDetached;
2148 :
2149 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2150 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2151 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2152 :
2153 0 : let span = tracing::info_span!("detach_ancestor", tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id);
2154 :
2155 0 : async move {
2156 0 : let mut options = detach_ancestor::Options::default();
2157 :
2158 0 : let rewrite_concurrency =
2159 0 : parse_query_param::<_, std::num::NonZeroUsize>(&request, "rewrite_concurrency")?;
2160 0 : let copy_concurrency =
2161 0 : parse_query_param::<_, std::num::NonZeroUsize>(&request, "copy_concurrency")?;
2162 :
2163 0 : [
2164 0 : (&mut options.rewrite_concurrency, rewrite_concurrency),
2165 0 : (&mut options.copy_concurrency, copy_concurrency),
2166 0 : ]
2167 0 : .into_iter()
2168 0 : .filter_map(|(target, val)| val.map(|val| (target, val)))
2169 0 : .for_each(|(target, val)| *target = val);
2170 0 :
2171 0 : let state = get_state(&request);
2172 :
2173 0 : let tenant = state
2174 0 : .tenant_manager
2175 0 : .get_attached_tenant_shard(tenant_shard_id)?;
2176 :
2177 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
2178 :
2179 0 : let ctx = RequestContext::new(TaskKind::DetachAncestor, DownloadBehavior::Download);
2180 0 : let ctx = &ctx;
2181 :
2182 : // Flush the upload queues of all timelines before detaching ancestor. We do the same thing again
2183 : // during shutdown. This early upload ensures the pageserver does not need to upload too many
2184 : // things and creates downtime during timeline reloads.
2185 0 : for timeline in tenant.list_timelines() {
2186 0 : timeline
2187 0 : .remote_client
2188 0 : .wait_completion()
2189 0 : .await
2190 0 : .map_err(|e| {
2191 0 : ApiError::PreconditionFailed(format!("cannot drain upload queue: {e}").into())
2192 0 : })?;
2193 : }
2194 :
2195 0 : tracing::info!("all timeline upload queues are drained");
2196 :
2197 0 : let timeline = tenant.get_timeline(timeline_id, true)?;
2198 :
2199 0 : let progress = timeline
2200 0 : .prepare_to_detach_from_ancestor(&tenant, options, ctx)
2201 0 : .await?;
2202 :
2203 : // uncomment to allow early as possible Tenant::drop
2204 : // drop(tenant);
2205 :
2206 0 : let resp = match progress {
2207 0 : detach_ancestor::Progress::Prepared(attempt, prepared) => {
2208 : // it would be great to tag the guard on to the tenant activation future
2209 0 : let reparented_timelines = state
2210 0 : .tenant_manager
2211 0 : .complete_detaching_timeline_ancestor(
2212 0 : tenant_shard_id,
2213 0 : timeline_id,
2214 0 : prepared,
2215 0 : attempt,
2216 0 : ctx,
2217 0 : )
2218 0 : .await?;
2219 :
2220 0 : AncestorDetached {
2221 0 : reparented_timelines,
2222 0 : }
2223 : }
2224 0 : detach_ancestor::Progress::Done(resp) => resp,
2225 : };
2226 :
2227 0 : json_response(StatusCode::OK, resp)
2228 0 : }
2229 0 : .instrument(span)
2230 0 : .await
2231 0 : }
2232 :
2233 0 : async fn deletion_queue_flush(
2234 0 : r: Request<Body>,
2235 0 : cancel: CancellationToken,
2236 0 : ) -> Result<Response<Body>, ApiError> {
2237 0 : let state = get_state(&r);
2238 :
2239 0 : let execute = parse_query_param(&r, "execute")?.unwrap_or(false);
2240 0 :
2241 0 : let flush = async {
2242 0 : if execute {
2243 0 : state.deletion_queue_client.flush_execute().await
2244 : } else {
2245 0 : state.deletion_queue_client.flush().await
2246 : }
2247 0 : }
2248 : // DeletionQueueError's only case is shutting down.
2249 0 : .map_err(|_| ApiError::ShuttingDown);
2250 0 :
2251 0 : tokio::select! {
2252 0 : res = flush => {
2253 0 : res.map(|()| json_response(StatusCode::OK, ()))?
2254 : }
2255 0 : _ = cancel.cancelled() => {
2256 0 : Err(ApiError::ShuttingDown)
2257 : }
2258 : }
2259 0 : }
2260 :
2261 : /// Try if `GetPage@Lsn` is successful, useful for manual debugging.
2262 0 : async fn getpage_at_lsn_handler(
2263 0 : request: Request<Body>,
2264 0 : _cancel: CancellationToken,
2265 0 : ) -> Result<Response<Body>, ApiError> {
2266 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2267 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2268 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2269 0 : let state = get_state(&request);
2270 :
2271 : struct Key(pageserver_api::key::Key);
2272 :
2273 : impl std::str::FromStr for Key {
2274 : type Err = anyhow::Error;
2275 :
2276 0 : fn from_str(s: &str) -> std::result::Result<Self, Self::Err> {
2277 0 : pageserver_api::key::Key::from_hex(s).map(Key)
2278 0 : }
2279 : }
2280 :
2281 0 : let key: Key = parse_query_param(&request, "key")?
2282 0 : .ok_or_else(|| ApiError::BadRequest(anyhow!("missing 'key' query parameter")))?;
2283 0 : let lsn: Lsn = parse_query_param(&request, "lsn")?
2284 0 : .ok_or_else(|| ApiError::BadRequest(anyhow!("missing 'lsn' query parameter")))?;
2285 :
2286 0 : async {
2287 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
2288 0 : let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
2289 :
2290 0 : let page = timeline.get(key.0, lsn, &ctx).await?;
2291 :
2292 0 : Result::<_, ApiError>::Ok(
2293 0 : Response::builder()
2294 0 : .status(StatusCode::OK)
2295 0 : .header(header::CONTENT_TYPE, "application/octet-stream")
2296 0 : .body(hyper::Body::from(page))
2297 0 : .unwrap(),
2298 0 : )
2299 0 : }
2300 0 : .instrument(info_span!("timeline_get", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug(), %timeline_id))
2301 0 : .await
2302 0 : }
2303 :
2304 0 : async fn timeline_collect_keyspace(
2305 0 : request: Request<Body>,
2306 0 : _cancel: CancellationToken,
2307 0 : ) -> Result<Response<Body>, ApiError> {
2308 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2309 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2310 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2311 0 : let state = get_state(&request);
2312 :
2313 0 : let at_lsn: Option<Lsn> = parse_query_param(&request, "at_lsn")?;
2314 :
2315 0 : async {
2316 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
2317 0 : let timeline = active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id).await?;
2318 0 : let at_lsn = at_lsn.unwrap_or_else(|| timeline.get_last_record_lsn());
2319 0 : let (dense_ks, sparse_ks) = timeline
2320 0 : .collect_keyspace(at_lsn, &ctx)
2321 0 : .await
2322 0 : .map_err(|e| ApiError::InternalServerError(e.into()))?;
2323 :
2324 : // This API is currently used by pagebench. Pagebench will iterate all keys within the keyspace.
2325 : // Therefore, we split dense/sparse keys in this API.
2326 0 : let res = pageserver_api::models::partitioning::Partitioning { keys: dense_ks, sparse_keys: sparse_ks, at_lsn };
2327 0 :
2328 0 : json_response(StatusCode::OK, res)
2329 0 : }
2330 0 : .instrument(info_span!("timeline_collect_keyspace", tenant_id = %tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug(), %timeline_id))
2331 0 : .await
2332 0 : }
2333 :
2334 0 : async fn active_timeline_of_active_tenant(
2335 0 : tenant_manager: &TenantManager,
2336 0 : tenant_shard_id: TenantShardId,
2337 0 : timeline_id: TimelineId,
2338 0 : ) -> Result<Arc<Timeline>, ApiError> {
2339 0 : let tenant = tenant_manager.get_attached_tenant_shard(tenant_shard_id)?;
2340 :
2341 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
2342 :
2343 0 : Ok(tenant.get_timeline(timeline_id, true)?)
2344 0 : }
2345 :
2346 0 : async fn always_panic_handler(
2347 0 : req: Request<Body>,
2348 0 : _cancel: CancellationToken,
2349 0 : ) -> Result<Response<Body>, ApiError> {
2350 0 : // Deliberately cause a panic to exercise the panic hook registered via std::panic::set_hook().
2351 0 : // For pageserver, the relevant panic hook is `tracing_panic_hook` , and the `sentry` crate's wrapper around it.
2352 0 : // Use catch_unwind to ensure that tokio nor hyper are distracted by our panic.
2353 0 : let query = req.uri().query();
2354 0 : let _ = std::panic::catch_unwind(|| {
2355 0 : panic!("unconditional panic for testing panic hook integration; request query: {query:?}")
2356 0 : });
2357 0 : json_response(StatusCode::NO_CONTENT, ())
2358 0 : }
2359 :
2360 0 : async fn disk_usage_eviction_run(
2361 0 : mut r: Request<Body>,
2362 0 : cancel: CancellationToken,
2363 0 : ) -> Result<Response<Body>, ApiError> {
2364 0 : check_permission(&r, None)?;
2365 :
2366 0 : #[derive(Debug, Clone, Copy, serde::Serialize, serde::Deserialize)]
2367 : struct Config {
2368 : /// How many bytes to evict before reporting that pressure is relieved.
2369 : evict_bytes: u64,
2370 :
2371 : #[serde(default)]
2372 : eviction_order: pageserver_api::config::EvictionOrder,
2373 : }
2374 :
2375 : #[derive(Debug, Clone, Copy, serde::Serialize)]
2376 : struct Usage {
2377 : // remains unchanged after instantiation of the struct
2378 : evict_bytes: u64,
2379 : // updated by `add_available_bytes`
2380 : freed_bytes: u64,
2381 : }
2382 :
2383 : impl crate::disk_usage_eviction_task::Usage for Usage {
2384 0 : fn has_pressure(&self) -> bool {
2385 0 : self.evict_bytes > self.freed_bytes
2386 0 : }
2387 :
2388 0 : fn add_available_bytes(&mut self, bytes: u64) {
2389 0 : self.freed_bytes += bytes;
2390 0 : }
2391 : }
2392 :
2393 0 : let config = json_request::<Config>(&mut r).await?;
2394 :
2395 0 : let usage = Usage {
2396 0 : evict_bytes: config.evict_bytes,
2397 0 : freed_bytes: 0,
2398 0 : };
2399 0 :
2400 0 : let state = get_state(&r);
2401 0 : let eviction_state = state.disk_usage_eviction_state.clone();
2402 :
2403 0 : let res = crate::disk_usage_eviction_task::disk_usage_eviction_task_iteration_impl(
2404 0 : &eviction_state,
2405 0 : &state.remote_storage,
2406 0 : usage,
2407 0 : &state.tenant_manager,
2408 0 : config.eviction_order.into(),
2409 0 : &cancel,
2410 0 : )
2411 0 : .await;
2412 :
2413 0 : info!(?res, "disk_usage_eviction_task_iteration_impl finished");
2414 :
2415 0 : let res = res.map_err(ApiError::InternalServerError)?;
2416 :
2417 0 : json_response(StatusCode::OK, res)
2418 0 : }
2419 :
2420 0 : async fn secondary_upload_handler(
2421 0 : request: Request<Body>,
2422 0 : _cancel: CancellationToken,
2423 0 : ) -> Result<Response<Body>, ApiError> {
2424 0 : let state = get_state(&request);
2425 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2426 0 : state
2427 0 : .secondary_controller
2428 0 : .upload_tenant(tenant_shard_id)
2429 0 : .await
2430 0 : .map_err(ApiError::InternalServerError)?;
2431 :
2432 0 : json_response(StatusCode::OK, ())
2433 0 : }
2434 :
2435 0 : async fn tenant_scan_remote_handler(
2436 0 : request: Request<Body>,
2437 0 : cancel: CancellationToken,
2438 0 : ) -> Result<Response<Body>, ApiError> {
2439 0 : let state = get_state(&request);
2440 0 : let tenant_id: TenantId = parse_request_param(&request, "tenant_id")?;
2441 :
2442 0 : let mut response = TenantScanRemoteStorageResponse::default();
2443 :
2444 0 : let (shards, _other_keys) =
2445 0 : list_remote_tenant_shards(&state.remote_storage, tenant_id, cancel.clone())
2446 0 : .await
2447 0 : .map_err(|e| ApiError::InternalServerError(anyhow::anyhow!(e)))?;
2448 :
2449 0 : for tenant_shard_id in shards {
2450 0 : let (timeline_ids, _other_keys) =
2451 0 : list_remote_timelines(&state.remote_storage, tenant_shard_id, cancel.clone())
2452 0 : .await
2453 0 : .map_err(|e| ApiError::InternalServerError(anyhow::anyhow!(e)))?;
2454 :
2455 0 : let mut generation = Generation::none();
2456 0 : for timeline_id in timeline_ids {
2457 0 : match download_index_part(
2458 0 : &state.remote_storage,
2459 0 : &tenant_shard_id,
2460 0 : &timeline_id,
2461 0 : Generation::MAX,
2462 0 : &cancel,
2463 0 : )
2464 0 : .instrument(info_span!("download_index_part",
2465 : tenant_id=%tenant_shard_id.tenant_id,
2466 0 : shard_id=%tenant_shard_id.shard_slug(),
2467 : %timeline_id))
2468 0 : .await
2469 : {
2470 0 : Ok((index_part, index_generation, _index_mtime)) => {
2471 0 : tracing::info!("Found timeline {tenant_shard_id}/{timeline_id} metadata (gen {index_generation:?}, {} layers, {} consistent LSN)",
2472 0 : index_part.layer_metadata.len(), index_part.metadata.disk_consistent_lsn());
2473 0 : generation = std::cmp::max(generation, index_generation);
2474 : }
2475 : Err(DownloadError::NotFound) => {
2476 : // This is normal for tenants that were created with multiple shards: they have an unsharded path
2477 : // containing the timeline's initdb tarball but no index. Otherwise it is a bit strange.
2478 0 : tracing::info!("Timeline path {tenant_shard_id}/{timeline_id} exists in remote storage but has no index, skipping");
2479 0 : continue;
2480 : }
2481 0 : Err(e) => {
2482 0 : return Err(ApiError::InternalServerError(anyhow::anyhow!(e)));
2483 : }
2484 : };
2485 : }
2486 :
2487 0 : response.shards.push(TenantScanRemoteStorageShard {
2488 0 : tenant_shard_id,
2489 0 : generation: generation.into(),
2490 0 : });
2491 : }
2492 :
2493 0 : if response.shards.is_empty() {
2494 0 : return Err(ApiError::NotFound(
2495 0 : anyhow::anyhow!("No shards found for tenant ID {tenant_id}").into(),
2496 0 : ));
2497 0 : }
2498 0 :
2499 0 : json_response(StatusCode::OK, response)
2500 0 : }
2501 :
2502 0 : async fn secondary_download_handler(
2503 0 : request: Request<Body>,
2504 0 : _cancel: CancellationToken,
2505 0 : ) -> Result<Response<Body>, ApiError> {
2506 0 : let state = get_state(&request);
2507 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2508 0 : let wait = parse_query_param(&request, "wait_ms")?.map(Duration::from_millis);
2509 :
2510 : // We don't need this to issue the download request, but:
2511 : // - it enables us to cleanly return 404 if we get a request for an absent shard
2512 : // - we will use this to provide status feedback in the response
2513 0 : let Some(secondary_tenant) = state
2514 0 : .tenant_manager
2515 0 : .get_secondary_tenant_shard(tenant_shard_id)
2516 : else {
2517 0 : return Err(ApiError::NotFound(
2518 0 : anyhow::anyhow!("Shard {} not found", tenant_shard_id).into(),
2519 0 : ));
2520 : };
2521 :
2522 0 : let timeout = wait.unwrap_or(Duration::MAX);
2523 :
2524 0 : let result = tokio::time::timeout(
2525 0 : timeout,
2526 0 : state.secondary_controller.download_tenant(tenant_shard_id),
2527 0 : )
2528 0 : .await;
2529 :
2530 0 : let progress = secondary_tenant.progress.lock().unwrap().clone();
2531 :
2532 0 : let status = match result {
2533 : Ok(Ok(())) => {
2534 0 : if progress.layers_downloaded >= progress.layers_total {
2535 : // Download job ran to completion
2536 0 : StatusCode::OK
2537 : } else {
2538 : // Download dropped out without errors because it ran out of time budget
2539 0 : StatusCode::ACCEPTED
2540 : }
2541 : }
2542 : // Edge case: downloads aren't usually fallible: things like a missing heatmap are considered
2543 : // okay. We could get an error here in the unlikely edge case that the tenant
2544 : // was detached between our check above and executing the download job.
2545 0 : Ok(Err(e)) => return Err(ApiError::InternalServerError(e)),
2546 : // A timeout is not an error: we have started the download, we're just not done
2547 : // yet. The caller will get a response body indicating status.
2548 0 : Err(_) => StatusCode::ACCEPTED,
2549 : };
2550 :
2551 0 : json_response(status, progress)
2552 0 : }
2553 :
2554 0 : async fn secondary_status_handler(
2555 0 : request: Request<Body>,
2556 0 : _cancel: CancellationToken,
2557 0 : ) -> Result<Response<Body>, ApiError> {
2558 0 : let state = get_state(&request);
2559 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2560 :
2561 0 : let Some(secondary_tenant) = state
2562 0 : .tenant_manager
2563 0 : .get_secondary_tenant_shard(tenant_shard_id)
2564 : else {
2565 0 : return Err(ApiError::NotFound(
2566 0 : anyhow::anyhow!("Shard {} not found", tenant_shard_id).into(),
2567 0 : ));
2568 : };
2569 :
2570 0 : let progress = secondary_tenant.progress.lock().unwrap().clone();
2571 0 :
2572 0 : json_response(StatusCode::OK, progress)
2573 0 : }
2574 :
2575 0 : async fn handler_404(_: Request<Body>) -> Result<Response<Body>, ApiError> {
2576 0 : json_response(
2577 0 : StatusCode::NOT_FOUND,
2578 0 : HttpErrorBody::from_msg("page not found".to_owned()),
2579 0 : )
2580 0 : }
2581 :
2582 0 : async fn post_tracing_event_handler(
2583 0 : mut r: Request<Body>,
2584 0 : _cancel: CancellationToken,
2585 0 : ) -> Result<Response<Body>, ApiError> {
2586 0 : #[derive(Debug, serde::Deserialize)]
2587 : #[serde(rename_all = "lowercase")]
2588 : enum Level {
2589 : Error,
2590 : Warn,
2591 : Info,
2592 : Debug,
2593 : Trace,
2594 : }
2595 0 : #[derive(Debug, serde::Deserialize)]
2596 : struct Request {
2597 : level: Level,
2598 : message: String,
2599 : }
2600 0 : let body: Request = json_request(&mut r)
2601 0 : .await
2602 0 : .map_err(|_| ApiError::BadRequest(anyhow::anyhow!("invalid JSON body")))?;
2603 :
2604 0 : match body.level {
2605 0 : Level::Error => tracing::error!(?body.message),
2606 0 : Level::Warn => tracing::warn!(?body.message),
2607 0 : Level::Info => tracing::info!(?body.message),
2608 0 : Level::Debug => tracing::debug!(?body.message),
2609 0 : Level::Trace => tracing::trace!(?body.message),
2610 : }
2611 :
2612 0 : json_response(StatusCode::OK, ())
2613 0 : }
2614 :
2615 0 : async fn put_io_engine_handler(
2616 0 : mut r: Request<Body>,
2617 0 : _cancel: CancellationToken,
2618 0 : ) -> Result<Response<Body>, ApiError> {
2619 0 : check_permission(&r, None)?;
2620 0 : let kind: crate::virtual_file::IoEngineKind = json_request(&mut r).await?;
2621 0 : crate::virtual_file::io_engine::set(kind);
2622 0 : json_response(StatusCode::OK, ())
2623 0 : }
2624 :
2625 0 : async fn put_io_mode_handler(
2626 0 : mut r: Request<Body>,
2627 0 : _cancel: CancellationToken,
2628 0 : ) -> Result<Response<Body>, ApiError> {
2629 0 : check_permission(&r, None)?;
2630 0 : let mode: IoMode = json_request(&mut r).await?;
2631 0 : crate::virtual_file::set_io_mode(mode);
2632 0 : json_response(StatusCode::OK, ())
2633 0 : }
2634 :
2635 : /// Polled by control plane.
2636 : ///
2637 : /// See [`crate::utilization`].
2638 0 : async fn get_utilization(
2639 0 : r: Request<Body>,
2640 0 : _cancel: CancellationToken,
2641 0 : ) -> Result<Response<Body>, ApiError> {
2642 0 : fail::fail_point!("get-utilization-http-handler", |_| {
2643 0 : Err(ApiError::ResourceUnavailable("failpoint".into()))
2644 0 : });
2645 :
2646 : // this probably could be completely public, but lets make that change later.
2647 0 : check_permission(&r, None)?;
2648 :
2649 0 : let state = get_state(&r);
2650 0 : let mut g = state.latest_utilization.lock().await;
2651 :
2652 0 : let regenerate_every = Duration::from_secs(1);
2653 0 : let still_valid = g
2654 0 : .as_ref()
2655 0 : .is_some_and(|(captured_at, _)| captured_at.elapsed() < regenerate_every);
2656 0 :
2657 0 : // avoid needless statvfs calls even though those should be non-blocking fast.
2658 0 : // regenerate at most 1Hz to allow polling at any rate.
2659 0 : if !still_valid {
2660 0 : let path = state.conf.tenants_path();
2661 0 : let doc =
2662 0 : crate::utilization::regenerate(state.conf, path.as_std_path(), &state.tenant_manager)
2663 0 : .map_err(ApiError::InternalServerError)?;
2664 :
2665 0 : let mut buf = Vec::new();
2666 0 : serde_json::to_writer(&mut buf, &doc)
2667 0 : .context("serialize")
2668 0 : .map_err(ApiError::InternalServerError)?;
2669 :
2670 0 : let body = bytes::Bytes::from(buf);
2671 0 :
2672 0 : *g = Some((std::time::Instant::now(), body));
2673 0 : }
2674 :
2675 : // hyper 0.14 doesn't yet have Response::clone so this is a bit of extra legwork
2676 0 : let cached = g.as_ref().expect("just set").1.clone();
2677 0 :
2678 0 : Response::builder()
2679 0 : .header(hyper::http::header::CONTENT_TYPE, "application/json")
2680 0 : // thought of using http date header, but that is second precision which does not give any
2681 0 : // debugging aid
2682 0 : .status(StatusCode::OK)
2683 0 : .body(hyper::Body::from(cached))
2684 0 : .context("build response")
2685 0 : .map_err(ApiError::InternalServerError)
2686 0 : }
2687 :
2688 0 : async fn list_aux_files(
2689 0 : mut request: Request<Body>,
2690 0 : _cancel: CancellationToken,
2691 0 : ) -> Result<Response<Body>, ApiError> {
2692 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2693 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2694 0 : let body: ListAuxFilesRequest = json_request(&mut request).await?;
2695 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2696 :
2697 0 : let state = get_state(&request);
2698 :
2699 0 : let timeline =
2700 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
2701 0 : .await?;
2702 :
2703 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
2704 0 : let files = timeline.list_aux_files(body.lsn, &ctx).await?;
2705 0 : json_response(StatusCode::OK, files)
2706 0 : }
2707 :
2708 0 : async fn perf_info(
2709 0 : request: Request<Body>,
2710 0 : _cancel: CancellationToken,
2711 0 : ) -> Result<Response<Body>, ApiError> {
2712 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2713 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2714 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2715 :
2716 0 : let state = get_state(&request);
2717 :
2718 0 : let timeline =
2719 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
2720 0 : .await?;
2721 :
2722 0 : let result = timeline.perf_info().await;
2723 :
2724 0 : json_response(StatusCode::OK, result)
2725 0 : }
2726 :
2727 0 : async fn ingest_aux_files(
2728 0 : mut request: Request<Body>,
2729 0 : _cancel: CancellationToken,
2730 0 : ) -> Result<Response<Body>, ApiError> {
2731 0 : let tenant_shard_id: TenantShardId = parse_request_param(&request, "tenant_shard_id")?;
2732 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2733 0 : let body: IngestAuxFilesRequest = json_request(&mut request).await?;
2734 0 : check_permission(&request, Some(tenant_shard_id.tenant_id))?;
2735 :
2736 0 : let state = get_state(&request);
2737 :
2738 0 : let timeline =
2739 0 : active_timeline_of_active_tenant(&state.tenant_manager, tenant_shard_id, timeline_id)
2740 0 : .await?;
2741 :
2742 0 : let mut modification = timeline.begin_modification(
2743 0 : Lsn(timeline.get_last_record_lsn().0 + 8), /* advance LSN by 8 */
2744 0 : );
2745 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
2746 0 : for (fname, content) in body.aux_files {
2747 0 : modification
2748 0 : .put_file(&fname, content.as_bytes(), &ctx)
2749 0 : .await
2750 0 : .map_err(ApiError::InternalServerError)?;
2751 : }
2752 0 : modification
2753 0 : .commit(&ctx)
2754 0 : .await
2755 0 : .map_err(ApiError::InternalServerError)?;
2756 :
2757 0 : json_response(StatusCode::OK, ())
2758 0 : }
2759 :
2760 : /// Report on the largest tenants on this pageserver, for the storage controller to identify
2761 : /// candidates for splitting
2762 0 : async fn post_top_tenants(
2763 0 : mut r: Request<Body>,
2764 0 : _cancel: CancellationToken,
2765 0 : ) -> Result<Response<Body>, ApiError> {
2766 0 : check_permission(&r, None)?;
2767 0 : let request: TopTenantShardsRequest = json_request(&mut r).await?;
2768 0 : let state = get_state(&r);
2769 :
2770 0 : fn get_size_metric(sizes: &TopTenantShardItem, order_by: &TenantSorting) -> u64 {
2771 0 : match order_by {
2772 0 : TenantSorting::ResidentSize => sizes.resident_size,
2773 0 : TenantSorting::MaxLogicalSize => sizes.max_logical_size,
2774 : }
2775 0 : }
2776 :
2777 : #[derive(Eq, PartialEq)]
2778 : struct HeapItem {
2779 : metric: u64,
2780 : sizes: TopTenantShardItem,
2781 : }
2782 :
2783 : impl PartialOrd for HeapItem {
2784 0 : fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
2785 0 : Some(self.cmp(other))
2786 0 : }
2787 : }
2788 :
2789 : /// Heap items have reverse ordering on their metric: this enables using BinaryHeap, which
2790 : /// supports popping the greatest item but not the smallest.
2791 : impl Ord for HeapItem {
2792 0 : fn cmp(&self, other: &Self) -> std::cmp::Ordering {
2793 0 : Reverse(self.metric).cmp(&Reverse(other.metric))
2794 0 : }
2795 : }
2796 :
2797 0 : let mut top_n: BinaryHeap<HeapItem> = BinaryHeap::with_capacity(request.limit);
2798 :
2799 : // FIXME: this is a lot of clones to take this tenant list
2800 0 : for (tenant_shard_id, tenant_slot) in state.tenant_manager.list() {
2801 0 : if let Some(shards_lt) = request.where_shards_lt {
2802 : // Ignore tenants which already have >= this many shards
2803 0 : if tenant_shard_id.shard_count >= shards_lt {
2804 0 : continue;
2805 0 : }
2806 0 : }
2807 :
2808 0 : let sizes = match tenant_slot {
2809 0 : TenantSlot::Attached(tenant) => tenant.get_sizes(),
2810 : TenantSlot::Secondary(_) | TenantSlot::InProgress(_) => {
2811 0 : continue;
2812 : }
2813 : };
2814 0 : let metric = get_size_metric(&sizes, &request.order_by);
2815 :
2816 0 : if let Some(gt) = request.where_gt {
2817 : // Ignore tenants whose metric is <= the lower size threshold, to do less sorting work
2818 0 : if metric <= gt {
2819 0 : continue;
2820 0 : }
2821 0 : };
2822 :
2823 0 : match top_n.peek() {
2824 0 : None => {
2825 0 : // Top N list is empty: candidate becomes first member
2826 0 : top_n.push(HeapItem { metric, sizes });
2827 0 : }
2828 0 : Some(i) if i.metric > metric && top_n.len() < request.limit => {
2829 0 : // Lowest item in list is greater than our candidate, but we aren't at limit yet: push to end
2830 0 : top_n.push(HeapItem { metric, sizes });
2831 0 : }
2832 0 : Some(i) if i.metric > metric => {
2833 0 : // List is at limit and lowest value is greater than our candidate, drop it.
2834 0 : }
2835 0 : Some(_) => top_n.push(HeapItem { metric, sizes }),
2836 : }
2837 :
2838 0 : while top_n.len() > request.limit {
2839 0 : top_n.pop();
2840 0 : }
2841 : }
2842 :
2843 0 : json_response(
2844 0 : StatusCode::OK,
2845 0 : TopTenantShardsResponse {
2846 0 : shards: top_n.into_iter().map(|i| i.sizes).collect(),
2847 0 : },
2848 0 : )
2849 0 : }
2850 :
2851 0 : async fn put_tenant_timeline_import_basebackup(
2852 0 : request: Request<Body>,
2853 0 : _cancel: CancellationToken,
2854 0 : ) -> Result<Response<Body>, ApiError> {
2855 0 : let tenant_id: TenantId = parse_request_param(&request, "tenant_id")?;
2856 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2857 0 : let base_lsn: Lsn = must_parse_query_param(&request, "base_lsn")?;
2858 0 : let end_lsn: Lsn = must_parse_query_param(&request, "end_lsn")?;
2859 0 : let pg_version: u32 = must_parse_query_param(&request, "pg_version")?;
2860 :
2861 0 : check_permission(&request, Some(tenant_id))?;
2862 :
2863 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
2864 :
2865 0 : let span = info_span!("import_basebackup", tenant_id=%tenant_id, timeline_id=%timeline_id, base_lsn=%base_lsn, end_lsn=%end_lsn, pg_version=%pg_version);
2866 0 : async move {
2867 0 : let state = get_state(&request);
2868 0 : let tenant = state
2869 0 : .tenant_manager
2870 0 : .get_attached_tenant_shard(TenantShardId::unsharded(tenant_id))?;
2871 :
2872 0 : let broker_client = state.broker_client.clone();
2873 0 :
2874 0 : let mut body = StreamReader::new(request.into_body().map(|res| {
2875 0 : res.map_err(|error| {
2876 0 : std::io::Error::new(std::io::ErrorKind::Other, anyhow::anyhow!(error))
2877 0 : })
2878 0 : }));
2879 0 :
2880 0 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
2881 :
2882 0 : let timeline = tenant
2883 0 : .create_empty_timeline(timeline_id, base_lsn, pg_version, &ctx)
2884 0 : .map_err(ApiError::InternalServerError)
2885 0 : .await?;
2886 :
2887 : // TODO mark timeline as not ready until it reaches end_lsn.
2888 : // We might have some wal to import as well, and we should prevent compute
2889 : // from connecting before that and writing conflicting wal.
2890 : //
2891 : // This is not relevant for pageserver->pageserver migrations, since there's
2892 : // no wal to import. But should be fixed if we want to import from postgres.
2893 :
2894 : // TODO leave clean state on error. For now you can use detach to clean
2895 : // up broken state from a failed import.
2896 :
2897 : // Import basebackup provided via CopyData
2898 0 : info!("importing basebackup");
2899 :
2900 0 : timeline
2901 0 : .import_basebackup_from_tar(tenant.clone(), &mut body, base_lsn, broker_client, &ctx)
2902 0 : .await
2903 0 : .map_err(ApiError::InternalServerError)?;
2904 :
2905 : // Read the end of the tar archive.
2906 0 : read_tar_eof(body)
2907 0 : .await
2908 0 : .map_err(ApiError::InternalServerError)?;
2909 :
2910 : // TODO check checksum
2911 : // Meanwhile you can verify client-side by taking fullbackup
2912 : // and checking that it matches in size with what was imported.
2913 : // It wouldn't work if base came from vanilla postgres though,
2914 : // since we discard some log files.
2915 :
2916 0 : info!("done");
2917 0 : json_response(StatusCode::OK, ())
2918 0 : }
2919 0 : .instrument(span)
2920 0 : .await
2921 0 : }
2922 :
2923 0 : async fn put_tenant_timeline_import_wal(
2924 0 : request: Request<Body>,
2925 0 : _cancel: CancellationToken,
2926 0 : ) -> Result<Response<Body>, ApiError> {
2927 0 : let tenant_id: TenantId = parse_request_param(&request, "tenant_id")?;
2928 0 : let timeline_id: TimelineId = parse_request_param(&request, "timeline_id")?;
2929 0 : let start_lsn: Lsn = must_parse_query_param(&request, "start_lsn")?;
2930 0 : let end_lsn: Lsn = must_parse_query_param(&request, "end_lsn")?;
2931 :
2932 0 : check_permission(&request, Some(tenant_id))?;
2933 :
2934 0 : let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Warn);
2935 :
2936 0 : let span = info_span!("import_wal", tenant_id=%tenant_id, timeline_id=%timeline_id, start_lsn=%start_lsn, end_lsn=%end_lsn);
2937 0 : async move {
2938 0 : let state = get_state(&request);
2939 :
2940 0 : let timeline = active_timeline_of_active_tenant(&state.tenant_manager, TenantShardId::unsharded(tenant_id), timeline_id).await?;
2941 :
2942 0 : let mut body = StreamReader::new(request.into_body().map(|res| {
2943 0 : res.map_err(|error| {
2944 0 : std::io::Error::new(std::io::ErrorKind::Other, anyhow::anyhow!(error))
2945 0 : })
2946 0 : }));
2947 0 :
2948 0 : let last_record_lsn = timeline.get_last_record_lsn();
2949 0 : if last_record_lsn != start_lsn {
2950 0 : return Err(ApiError::InternalServerError(anyhow::anyhow!("Cannot import WAL from Lsn {start_lsn} because timeline does not start from the same lsn: {last_record_lsn}")));
2951 0 : }
2952 0 :
2953 0 : // TODO leave clean state on error. For now you can use detach to clean
2954 0 : // up broken state from a failed import.
2955 0 :
2956 0 : // Import wal provided via CopyData
2957 0 : info!("importing wal");
2958 0 : crate::import_datadir::import_wal_from_tar(&timeline, &mut body, start_lsn, end_lsn, &ctx).await.map_err(ApiError::InternalServerError)?;
2959 0 : info!("wal import complete");
2960 :
2961 : // Read the end of the tar archive.
2962 0 : read_tar_eof(body).await.map_err(ApiError::InternalServerError)?;
2963 :
2964 : // TODO Does it make sense to overshoot?
2965 0 : if timeline.get_last_record_lsn() < end_lsn {
2966 0 : return Err(ApiError::InternalServerError(anyhow::anyhow!("Cannot import WAL from Lsn {start_lsn} because timeline does not start from the same lsn: {last_record_lsn}")));
2967 0 : }
2968 0 :
2969 0 : // Flush data to disk, then upload to s3. No need for a forced checkpoint.
2970 0 : // We only want to persist the data, and it doesn't matter if it's in the
2971 0 : // shape of deltas or images.
2972 0 : info!("flushing layers");
2973 0 : timeline.freeze_and_flush().await.map_err(|e| match e {
2974 0 : tenant::timeline::FlushLayerError::Cancelled => ApiError::ShuttingDown,
2975 0 : other => ApiError::InternalServerError(anyhow::anyhow!(other)),
2976 0 : })?;
2977 :
2978 0 : info!("done");
2979 :
2980 0 : json_response(StatusCode::OK, ())
2981 0 : }.instrument(span).await
2982 0 : }
2983 :
2984 : /// Read the end of a tar archive.
2985 : ///
2986 : /// A tar archive normally ends with two consecutive blocks of zeros, 512 bytes each.
2987 : /// `tokio_tar` already read the first such block. Read the second all-zeros block,
2988 : /// and check that there is no more data after the EOF marker.
2989 : ///
2990 : /// 'tar' command can also write extra blocks of zeros, up to a record
2991 : /// size, controlled by the --record-size argument. Ignore them too.
2992 0 : async fn read_tar_eof(mut reader: (impl tokio::io::AsyncRead + Unpin)) -> anyhow::Result<()> {
2993 : use tokio::io::AsyncReadExt;
2994 0 : let mut buf = [0u8; 512];
2995 0 :
2996 0 : // Read the all-zeros block, and verify it
2997 0 : let mut total_bytes = 0;
2998 0 : while total_bytes < 512 {
2999 0 : let nbytes = reader.read(&mut buf[total_bytes..]).await?;
3000 0 : total_bytes += nbytes;
3001 0 : if nbytes == 0 {
3002 0 : break;
3003 0 : }
3004 : }
3005 0 : if total_bytes < 512 {
3006 0 : anyhow::bail!("incomplete or invalid tar EOF marker");
3007 0 : }
3008 0 : if !buf.iter().all(|&x| x == 0) {
3009 0 : anyhow::bail!("invalid tar EOF marker");
3010 0 : }
3011 0 :
3012 0 : // Drain any extra zero-blocks after the EOF marker
3013 0 : let mut trailing_bytes = 0;
3014 0 : let mut seen_nonzero_bytes = false;
3015 : loop {
3016 0 : let nbytes = reader.read(&mut buf).await?;
3017 0 : trailing_bytes += nbytes;
3018 0 : if !buf.iter().all(|&x| x == 0) {
3019 0 : seen_nonzero_bytes = true;
3020 0 : }
3021 0 : if nbytes == 0 {
3022 0 : break;
3023 0 : }
3024 : }
3025 0 : if seen_nonzero_bytes {
3026 0 : anyhow::bail!("unexpected non-zero bytes after the tar archive");
3027 0 : }
3028 0 : if trailing_bytes % 512 != 0 {
3029 0 : anyhow::bail!("unexpected number of zeros ({trailing_bytes}), not divisible by tar block size (512 bytes), after the tar archive");
3030 0 : }
3031 0 : Ok(())
3032 0 : }
3033 :
3034 : /// Common functionality of all the HTTP API handlers.
3035 : ///
3036 : /// - Adds a tracing span to each request (by `request_span`)
3037 : /// - Logs the request depending on the request method (by `request_span`)
3038 : /// - Logs the response if it was not successful (by `request_span`
3039 : /// - Shields the handler function from async cancellations. Hyper can drop the handler
3040 : /// Future if the connection to the client is lost, but most of the pageserver code is
3041 : /// not async cancellation safe. This converts the dropped future into a graceful cancellation
3042 : /// request with a CancellationToken.
3043 0 : async fn api_handler<R, H>(request: Request<Body>, handler: H) -> Result<Response<Body>, ApiError>
3044 0 : where
3045 0 : R: std::future::Future<Output = Result<Response<Body>, ApiError>> + Send + 'static,
3046 0 : H: FnOnce(Request<Body>, CancellationToken) -> R + Send + Sync + 'static,
3047 0 : {
3048 0 : if request.uri() != &"/v1/failpoints".parse::<Uri>().unwrap() {
3049 0 : fail::fail_point!("api-503", |_| Err(ApiError::ResourceUnavailable(
3050 0 : "failpoint".into()
3051 0 : )));
3052 :
3053 0 : fail::fail_point!("api-500", |_| Err(ApiError::InternalServerError(
3054 0 : anyhow::anyhow!("failpoint")
3055 0 : )));
3056 0 : }
3057 :
3058 : // Spawn a new task to handle the request, to protect the handler from unexpected
3059 : // async cancellations. Most pageserver functions are not async cancellation safe.
3060 : // We arm a drop-guard, so that if Hyper drops the Future, we signal the task
3061 : // with the cancellation token.
3062 0 : let token = CancellationToken::new();
3063 0 : let cancel_guard = token.clone().drop_guard();
3064 0 : let result = request_span(request, move |r| async {
3065 0 : let handle = tokio::spawn(
3066 0 : async {
3067 0 : let token_cloned = token.clone();
3068 0 : let result = handler(r, token).await;
3069 0 : if token_cloned.is_cancelled() {
3070 : // dropguard has executed: we will never turn this result into response.
3071 : //
3072 : // at least temporarily do {:?} logging; these failures are rare enough but
3073 : // could hide difficult errors.
3074 0 : match &result {
3075 0 : Ok(response) => {
3076 0 : let status = response.status();
3077 0 : info!(%status, "Cancelled request finished successfully")
3078 : }
3079 0 : Err(e) => error!("Cancelled request finished with an error: {e:?}"),
3080 : }
3081 0 : }
3082 : // only logging for cancelled panicked request handlers is the tracing_panic_hook,
3083 : // which should suffice.
3084 : //
3085 : // there is still a chance to lose the result due to race between
3086 : // returning from here and the actual connection closing happening
3087 : // before outer task gets to execute. leaving that up for #5815.
3088 0 : result
3089 0 : }
3090 0 : .in_current_span(),
3091 0 : );
3092 0 :
3093 0 : match handle.await {
3094 : // TODO: never actually return Err from here, always Ok(...) so that we can log
3095 : // spanned errors. Call api_error_handler instead and return appropriate Body.
3096 0 : Ok(result) => result,
3097 0 : Err(e) => {
3098 0 : // The handler task panicked. We have a global panic handler that logs the
3099 0 : // panic with its backtrace, so no need to log that here. Only log a brief
3100 0 : // message to make it clear that we returned the error to the client.
3101 0 : error!("HTTP request handler task panicked: {e:#}");
3102 :
3103 : // Don't return an Error here, because then fallback error handler that was
3104 : // installed in make_router() will print the error. Instead, construct the
3105 : // HTTP error response and return that.
3106 0 : Ok(
3107 0 : ApiError::InternalServerError(anyhow!("HTTP request handler task panicked"))
3108 0 : .into_response(),
3109 0 : )
3110 : }
3111 : }
3112 0 : })
3113 0 : .await;
3114 :
3115 0 : cancel_guard.disarm();
3116 0 :
3117 0 : result
3118 0 : }
3119 :
3120 : /// Like api_handler, but returns an error response if the server is built without
3121 : /// the 'testing' feature.
3122 0 : async fn testing_api_handler<R, H>(
3123 0 : desc: &str,
3124 0 : request: Request<Body>,
3125 0 : handler: H,
3126 0 : ) -> Result<Response<Body>, ApiError>
3127 0 : where
3128 0 : R: std::future::Future<Output = Result<Response<Body>, ApiError>> + Send + 'static,
3129 0 : H: FnOnce(Request<Body>, CancellationToken) -> R + Send + Sync + 'static,
3130 0 : {
3131 0 : if cfg!(feature = "testing") {
3132 0 : api_handler(request, handler).await
3133 : } else {
3134 0 : std::future::ready(Err(ApiError::BadRequest(anyhow!(
3135 0 : "Cannot {desc} because pageserver was compiled without testing APIs",
3136 0 : ))))
3137 0 : .await
3138 : }
3139 0 : }
3140 :
3141 0 : pub fn make_router(
3142 0 : state: Arc<State>,
3143 0 : launch_ts: &'static LaunchTimestamp,
3144 0 : auth: Option<Arc<SwappableJwtAuth>>,
3145 0 : ) -> anyhow::Result<RouterBuilder<hyper::Body, ApiError>> {
3146 0 : let spec = include_bytes!("openapi_spec.yml");
3147 0 : let mut router = attach_openapi_ui(endpoint::make_router(), spec, "/swagger.yml", "/v1/doc");
3148 0 : if auth.is_some() {
3149 0 : router = router.middleware(auth_middleware(|request| {
3150 0 : let state = get_state(request);
3151 0 : if state.allowlist_routes.contains(request.uri()) {
3152 0 : None
3153 : } else {
3154 0 : state.auth.as_deref()
3155 : }
3156 0 : }))
3157 0 : }
3158 :
3159 0 : router = router.middleware(
3160 0 : endpoint::add_response_header_middleware(
3161 0 : "PAGESERVER_LAUNCH_TIMESTAMP",
3162 0 : &launch_ts.to_string(),
3163 0 : )
3164 0 : .expect("construct launch timestamp header middleware"),
3165 0 : );
3166 0 :
3167 0 : Ok(router
3168 0 : .data(state)
3169 0 : .get("/metrics", |r| request_span(r, prometheus_metrics_handler))
3170 0 : .get("/v1/status", |r| api_handler(r, status_handler))
3171 0 : .put("/v1/failpoints", |r| {
3172 0 : testing_api_handler("manage failpoints", r, failpoints_handler)
3173 0 : })
3174 0 : .post("/v1/reload_auth_validation_keys", |r| {
3175 0 : api_handler(r, reload_auth_validation_keys_handler)
3176 0 : })
3177 0 : .get("/v1/tenant", |r| api_handler(r, tenant_list_handler))
3178 0 : .get("/v1/tenant/:tenant_shard_id", |r| {
3179 0 : api_handler(r, tenant_status)
3180 0 : })
3181 0 : .delete("/v1/tenant/:tenant_shard_id", |r| {
3182 0 : api_handler(r, tenant_delete_handler)
3183 0 : })
3184 0 : .get("/v1/tenant/:tenant_shard_id/synthetic_size", |r| {
3185 0 : api_handler(r, tenant_size_handler)
3186 0 : })
3187 0 : .put("/v1/tenant/config", |r| {
3188 0 : api_handler(r, update_tenant_config_handler)
3189 0 : })
3190 0 : .put("/v1/tenant/:tenant_shard_id/shard_split", |r| {
3191 0 : api_handler(r, tenant_shard_split_handler)
3192 0 : })
3193 0 : .get("/v1/tenant/:tenant_shard_id/config", |r| {
3194 0 : api_handler(r, get_tenant_config_handler)
3195 0 : })
3196 0 : .put("/v1/tenant/:tenant_shard_id/location_config", |r| {
3197 0 : api_handler(r, put_tenant_location_config_handler)
3198 0 : })
3199 0 : .get("/v1/location_config", |r| {
3200 0 : api_handler(r, list_location_config_handler)
3201 0 : })
3202 0 : .get("/v1/location_config/:tenant_shard_id", |r| {
3203 0 : api_handler(r, get_location_config_handler)
3204 0 : })
3205 0 : .put(
3206 0 : "/v1/tenant/:tenant_shard_id/time_travel_remote_storage",
3207 0 : |r| api_handler(r, tenant_time_travel_remote_storage_handler),
3208 0 : )
3209 0 : .get("/v1/tenant/:tenant_shard_id/timeline", |r| {
3210 0 : api_handler(r, timeline_list_handler)
3211 0 : })
3212 0 : .get("/v1/tenant/:tenant_shard_id/timeline_and_offloaded", |r| {
3213 0 : api_handler(r, timeline_and_offloaded_list_handler)
3214 0 : })
3215 0 : .post("/v1/tenant/:tenant_shard_id/timeline", |r| {
3216 0 : api_handler(r, timeline_create_handler)
3217 0 : })
3218 0 : .post("/v1/tenant/:tenant_shard_id/reset", |r| {
3219 0 : api_handler(r, tenant_reset_handler)
3220 0 : })
3221 0 : .post(
3222 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/preserve_initdb_archive",
3223 0 : |r| api_handler(r, timeline_preserve_initdb_handler),
3224 0 : )
3225 0 : .put(
3226 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/archival_config",
3227 0 : |r| api_handler(r, timeline_archival_config_handler),
3228 0 : )
3229 0 : .get("/v1/tenant/:tenant_shard_id/timeline/:timeline_id", |r| {
3230 0 : api_handler(r, timeline_detail_handler)
3231 0 : })
3232 0 : .get(
3233 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/get_lsn_by_timestamp",
3234 0 : |r| api_handler(r, get_lsn_by_timestamp_handler),
3235 0 : )
3236 0 : .get(
3237 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/get_timestamp_of_lsn",
3238 0 : |r| api_handler(r, get_timestamp_of_lsn_handler),
3239 0 : )
3240 0 : .post(
3241 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/lsn_lease",
3242 0 : |r| api_handler(r, lsn_lease_handler),
3243 0 : )
3244 0 : .put(
3245 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/do_gc",
3246 0 : |r| api_handler(r, timeline_gc_handler),
3247 0 : )
3248 0 : .put(
3249 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/compact",
3250 0 : |r| api_handler(r, timeline_compact_handler),
3251 0 : )
3252 0 : .put(
3253 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/offload",
3254 0 : |r| testing_api_handler("attempt timeline offload", r, timeline_offload_handler),
3255 0 : )
3256 0 : .put(
3257 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/checkpoint",
3258 0 : |r| testing_api_handler("run timeline checkpoint", r, timeline_checkpoint_handler),
3259 0 : )
3260 0 : .post(
3261 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/download_remote_layers",
3262 0 : |r| api_handler(r, timeline_download_remote_layers_handler_post),
3263 0 : )
3264 0 : .get(
3265 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/download_remote_layers",
3266 0 : |r| api_handler(r, timeline_download_remote_layers_handler_get),
3267 0 : )
3268 0 : .put(
3269 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/detach_ancestor",
3270 0 : |r| api_handler(r, timeline_detach_ancestor_handler),
3271 0 : )
3272 0 : .delete("/v1/tenant/:tenant_shard_id/timeline/:timeline_id", |r| {
3273 0 : api_handler(r, timeline_delete_handler)
3274 0 : })
3275 0 : .get(
3276 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/layer",
3277 0 : |r| api_handler(r, layer_map_info_handler),
3278 0 : )
3279 0 : .get(
3280 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/layer/:layer_file_name",
3281 0 : |r| api_handler(r, layer_download_handler),
3282 0 : )
3283 0 : .delete(
3284 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/layer/:layer_file_name",
3285 0 : |r| api_handler(r, evict_timeline_layer_handler),
3286 0 : )
3287 0 : .post(
3288 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/layer/:layer_name/scan_disposable_keys",
3289 0 : |r| testing_api_handler("timeline_layer_scan_disposable_keys", r, timeline_layer_scan_disposable_keys),
3290 0 : )
3291 0 : .post(
3292 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/block_gc",
3293 0 : |r| api_handler(r, timeline_gc_blocking_handler),
3294 0 : )
3295 0 : .post(
3296 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/unblock_gc",
3297 0 : |r| api_handler(r, timeline_gc_unblocking_handler),
3298 0 : )
3299 0 : .post("/v1/tenant/:tenant_shard_id/heatmap_upload", |r| {
3300 0 : api_handler(r, secondary_upload_handler)
3301 0 : })
3302 0 : .get("/v1/tenant/:tenant_id/scan_remote_storage", |r| {
3303 0 : api_handler(r, tenant_scan_remote_handler)
3304 0 : })
3305 0 : .put("/v1/disk_usage_eviction/run", |r| {
3306 0 : api_handler(r, disk_usage_eviction_run)
3307 0 : })
3308 0 : .put("/v1/deletion_queue/flush", |r| {
3309 0 : api_handler(r, deletion_queue_flush)
3310 0 : })
3311 0 : .get("/v1/tenant/:tenant_shard_id/secondary/status", |r| {
3312 0 : api_handler(r, secondary_status_handler)
3313 0 : })
3314 0 : .post("/v1/tenant/:tenant_shard_id/secondary/download", |r| {
3315 0 : api_handler(r, secondary_download_handler)
3316 0 : })
3317 0 : .put("/v1/tenant/:tenant_shard_id/break", |r| {
3318 0 : testing_api_handler("set tenant state to broken", r, handle_tenant_break)
3319 0 : })
3320 0 : .get("/v1/panic", |r| api_handler(r, always_panic_handler))
3321 0 : .post("/v1/tracing/event", |r| {
3322 0 : testing_api_handler("emit a tracing event", r, post_tracing_event_handler)
3323 0 : })
3324 0 : .get(
3325 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/getpage",
3326 0 : |r| testing_api_handler("getpage@lsn", r, getpage_at_lsn_handler),
3327 0 : )
3328 0 : .get(
3329 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/keyspace",
3330 0 : |r| api_handler(r, timeline_collect_keyspace),
3331 0 : )
3332 0 : .put("/v1/io_engine", |r| api_handler(r, put_io_engine_handler))
3333 0 : .put("/v1/io_mode", |r| api_handler(r, put_io_mode_handler))
3334 0 : .get("/v1/utilization", |r| api_handler(r, get_utilization))
3335 0 : .post(
3336 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/ingest_aux_files",
3337 0 : |r| testing_api_handler("ingest_aux_files", r, ingest_aux_files),
3338 0 : )
3339 0 : .post(
3340 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/list_aux_files",
3341 0 : |r| testing_api_handler("list_aux_files", r, list_aux_files),
3342 0 : )
3343 0 : .post("/v1/top_tenants", |r| api_handler(r, post_top_tenants))
3344 0 : .post(
3345 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/perf_info",
3346 0 : |r| testing_api_handler("perf_info", r, perf_info),
3347 0 : )
3348 0 : .put(
3349 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/import_basebackup",
3350 0 : |r| api_handler(r, put_tenant_timeline_import_basebackup),
3351 0 : )
3352 0 : .put(
3353 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/import_wal",
3354 0 : |r| api_handler(r, put_tenant_timeline_import_wal),
3355 0 : )
3356 0 : .any(handler_404))
3357 0 : }
|
