Line data Source code
1 : use std::str::FromStr;
2 : use std::sync::{Arc, LazyLock};
3 : use std::time::{Duration, Instant};
4 :
5 : use anyhow::Context;
6 : use control_plane::storage_controller::{AttachHookRequest, InspectRequest};
7 : use futures::Future;
8 : use http_utils::endpoint::{
9 : self, auth_middleware, check_permission_with, profile_cpu_handler, profile_heap_handler,
10 : request_span,
11 : };
12 : use http_utils::error::ApiError;
13 : use http_utils::failpoints::failpoints_handler;
14 : use http_utils::json::{json_request, json_response};
15 : use http_utils::request::{must_get_query_param, parse_query_param, parse_request_param};
16 : use http_utils::{RequestExt, RouterBuilder};
17 : use hyper::header::CONTENT_TYPE;
18 : use hyper::{Body, Request, Response, StatusCode, Uri};
19 : use metrics::{BuildInfo, NeonMetrics};
20 : use pageserver_api::controller_api::{
21 : MetadataHealthListOutdatedRequest, MetadataHealthListOutdatedResponse,
22 : MetadataHealthListUnhealthyResponse, MetadataHealthUpdateRequest, MetadataHealthUpdateResponse,
23 : NodeAvailability, NodeConfigureRequest, NodeRegisterRequest, SafekeeperSchedulingPolicyRequest,
24 : ShardsPreferredAzsRequest, TenantCreateRequest, TenantPolicyRequest, TenantShardMigrateRequest,
25 : TimelineImportRequest,
26 : };
27 : use pageserver_api::models::{
28 : DetachBehavior, LsnLeaseRequest, TenantConfigPatchRequest, TenantConfigRequest,
29 : TenantLocationConfigRequest, TenantShardSplitRequest, TenantTimeTravelRequest,
30 : TimelineArchivalConfigRequest, TimelineCreateRequest,
31 : };
32 : use pageserver_api::shard::TenantShardId;
33 : use pageserver_api::upcall_api::{ReAttachRequest, ValidateRequest};
34 : use pageserver_client::{BlockUnblock, mgmt_api};
35 : use routerify::Middleware;
36 : use tokio_util::sync::CancellationToken;
37 : use tracing::warn;
38 : use utils::auth::{Scope, SwappableJwtAuth};
39 : use utils::id::{NodeId, TenantId, TimelineId};
40 :
41 : use crate::http;
42 : use crate::metrics::{
43 : HttpRequestLatencyLabelGroup, HttpRequestStatusLabelGroup, METRICS_REGISTRY,
44 : PageserverRequestLabelGroup,
45 : };
46 : use crate::persistence::SafekeeperUpsert;
47 : use crate::reconciler::ReconcileError;
48 : use crate::service::{LeadershipStatus, RECONCILE_TIMEOUT, STARTUP_RECONCILE_TIMEOUT, Service};
49 :
50 : /// State available to HTTP request handlers
51 : pub struct HttpState {
52 : service: Arc<crate::service::Service>,
53 : auth: Option<Arc<SwappableJwtAuth>>,
54 : rate_limiter: governor::DefaultKeyedRateLimiter<TenantId>,
55 : neon_metrics: NeonMetrics,
56 : allowlist_routes: &'static [&'static str],
57 : }
58 :
59 : impl HttpState {
60 0 : pub fn new(
61 0 : service: Arc<crate::service::Service>,
62 0 : auth: Option<Arc<SwappableJwtAuth>>,
63 0 : build_info: BuildInfo,
64 0 : ) -> Self {
65 0 : let quota = governor::Quota::per_second(service.get_config().tenant_rate_limit);
66 0 : Self {
67 0 : service,
68 0 : auth,
69 0 : rate_limiter: governor::RateLimiter::keyed(quota),
70 0 : neon_metrics: NeonMetrics::new(build_info),
71 0 : allowlist_routes: &[
72 0 : "/status",
73 0 : "/ready",
74 0 : "/metrics",
75 0 : "/profile/cpu",
76 0 : "/profile/heap",
77 0 : ],
78 0 : }
79 0 : }
80 : }
81 :
82 : #[inline(always)]
83 0 : fn get_state(request: &Request<Body>) -> &HttpState {
84 0 : request
85 0 : .data::<Arc<HttpState>>()
86 0 : .expect("unknown state type")
87 0 : .as_ref()
88 0 : }
89 :
90 : /// Rate limits tenant requests.
91 : ///
92 : /// TODO: this should be a request middleware, but requires us to extract the tenant ID from
93 : /// different URLs in a systematic way.
94 : ///
95 : /// TODO: consider returning a 429 response if these start piling up.
96 0 : async fn maybe_rate_limit(request: &Request<Body>, tenant_id: TenantId) {
97 0 : // Check if the tenant should be rate-limited.
98 0 : let rate_limiter = &get_state(request).rate_limiter;
99 0 : if rate_limiter.check_key(&tenant_id).is_ok() {
100 0 : return;
101 0 : }
102 0 :
103 0 : // Measure the rate limiting delay.
104 0 : let _timer = METRICS_REGISTRY
105 0 : .metrics_group
106 0 : .storage_controller_http_request_rate_limited
107 0 : .start_timer();
108 :
109 : // Log rate limited tenants once every 10 seconds.
110 : static LOG_RATE_LIMITER: LazyLock<governor::DefaultKeyedRateLimiter<TenantId>> =
111 0 : LazyLock::new(|| {
112 0 : let quota = governor::Quota::with_period(Duration::from_secs(10)).unwrap();
113 0 : governor::RateLimiter::keyed(quota)
114 0 : });
115 :
116 0 : if LOG_RATE_LIMITER.check_key(&tenant_id).is_ok() {
117 0 : warn!("tenant {tenant_id} is rate limited")
118 0 : }
119 :
120 : // Wait for quota.
121 0 : rate_limiter.until_key_ready(&tenant_id).await;
122 0 : }
123 :
124 : /// Pageserver calls into this on startup, to learn which tenants it should attach
125 0 : async fn handle_re_attach(req: Request<Body>) -> Result<Response<Body>, ApiError> {
126 0 : check_permissions(&req, Scope::GenerationsApi)?;
127 :
128 0 : let mut req = match maybe_forward(req).await {
129 0 : ForwardOutcome::Forwarded(res) => {
130 0 : return res;
131 : }
132 0 : ForwardOutcome::NotForwarded(req) => req,
133 : };
134 :
135 0 : let reattach_req = json_request::<ReAttachRequest>(&mut req).await?;
136 0 : let state = get_state(&req);
137 0 : json_response(StatusCode::OK, state.service.re_attach(reattach_req).await?)
138 0 : }
139 :
140 : /// Pageserver calls into this before doing deletions, to confirm that it still
141 : /// holds the latest generation for the tenants with deletions enqueued
142 0 : async fn handle_validate(req: Request<Body>) -> Result<Response<Body>, ApiError> {
143 0 : check_permissions(&req, Scope::GenerationsApi)?;
144 :
145 0 : let mut req = match maybe_forward(req).await {
146 0 : ForwardOutcome::Forwarded(res) => {
147 0 : return res;
148 : }
149 0 : ForwardOutcome::NotForwarded(req) => req,
150 : };
151 :
152 0 : let validate_req = json_request::<ValidateRequest>(&mut req).await?;
153 0 : let state = get_state(&req);
154 0 : json_response(StatusCode::OK, state.service.validate(validate_req).await?)
155 0 : }
156 :
157 : /// Call into this before attaching a tenant to a pageserver, to acquire a generation number
158 : /// (in the real control plane this is unnecessary, because the same program is managing
159 : /// generation numbers and doing attachments).
160 0 : async fn handle_attach_hook(req: Request<Body>) -> Result<Response<Body>, ApiError> {
161 0 : check_permissions(&req, Scope::Admin)?;
162 :
163 0 : let mut req = match maybe_forward(req).await {
164 0 : ForwardOutcome::Forwarded(res) => {
165 0 : return res;
166 : }
167 0 : ForwardOutcome::NotForwarded(req) => req,
168 : };
169 :
170 0 : let attach_req = json_request::<AttachHookRequest>(&mut req).await?;
171 0 : let state = get_state(&req);
172 0 :
173 0 : json_response(
174 0 : StatusCode::OK,
175 0 : state
176 0 : .service
177 0 : .attach_hook(attach_req)
178 0 : .await
179 0 : .map_err(ApiError::InternalServerError)?,
180 : )
181 0 : }
182 :
183 0 : async fn handle_inspect(req: Request<Body>) -> Result<Response<Body>, ApiError> {
184 0 : check_permissions(&req, Scope::Admin)?;
185 :
186 0 : let mut req = match maybe_forward(req).await {
187 0 : ForwardOutcome::Forwarded(res) => {
188 0 : return res;
189 : }
190 0 : ForwardOutcome::NotForwarded(req) => req,
191 : };
192 :
193 0 : let inspect_req = json_request::<InspectRequest>(&mut req).await?;
194 :
195 0 : let state = get_state(&req);
196 0 :
197 0 : json_response(StatusCode::OK, state.service.inspect(inspect_req))
198 0 : }
199 :
200 0 : async fn handle_tenant_create(
201 0 : service: Arc<Service>,
202 0 : req: Request<Body>,
203 0 : ) -> Result<Response<Body>, ApiError> {
204 0 : check_permissions(&req, Scope::PageServerApi)?;
205 :
206 0 : let mut req = match maybe_forward(req).await {
207 0 : ForwardOutcome::Forwarded(res) => {
208 0 : return res;
209 : }
210 0 : ForwardOutcome::NotForwarded(req) => req,
211 : };
212 :
213 0 : let create_req = json_request::<TenantCreateRequest>(&mut req).await?;
214 :
215 : json_response(
216 : StatusCode::CREATED,
217 0 : service.tenant_create(create_req).await?,
218 : )
219 0 : }
220 :
221 0 : async fn handle_tenant_location_config(
222 0 : service: Arc<Service>,
223 0 : req: Request<Body>,
224 0 : ) -> Result<Response<Body>, ApiError> {
225 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
226 0 : check_permissions(&req, Scope::PageServerApi)?;
227 :
228 0 : let mut req = match maybe_forward(req).await {
229 0 : ForwardOutcome::Forwarded(res) => {
230 0 : return res;
231 : }
232 0 : ForwardOutcome::NotForwarded(req) => req,
233 : };
234 :
235 0 : let config_req = json_request::<TenantLocationConfigRequest>(&mut req).await?;
236 : json_response(
237 : StatusCode::OK,
238 0 : service
239 0 : .tenant_location_config(tenant_shard_id, config_req)
240 0 : .await?,
241 : )
242 0 : }
243 :
244 0 : async fn handle_tenant_config_patch(
245 0 : service: Arc<Service>,
246 0 : req: Request<Body>,
247 0 : ) -> Result<Response<Body>, ApiError> {
248 0 : check_permissions(&req, Scope::PageServerApi)?;
249 :
250 0 : let mut req = match maybe_forward(req).await {
251 0 : ForwardOutcome::Forwarded(res) => {
252 0 : return res;
253 : }
254 0 : ForwardOutcome::NotForwarded(req) => req,
255 : };
256 :
257 0 : let config_req = json_request::<TenantConfigPatchRequest>(&mut req).await?;
258 :
259 : json_response(
260 : StatusCode::OK,
261 0 : service.tenant_config_patch(config_req).await?,
262 : )
263 0 : }
264 :
265 0 : async fn handle_tenant_config_set(
266 0 : service: Arc<Service>,
267 0 : req: Request<Body>,
268 0 : ) -> Result<Response<Body>, ApiError> {
269 0 : check_permissions(&req, Scope::PageServerApi)?;
270 :
271 0 : let mut req = match maybe_forward(req).await {
272 0 : ForwardOutcome::Forwarded(res) => {
273 0 : return res;
274 : }
275 0 : ForwardOutcome::NotForwarded(req) => req,
276 : };
277 :
278 0 : let config_req = json_request::<TenantConfigRequest>(&mut req).await?;
279 :
280 0 : json_response(StatusCode::OK, service.tenant_config_set(config_req).await?)
281 0 : }
282 :
283 0 : async fn handle_tenant_config_get(
284 0 : service: Arc<Service>,
285 0 : req: Request<Body>,
286 0 : ) -> Result<Response<Body>, ApiError> {
287 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
288 0 : check_permissions(&req, Scope::PageServerApi)?;
289 0 : maybe_rate_limit(&req, tenant_id).await;
290 :
291 0 : match maybe_forward(req).await {
292 0 : ForwardOutcome::Forwarded(res) => {
293 0 : return res;
294 : }
295 0 : ForwardOutcome::NotForwarded(_req) => {}
296 0 : };
297 0 :
298 0 : json_response(StatusCode::OK, service.tenant_config_get(tenant_id)?)
299 0 : }
300 :
301 0 : async fn handle_tenant_time_travel_remote_storage(
302 0 : service: Arc<Service>,
303 0 : req: Request<Body>,
304 0 : ) -> Result<Response<Body>, ApiError> {
305 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
306 0 : check_permissions(&req, Scope::PageServerApi)?;
307 0 : maybe_rate_limit(&req, tenant_id).await;
308 :
309 0 : let mut req = match maybe_forward(req).await {
310 0 : ForwardOutcome::Forwarded(res) => {
311 0 : return res;
312 : }
313 0 : ForwardOutcome::NotForwarded(req) => req,
314 : };
315 :
316 0 : let time_travel_req = json_request::<TenantTimeTravelRequest>(&mut req).await?;
317 :
318 0 : let timestamp_raw = must_get_query_param(&req, "travel_to")?;
319 0 : let _timestamp = humantime::parse_rfc3339(×tamp_raw).map_err(|_e| {
320 0 : ApiError::BadRequest(anyhow::anyhow!(
321 0 : "Invalid time for travel_to: {timestamp_raw:?}"
322 0 : ))
323 0 : })?;
324 :
325 0 : let done_if_after_raw = must_get_query_param(&req, "done_if_after")?;
326 0 : let _done_if_after = humantime::parse_rfc3339(&done_if_after_raw).map_err(|_e| {
327 0 : ApiError::BadRequest(anyhow::anyhow!(
328 0 : "Invalid time for done_if_after: {done_if_after_raw:?}"
329 0 : ))
330 0 : })?;
331 :
332 0 : service
333 0 : .tenant_time_travel_remote_storage(
334 0 : &time_travel_req,
335 0 : tenant_id,
336 0 : timestamp_raw,
337 0 : done_if_after_raw,
338 0 : )
339 0 : .await?;
340 0 : json_response(StatusCode::OK, ())
341 0 : }
342 :
343 0 : fn map_reqwest_hyper_status(status: reqwest::StatusCode) -> Result<hyper::StatusCode, ApiError> {
344 0 : hyper::StatusCode::from_u16(status.as_u16())
345 0 : .context("invalid status code")
346 0 : .map_err(ApiError::InternalServerError)
347 0 : }
348 :
349 0 : async fn handle_tenant_secondary_download(
350 0 : service: Arc<Service>,
351 0 : req: Request<Body>,
352 0 : ) -> Result<Response<Body>, ApiError> {
353 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
354 0 : let wait = parse_query_param(&req, "wait_ms")?.map(Duration::from_millis);
355 0 : maybe_rate_limit(&req, tenant_id).await;
356 :
357 0 : match maybe_forward(req).await {
358 0 : ForwardOutcome::Forwarded(res) => {
359 0 : return res;
360 : }
361 0 : ForwardOutcome::NotForwarded(_req) => {}
362 : };
363 :
364 0 : let (status, progress) = service.tenant_secondary_download(tenant_id, wait).await?;
365 0 : json_response(map_reqwest_hyper_status(status)?, progress)
366 0 : }
367 :
368 0 : async fn handle_tenant_delete(
369 0 : service: Arc<Service>,
370 0 : req: Request<Body>,
371 0 : ) -> Result<Response<Body>, ApiError> {
372 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
373 0 : check_permissions(&req, Scope::PageServerApi)?;
374 0 : maybe_rate_limit(&req, tenant_id).await;
375 :
376 0 : match maybe_forward(req).await {
377 0 : ForwardOutcome::Forwarded(res) => {
378 0 : return res;
379 : }
380 0 : ForwardOutcome::NotForwarded(_req) => {}
381 : };
382 :
383 0 : let status_code = service
384 0 : .tenant_delete(tenant_id)
385 0 : .await
386 0 : .and_then(map_reqwest_hyper_status)?;
387 :
388 0 : if status_code == StatusCode::NOT_FOUND {
389 : // The pageserver uses 404 for successful deletion, but we use 200
390 0 : json_response(StatusCode::OK, ())
391 : } else {
392 0 : json_response(status_code, ())
393 : }
394 0 : }
395 :
396 0 : async fn handle_tenant_timeline_create(
397 0 : service: Arc<Service>,
398 0 : req: Request<Body>,
399 0 : ) -> Result<Response<Body>, ApiError> {
400 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
401 0 : check_permissions(&req, Scope::PageServerApi)?;
402 0 : maybe_rate_limit(&req, tenant_id).await;
403 :
404 0 : let mut req = match maybe_forward(req).await {
405 0 : ForwardOutcome::Forwarded(res) => {
406 0 : return res;
407 : }
408 0 : ForwardOutcome::NotForwarded(req) => req,
409 : };
410 :
411 0 : let create_req = json_request::<TimelineCreateRequest>(&mut req).await?;
412 : json_response(
413 : StatusCode::CREATED,
414 0 : service
415 0 : .tenant_timeline_create(tenant_id, create_req)
416 0 : .await?,
417 : )
418 0 : }
419 :
420 0 : async fn handle_tenant_timeline_delete(
421 0 : service: Arc<Service>,
422 0 : req: Request<Body>,
423 0 : ) -> Result<Response<Body>, ApiError> {
424 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
425 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
426 :
427 0 : check_permissions(&req, Scope::PageServerApi)?;
428 0 : maybe_rate_limit(&req, tenant_id).await;
429 :
430 0 : match maybe_forward(req).await {
431 0 : ForwardOutcome::Forwarded(res) => {
432 0 : return res;
433 : }
434 0 : ForwardOutcome::NotForwarded(_req) => {}
435 : };
436 :
437 : // For timeline deletions, which both implement an "initially return 202, then 404 once
438 : // we're done" semantic, we wrap with a retry loop to expose a simpler API upstream.
439 0 : async fn deletion_wrapper<R, F>(service: Arc<Service>, f: F) -> Result<Response<Body>, ApiError>
440 0 : where
441 0 : R: std::future::Future<Output = Result<StatusCode, ApiError>> + Send + 'static,
442 0 : F: Fn(Arc<Service>) -> R + Send + Sync + 'static,
443 0 : {
444 : // On subsequent retries, wait longer.
445 : // Enable callers with a 25 second request timeout to reliably get a response
446 : const MAX_WAIT: Duration = Duration::from_secs(25);
447 : const MAX_RETRY_PERIOD: Duration = Duration::from_secs(5);
448 :
449 0 : let started_at = Instant::now();
450 0 :
451 0 : // To keep deletion reasonably snappy for small tenants, initially check after 1 second if deletion
452 0 : // completed.
453 0 : let mut retry_period = Duration::from_secs(1);
454 :
455 : loop {
456 0 : let status = f(service.clone()).await?;
457 0 : match status {
458 : StatusCode::ACCEPTED => {
459 0 : tracing::info!("Deletion accepted, waiting to try again...");
460 0 : tokio::time::sleep(retry_period).await;
461 0 : retry_period = MAX_RETRY_PERIOD;
462 : }
463 : StatusCode::CONFLICT => {
464 0 : tracing::info!("Deletion already in progress, waiting to try again...");
465 0 : tokio::time::sleep(retry_period).await;
466 : }
467 : StatusCode::NOT_FOUND => {
468 0 : tracing::info!("Deletion complete");
469 0 : return json_response(StatusCode::OK, ());
470 : }
471 : _ => {
472 0 : tracing::warn!("Unexpected status {status}");
473 0 : return json_response(status, ());
474 : }
475 : }
476 :
477 0 : let now = Instant::now();
478 0 : if now + retry_period > started_at + MAX_WAIT {
479 0 : tracing::info!("Deletion timed out waiting for 404");
480 : // REQUEST_TIMEOUT would be more appropriate, but CONFLICT is already part of
481 : // the pageserver's swagger definition for this endpoint, and has the same desired
482 : // effect of causing the control plane to retry later.
483 0 : return json_response(StatusCode::CONFLICT, ());
484 0 : }
485 : }
486 0 : }
487 :
488 0 : deletion_wrapper(service, move |service| async move {
489 0 : service
490 0 : .tenant_timeline_delete(tenant_id, timeline_id)
491 0 : .await
492 0 : .and_then(map_reqwest_hyper_status)
493 0 : })
494 0 : .await
495 0 : }
496 :
497 0 : async fn handle_tenant_timeline_archival_config(
498 0 : service: Arc<Service>,
499 0 : req: Request<Body>,
500 0 : ) -> Result<Response<Body>, ApiError> {
501 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
502 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
503 :
504 0 : check_permissions(&req, Scope::PageServerApi)?;
505 0 : maybe_rate_limit(&req, tenant_id).await;
506 :
507 0 : let mut req = match maybe_forward(req).await {
508 0 : ForwardOutcome::Forwarded(res) => {
509 0 : return res;
510 : }
511 0 : ForwardOutcome::NotForwarded(req) => req,
512 : };
513 :
514 0 : let create_req = json_request::<TimelineArchivalConfigRequest>(&mut req).await?;
515 :
516 0 : service
517 0 : .tenant_timeline_archival_config(tenant_id, timeline_id, create_req)
518 0 : .await?;
519 :
520 0 : json_response(StatusCode::OK, ())
521 0 : }
522 :
523 0 : async fn handle_tenant_timeline_detach_ancestor(
524 0 : service: Arc<Service>,
525 0 : req: Request<Body>,
526 0 : ) -> Result<Response<Body>, ApiError> {
527 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
528 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
529 0 : let behavior: Option<DetachBehavior> = parse_query_param(&req, "detach_behavior")?;
530 :
531 0 : check_permissions(&req, Scope::PageServerApi)?;
532 0 : maybe_rate_limit(&req, tenant_id).await;
533 :
534 0 : match maybe_forward(req).await {
535 0 : ForwardOutcome::Forwarded(res) => {
536 0 : return res;
537 : }
538 0 : ForwardOutcome::NotForwarded(_req) => {}
539 : };
540 :
541 0 : let res = service
542 0 : .tenant_timeline_detach_ancestor(tenant_id, timeline_id, behavior)
543 0 : .await?;
544 :
545 0 : json_response(StatusCode::OK, res)
546 0 : }
547 :
548 0 : async fn handle_tenant_timeline_block_unblock_gc(
549 0 : service: Arc<Service>,
550 0 : req: Request<Body>,
551 0 : dir: BlockUnblock,
552 0 : ) -> Result<Response<Body>, ApiError> {
553 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
554 0 : check_permissions(&req, Scope::PageServerApi)?;
555 0 : maybe_rate_limit(&req, tenant_id).await;
556 :
557 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
558 :
559 0 : service
560 0 : .tenant_timeline_block_unblock_gc(tenant_id, timeline_id, dir)
561 0 : .await?;
562 :
563 0 : json_response(StatusCode::OK, ())
564 0 : }
565 :
566 0 : async fn handle_tenant_timeline_download_heatmap_layers(
567 0 : service: Arc<Service>,
568 0 : req: Request<Body>,
569 0 : ) -> Result<Response<Body>, ApiError> {
570 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
571 :
572 0 : check_permissions(&req, Scope::PageServerApi)?;
573 0 : maybe_rate_limit(&req, tenant_shard_id.tenant_id).await;
574 :
575 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
576 0 : let concurrency: Option<usize> = parse_query_param(&req, "concurrency")?;
577 0 : let recurse = parse_query_param(&req, "recurse")?.unwrap_or(false);
578 0 :
579 0 : service
580 0 : .tenant_timeline_download_heatmap_layers(tenant_shard_id, timeline_id, concurrency, recurse)
581 0 : .await?;
582 :
583 0 : json_response(StatusCode::OK, ())
584 0 : }
585 :
586 0 : async fn handle_tenant_timeline_lsn_lease(
587 0 : service: Arc<Service>,
588 0 : req: Request<Body>,
589 0 : ) -> Result<Response<Body>, ApiError> {
590 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
591 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
592 :
593 0 : check_permissions(&req, Scope::PageServerApi)?;
594 0 : maybe_rate_limit(&req, tenant_id).await;
595 :
596 0 : let mut req = match maybe_forward(req).await {
597 0 : ForwardOutcome::Forwarded(res) => {
598 0 : return res;
599 : }
600 0 : ForwardOutcome::NotForwarded(req) => req,
601 : };
602 :
603 0 : let lsn_lease_request = json_request::<LsnLeaseRequest>(&mut req).await?;
604 :
605 0 : service
606 0 : .tenant_timeline_lsn_lease(tenant_id, timeline_id, lsn_lease_request.lsn)
607 0 : .await?;
608 :
609 0 : json_response(StatusCode::OK, ())
610 0 : }
611 :
612 : // For metric labels where we would like to include the approximate path, but exclude high-cardinality fields like query parameters
613 : // and tenant/timeline IDs. Since we are proxying to arbitrary paths, we don't have routing templates to
614 : // compare to, so we can just filter out our well known ID format with regexes.
615 3 : fn path_without_ids(path: &str) -> String {
616 : static ID_REGEX: std::sync::OnceLock<regex::Regex> = std::sync::OnceLock::new();
617 3 : ID_REGEX
618 3 : .get_or_init(|| regex::Regex::new(r"([0-9a-fA-F]{32}(-[0-9]{4})?|\?.*)").unwrap())
619 3 : .replace_all(path, "")
620 3 : .to_string()
621 3 : }
622 :
623 0 : async fn handle_tenant_timeline_passthrough(
624 0 : service: Arc<Service>,
625 0 : req: Request<Body>,
626 0 : ) -> Result<Response<Body>, ApiError> {
627 0 : let tenant_or_shard_id: TenantShardId = parse_request_param(&req, "tenant_id")?;
628 0 : check_permissions(&req, Scope::PageServerApi)?;
629 0 : maybe_rate_limit(&req, tenant_or_shard_id.tenant_id).await;
630 :
631 0 : let req = match maybe_forward(req).await {
632 0 : ForwardOutcome::Forwarded(res) => {
633 0 : return res;
634 : }
635 0 : ForwardOutcome::NotForwarded(req) => req,
636 : };
637 :
638 0 : let Some(path) = req.uri().path_and_query() else {
639 : // This should never happen, our request router only calls us if there is a path
640 0 : return Err(ApiError::BadRequest(anyhow::anyhow!("Missing path")));
641 : };
642 :
643 0 : let method = match *req.method() {
644 0 : hyper::Method::GET => reqwest::Method::GET,
645 0 : hyper::Method::POST => reqwest::Method::POST,
646 0 : hyper::Method::PUT => reqwest::Method::PUT,
647 0 : hyper::Method::DELETE => reqwest::Method::DELETE,
648 0 : hyper::Method::PATCH => reqwest::Method::PATCH,
649 0 : _ => return Err(ApiError::BadRequest(anyhow::anyhow!("Unsupported method"))),
650 : };
651 :
652 0 : tracing::info!(
653 0 : "Proxying request for tenant {} ({})",
654 : tenant_or_shard_id.tenant_id,
655 : path
656 : );
657 :
658 : // Find the node that holds shard zero
659 0 : let (node, tenant_shard_id) = if tenant_or_shard_id.is_unsharded() {
660 0 : service
661 0 : .tenant_shard0_node(tenant_or_shard_id.tenant_id)
662 0 : .await?
663 : } else {
664 : (
665 0 : service.tenant_shard_node(tenant_or_shard_id).await?,
666 0 : tenant_or_shard_id,
667 : )
668 : };
669 :
670 : // Callers will always pass an unsharded tenant ID. Before proxying, we must
671 : // rewrite this to a shard-aware shard zero ID.
672 0 : let path = format!("{}", path);
673 0 : let tenant_str = tenant_or_shard_id.tenant_id.to_string();
674 0 : let tenant_shard_str = format!("{}", tenant_shard_id);
675 0 : let path = path.replace(&tenant_str, &tenant_shard_str);
676 0 :
677 0 : let latency = &METRICS_REGISTRY
678 0 : .metrics_group
679 0 : .storage_controller_passthrough_request_latency;
680 0 :
681 0 : let path_label = path_without_ids(&path)
682 0 : .split('/')
683 0 : .filter(|token| !token.is_empty())
684 0 : .collect::<Vec<_>>()
685 0 : .join("_");
686 0 : let labels = PageserverRequestLabelGroup {
687 0 : pageserver_id: &node.get_id().to_string(),
688 0 : path: &path_label,
689 0 : method: crate::metrics::Method::Get,
690 0 : };
691 0 :
692 0 : let _timer = latency.start_timer(labels.clone());
693 0 :
694 0 : let client = mgmt_api::Client::new(
695 0 : service.get_http_client().clone(),
696 0 : node.base_url(),
697 0 : service.get_config().pageserver_jwt_token.as_deref(),
698 0 : );
699 0 : let resp = client.op_raw(method, path).await.map_err(|e|
700 : // We return 503 here because if we can't successfully send a request to the pageserver,
701 : // either we aren't available or the pageserver is unavailable.
702 0 : ApiError::ResourceUnavailable(format!("Error sending pageserver API request to {node}: {e}").into()))?;
703 :
704 0 : if !resp.status().is_success() {
705 0 : let error_counter = &METRICS_REGISTRY
706 0 : .metrics_group
707 0 : .storage_controller_passthrough_request_error;
708 0 : error_counter.inc(labels);
709 0 : }
710 :
711 : // Transform 404 into 503 if we raced with a migration
712 0 : if resp.status() == reqwest::StatusCode::NOT_FOUND {
713 : // Look up node again: if we migrated it will be different
714 0 : let new_node = service.tenant_shard_node(tenant_shard_id).await?;
715 0 : if new_node.get_id() != node.get_id() {
716 : // Rather than retry here, send the client a 503 to prompt a retry: this matches
717 : // the pageserver's use of 503, and all clients calling this API should retry on 503.
718 0 : return Err(ApiError::ResourceUnavailable(
719 0 : format!("Pageserver {node} returned 404, was migrated to {new_node}").into(),
720 0 : ));
721 0 : }
722 0 : }
723 :
724 : // We have a reqest::Response, would like a http::Response
725 0 : let mut builder = hyper::Response::builder().status(map_reqwest_hyper_status(resp.status())?);
726 0 : for (k, v) in resp.headers() {
727 0 : builder = builder.header(k.as_str(), v.as_bytes());
728 0 : }
729 :
730 0 : let response = builder
731 0 : .body(Body::wrap_stream(resp.bytes_stream()))
732 0 : .map_err(|e| ApiError::InternalServerError(e.into()))?;
733 :
734 0 : Ok(response)
735 0 : }
736 :
737 0 : async fn handle_tenant_locate(
738 0 : service: Arc<Service>,
739 0 : req: Request<Body>,
740 0 : ) -> Result<Response<Body>, ApiError> {
741 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
742 :
743 0 : check_permissions(&req, Scope::Admin)?;
744 : // NB: don't rate limit: admin operation.
745 :
746 0 : match maybe_forward(req).await {
747 0 : ForwardOutcome::Forwarded(res) => {
748 0 : return res;
749 : }
750 0 : ForwardOutcome::NotForwarded(_req) => {}
751 0 : };
752 0 :
753 0 : json_response(StatusCode::OK, service.tenant_locate(tenant_id)?)
754 0 : }
755 :
756 0 : async fn handle_tenant_describe(
757 0 : service: Arc<Service>,
758 0 : req: Request<Body>,
759 0 : ) -> Result<Response<Body>, ApiError> {
760 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
761 0 : check_permissions(&req, Scope::Scrubber)?;
762 : // NB: don't rate limit: scrubber operation.
763 :
764 0 : match maybe_forward(req).await {
765 0 : ForwardOutcome::Forwarded(res) => {
766 0 : return res;
767 : }
768 0 : ForwardOutcome::NotForwarded(_req) => {}
769 0 : };
770 0 :
771 0 : json_response(StatusCode::OK, service.tenant_describe(tenant_id)?)
772 0 : }
773 :
774 0 : async fn handle_tenant_list(
775 0 : service: Arc<Service>,
776 0 : req: Request<Body>,
777 0 : ) -> Result<Response<Body>, ApiError> {
778 0 : check_permissions(&req, Scope::Admin)?;
779 :
780 0 : let limit: Option<usize> = parse_query_param(&req, "limit")?;
781 0 : let start_after: Option<TenantId> = parse_query_param(&req, "start_after")?;
782 0 : tracing::info!("start_after: {:?}", start_after);
783 :
784 0 : match maybe_forward(req).await {
785 0 : ForwardOutcome::Forwarded(res) => {
786 0 : return res;
787 : }
788 0 : ForwardOutcome::NotForwarded(_req) => {}
789 0 : };
790 0 :
791 0 : json_response(StatusCode::OK, service.tenant_list(limit, start_after))
792 0 : }
793 :
794 0 : async fn handle_node_register(req: Request<Body>) -> Result<Response<Body>, ApiError> {
795 0 : check_permissions(&req, Scope::Infra)?;
796 :
797 0 : let mut req = match maybe_forward(req).await {
798 0 : ForwardOutcome::Forwarded(res) => {
799 0 : return res;
800 : }
801 0 : ForwardOutcome::NotForwarded(req) => req,
802 : };
803 :
804 0 : let register_req = json_request::<NodeRegisterRequest>(&mut req).await?;
805 0 : let state = get_state(&req);
806 0 : state.service.node_register(register_req).await?;
807 0 : json_response(StatusCode::OK, ())
808 0 : }
809 :
810 0 : async fn handle_node_list(req: Request<Body>) -> Result<Response<Body>, ApiError> {
811 0 : check_permissions(&req, Scope::Infra)?;
812 :
813 0 : let req = match maybe_forward(req).await {
814 0 : ForwardOutcome::Forwarded(res) => {
815 0 : return res;
816 : }
817 0 : ForwardOutcome::NotForwarded(req) => req,
818 0 : };
819 0 :
820 0 : let state = get_state(&req);
821 0 : let mut nodes = state.service.node_list().await?;
822 0 : nodes.sort_by_key(|n| n.get_id());
823 0 : let api_nodes = nodes.into_iter().map(|n| n.describe()).collect::<Vec<_>>();
824 0 :
825 0 : json_response(StatusCode::OK, api_nodes)
826 0 : }
827 :
828 0 : async fn handle_node_drop(req: Request<Body>) -> Result<Response<Body>, ApiError> {
829 0 : check_permissions(&req, Scope::Admin)?;
830 :
831 0 : let req = match maybe_forward(req).await {
832 0 : ForwardOutcome::Forwarded(res) => {
833 0 : return res;
834 : }
835 0 : ForwardOutcome::NotForwarded(req) => req,
836 0 : };
837 0 :
838 0 : let state = get_state(&req);
839 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
840 0 : json_response(StatusCode::OK, state.service.node_drop(node_id).await?)
841 0 : }
842 :
843 0 : async fn handle_node_delete(req: Request<Body>) -> Result<Response<Body>, ApiError> {
844 0 : check_permissions(&req, Scope::Admin)?;
845 :
846 0 : let req = match maybe_forward(req).await {
847 0 : ForwardOutcome::Forwarded(res) => {
848 0 : return res;
849 : }
850 0 : ForwardOutcome::NotForwarded(req) => req,
851 0 : };
852 0 :
853 0 : let state = get_state(&req);
854 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
855 0 : json_response(StatusCode::OK, state.service.node_delete(node_id).await?)
856 0 : }
857 :
858 0 : async fn handle_node_configure(req: Request<Body>) -> Result<Response<Body>, ApiError> {
859 0 : check_permissions(&req, Scope::Admin)?;
860 :
861 0 : let mut req = match maybe_forward(req).await {
862 0 : ForwardOutcome::Forwarded(res) => {
863 0 : return res;
864 : }
865 0 : ForwardOutcome::NotForwarded(req) => req,
866 : };
867 :
868 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
869 0 : let config_req = json_request::<NodeConfigureRequest>(&mut req).await?;
870 0 : if node_id != config_req.node_id {
871 0 : return Err(ApiError::BadRequest(anyhow::anyhow!(
872 0 : "Path and body node_id differ"
873 0 : )));
874 0 : }
875 0 : let state = get_state(&req);
876 0 :
877 0 : json_response(
878 0 : StatusCode::OK,
879 0 : state
880 0 : .service
881 0 : .external_node_configure(
882 0 : config_req.node_id,
883 0 : config_req.availability.map(NodeAvailability::from),
884 0 : config_req.scheduling,
885 0 : )
886 0 : .await?,
887 : )
888 0 : }
889 :
890 0 : async fn handle_node_status(req: Request<Body>) -> Result<Response<Body>, ApiError> {
891 0 : check_permissions(&req, Scope::Infra)?;
892 :
893 0 : let req = match maybe_forward(req).await {
894 0 : ForwardOutcome::Forwarded(res) => {
895 0 : return res;
896 : }
897 0 : ForwardOutcome::NotForwarded(req) => req,
898 0 : };
899 0 :
900 0 : let state = get_state(&req);
901 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
902 :
903 0 : let node_status = state.service.get_node(node_id).await?.describe();
904 0 :
905 0 : json_response(StatusCode::OK, node_status)
906 0 : }
907 :
908 0 : async fn handle_node_shards(req: Request<Body>) -> Result<Response<Body>, ApiError> {
909 0 : check_permissions(&req, Scope::Admin)?;
910 :
911 0 : let state = get_state(&req);
912 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
913 :
914 0 : let node_status = state.service.get_node_shards(node_id).await?;
915 :
916 0 : json_response(StatusCode::OK, node_status)
917 0 : }
918 :
919 0 : async fn handle_get_leader(req: Request<Body>) -> Result<Response<Body>, ApiError> {
920 0 : check_permissions(&req, Scope::Admin)?;
921 :
922 0 : let req = match maybe_forward(req).await {
923 0 : ForwardOutcome::Forwarded(res) => {
924 0 : return res;
925 : }
926 0 : ForwardOutcome::NotForwarded(req) => req,
927 0 : };
928 0 :
929 0 : let state = get_state(&req);
930 0 : let leader = state.service.get_leader().await.map_err(|err| {
931 0 : ApiError::InternalServerError(anyhow::anyhow!(
932 0 : "Failed to read leader from database: {err}"
933 0 : ))
934 0 : })?;
935 :
936 0 : json_response(StatusCode::OK, leader)
937 0 : }
938 :
939 0 : async fn handle_node_drain(req: Request<Body>) -> Result<Response<Body>, ApiError> {
940 0 : check_permissions(&req, Scope::Infra)?;
941 :
942 0 : let req = match maybe_forward(req).await {
943 0 : ForwardOutcome::Forwarded(res) => {
944 0 : return res;
945 : }
946 0 : ForwardOutcome::NotForwarded(req) => req,
947 0 : };
948 0 :
949 0 : let state = get_state(&req);
950 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
951 :
952 0 : state.service.start_node_drain(node_id).await?;
953 :
954 0 : json_response(StatusCode::ACCEPTED, ())
955 0 : }
956 :
957 0 : async fn handle_cancel_node_drain(req: Request<Body>) -> Result<Response<Body>, ApiError> {
958 0 : check_permissions(&req, Scope::Infra)?;
959 :
960 0 : let req = match maybe_forward(req).await {
961 0 : ForwardOutcome::Forwarded(res) => {
962 0 : return res;
963 : }
964 0 : ForwardOutcome::NotForwarded(req) => req,
965 0 : };
966 0 :
967 0 : let state = get_state(&req);
968 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
969 :
970 0 : state.service.cancel_node_drain(node_id).await?;
971 :
972 0 : json_response(StatusCode::ACCEPTED, ())
973 0 : }
974 :
975 0 : async fn handle_node_fill(req: Request<Body>) -> Result<Response<Body>, ApiError> {
976 0 : check_permissions(&req, Scope::Infra)?;
977 :
978 0 : let req = match maybe_forward(req).await {
979 0 : ForwardOutcome::Forwarded(res) => {
980 0 : return res;
981 : }
982 0 : ForwardOutcome::NotForwarded(req) => req,
983 0 : };
984 0 :
985 0 : let state = get_state(&req);
986 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
987 :
988 0 : state.service.start_node_fill(node_id).await?;
989 :
990 0 : json_response(StatusCode::ACCEPTED, ())
991 0 : }
992 :
993 0 : async fn handle_cancel_node_fill(req: Request<Body>) -> Result<Response<Body>, ApiError> {
994 0 : check_permissions(&req, Scope::Infra)?;
995 :
996 0 : let req = match maybe_forward(req).await {
997 0 : ForwardOutcome::Forwarded(res) => {
998 0 : return res;
999 : }
1000 0 : ForwardOutcome::NotForwarded(req) => req,
1001 0 : };
1002 0 :
1003 0 : let state = get_state(&req);
1004 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
1005 :
1006 0 : state.service.cancel_node_fill(node_id).await?;
1007 :
1008 0 : json_response(StatusCode::ACCEPTED, ())
1009 0 : }
1010 :
1011 0 : async fn handle_safekeeper_list(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1012 0 : check_permissions(&req, Scope::Infra)?;
1013 :
1014 0 : let req = match maybe_forward(req).await {
1015 0 : ForwardOutcome::Forwarded(res) => {
1016 0 : return res;
1017 : }
1018 0 : ForwardOutcome::NotForwarded(req) => req,
1019 0 : };
1020 0 :
1021 0 : let state = get_state(&req);
1022 0 : let safekeepers = state.service.safekeepers_list().await?;
1023 0 : json_response(StatusCode::OK, safekeepers)
1024 0 : }
1025 :
1026 0 : async fn handle_metadata_health_update(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1027 0 : check_permissions(&req, Scope::Scrubber)?;
1028 :
1029 0 : let mut req = match maybe_forward(req).await {
1030 0 : ForwardOutcome::Forwarded(res) => {
1031 0 : return res;
1032 : }
1033 0 : ForwardOutcome::NotForwarded(req) => req,
1034 : };
1035 :
1036 0 : let update_req = json_request::<MetadataHealthUpdateRequest>(&mut req).await?;
1037 0 : let state = get_state(&req);
1038 0 :
1039 0 : state.service.metadata_health_update(update_req).await?;
1040 :
1041 0 : json_response(StatusCode::OK, MetadataHealthUpdateResponse {})
1042 0 : }
1043 :
1044 0 : async fn handle_metadata_health_list_unhealthy(
1045 0 : req: Request<Body>,
1046 0 : ) -> Result<Response<Body>, ApiError> {
1047 0 : check_permissions(&req, Scope::Admin)?;
1048 :
1049 0 : let req = match maybe_forward(req).await {
1050 0 : ForwardOutcome::Forwarded(res) => {
1051 0 : return res;
1052 : }
1053 0 : ForwardOutcome::NotForwarded(req) => req,
1054 0 : };
1055 0 :
1056 0 : let state = get_state(&req);
1057 0 : let unhealthy_tenant_shards = state.service.metadata_health_list_unhealthy().await?;
1058 :
1059 0 : json_response(
1060 0 : StatusCode::OK,
1061 0 : MetadataHealthListUnhealthyResponse {
1062 0 : unhealthy_tenant_shards,
1063 0 : },
1064 0 : )
1065 0 : }
1066 :
1067 0 : async fn handle_metadata_health_list_outdated(
1068 0 : req: Request<Body>,
1069 0 : ) -> Result<Response<Body>, ApiError> {
1070 0 : check_permissions(&req, Scope::Admin)?;
1071 :
1072 0 : let mut req = match maybe_forward(req).await {
1073 0 : ForwardOutcome::Forwarded(res) => {
1074 0 : return res;
1075 : }
1076 0 : ForwardOutcome::NotForwarded(req) => req,
1077 : };
1078 :
1079 0 : let list_outdated_req = json_request::<MetadataHealthListOutdatedRequest>(&mut req).await?;
1080 0 : let state = get_state(&req);
1081 0 : let health_records = state
1082 0 : .service
1083 0 : .metadata_health_list_outdated(list_outdated_req.not_scrubbed_for)
1084 0 : .await?;
1085 :
1086 0 : json_response(
1087 0 : StatusCode::OK,
1088 0 : MetadataHealthListOutdatedResponse { health_records },
1089 0 : )
1090 0 : }
1091 :
1092 0 : async fn handle_tenant_shard_split(
1093 0 : service: Arc<Service>,
1094 0 : req: Request<Body>,
1095 0 : ) -> Result<Response<Body>, ApiError> {
1096 0 : check_permissions(&req, Scope::Admin)?;
1097 : // NB: don't rate limit: admin operation.
1098 :
1099 0 : let mut req = match maybe_forward(req).await {
1100 0 : ForwardOutcome::Forwarded(res) => {
1101 0 : return res;
1102 : }
1103 0 : ForwardOutcome::NotForwarded(req) => req,
1104 : };
1105 :
1106 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1107 0 : let split_req = json_request::<TenantShardSplitRequest>(&mut req).await?;
1108 :
1109 : json_response(
1110 : StatusCode::OK,
1111 0 : service.tenant_shard_split(tenant_id, split_req).await?,
1112 : )
1113 0 : }
1114 :
1115 0 : async fn handle_tenant_shard_migrate(
1116 0 : service: Arc<Service>,
1117 0 : req: Request<Body>,
1118 0 : ) -> Result<Response<Body>, ApiError> {
1119 0 : check_permissions(&req, Scope::Admin)?;
1120 : // NB: don't rate limit: admin operation.
1121 :
1122 0 : let mut req = match maybe_forward(req).await {
1123 0 : ForwardOutcome::Forwarded(res) => {
1124 0 : return res;
1125 : }
1126 0 : ForwardOutcome::NotForwarded(req) => req,
1127 : };
1128 :
1129 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
1130 0 : let migrate_req = json_request::<TenantShardMigrateRequest>(&mut req).await?;
1131 : json_response(
1132 : StatusCode::OK,
1133 0 : service
1134 0 : .tenant_shard_migrate(tenant_shard_id, migrate_req)
1135 0 : .await?,
1136 : )
1137 0 : }
1138 :
1139 0 : async fn handle_tenant_shard_migrate_secondary(
1140 0 : service: Arc<Service>,
1141 0 : req: Request<Body>,
1142 0 : ) -> Result<Response<Body>, ApiError> {
1143 0 : check_permissions(&req, Scope::Admin)?;
1144 : // NB: don't rate limit: admin operation.
1145 :
1146 0 : let mut req = match maybe_forward(req).await {
1147 0 : ForwardOutcome::Forwarded(res) => {
1148 0 : return res;
1149 : }
1150 0 : ForwardOutcome::NotForwarded(req) => req,
1151 : };
1152 :
1153 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
1154 0 : let migrate_req = json_request::<TenantShardMigrateRequest>(&mut req).await?;
1155 : json_response(
1156 : StatusCode::OK,
1157 0 : service
1158 0 : .tenant_shard_migrate_secondary(tenant_shard_id, migrate_req)
1159 0 : .await?,
1160 : )
1161 0 : }
1162 :
1163 0 : async fn handle_tenant_shard_cancel_reconcile(
1164 0 : service: Arc<Service>,
1165 0 : req: Request<Body>,
1166 0 : ) -> Result<Response<Body>, ApiError> {
1167 0 : check_permissions(&req, Scope::Admin)?;
1168 : // NB: don't rate limit: admin operation.
1169 :
1170 0 : let req = match maybe_forward(req).await {
1171 0 : ForwardOutcome::Forwarded(res) => {
1172 0 : return res;
1173 : }
1174 0 : ForwardOutcome::NotForwarded(req) => req,
1175 : };
1176 :
1177 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
1178 : json_response(
1179 : StatusCode::OK,
1180 0 : service
1181 0 : .tenant_shard_cancel_reconcile(tenant_shard_id)
1182 0 : .await?,
1183 : )
1184 0 : }
1185 :
1186 0 : async fn handle_tenant_update_policy(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1187 0 : check_permissions(&req, Scope::Admin)?;
1188 : // NB: don't rate limit: admin operation.
1189 :
1190 0 : let mut req = match maybe_forward(req).await {
1191 0 : ForwardOutcome::Forwarded(res) => {
1192 0 : return res;
1193 : }
1194 0 : ForwardOutcome::NotForwarded(req) => req,
1195 : };
1196 :
1197 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1198 0 : let update_req = json_request::<TenantPolicyRequest>(&mut req).await?;
1199 0 : let state = get_state(&req);
1200 0 :
1201 0 : json_response(
1202 0 : StatusCode::OK,
1203 0 : state
1204 0 : .service
1205 0 : .tenant_update_policy(tenant_id, update_req)
1206 0 : .await?,
1207 : )
1208 0 : }
1209 :
1210 0 : async fn handle_update_preferred_azs(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1211 0 : check_permissions(&req, Scope::Admin)?;
1212 :
1213 0 : let mut req = match maybe_forward(req).await {
1214 0 : ForwardOutcome::Forwarded(res) => {
1215 0 : return res;
1216 : }
1217 0 : ForwardOutcome::NotForwarded(req) => req,
1218 : };
1219 :
1220 0 : let azs_req = json_request::<ShardsPreferredAzsRequest>(&mut req).await?;
1221 0 : let state = get_state(&req);
1222 0 :
1223 0 : json_response(
1224 0 : StatusCode::OK,
1225 0 : state.service.update_shards_preferred_azs(azs_req).await?,
1226 : )
1227 0 : }
1228 :
1229 0 : async fn handle_step_down(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1230 0 : check_permissions(&req, Scope::ControllerPeer)?;
1231 :
1232 0 : let req = match maybe_forward(req).await {
1233 0 : ForwardOutcome::Forwarded(res) => {
1234 0 : return res;
1235 : }
1236 0 : ForwardOutcome::NotForwarded(req) => req,
1237 0 : };
1238 0 :
1239 0 : // Spawn a background task: once we start stepping down, we must finish: if the client drops
1240 0 : // their request we should avoid stopping in some part-stepped-down state.
1241 0 : let handle = tokio::spawn(async move {
1242 0 : let state = get_state(&req);
1243 0 : state.service.step_down().await
1244 0 : });
1245 :
1246 0 : let result = handle
1247 0 : .await
1248 0 : .map_err(|e| ApiError::InternalServerError(e.into()))?;
1249 :
1250 0 : json_response(StatusCode::OK, result)
1251 0 : }
1252 :
1253 0 : async fn handle_tenant_drop(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1254 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1255 0 : check_permissions(&req, Scope::PageServerApi)?;
1256 0 : maybe_rate_limit(&req, tenant_id).await;
1257 :
1258 0 : let req = match maybe_forward(req).await {
1259 0 : ForwardOutcome::Forwarded(res) => {
1260 0 : return res;
1261 : }
1262 0 : ForwardOutcome::NotForwarded(req) => req,
1263 0 : };
1264 0 :
1265 0 : let state = get_state(&req);
1266 0 :
1267 0 : json_response(StatusCode::OK, state.service.tenant_drop(tenant_id).await?)
1268 0 : }
1269 :
1270 0 : async fn handle_tenant_import(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1271 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1272 0 : check_permissions(&req, Scope::PageServerApi)?;
1273 0 : maybe_rate_limit(&req, tenant_id).await;
1274 :
1275 0 : let req = match maybe_forward(req).await {
1276 0 : ForwardOutcome::Forwarded(res) => {
1277 0 : return res;
1278 : }
1279 0 : ForwardOutcome::NotForwarded(req) => req,
1280 0 : };
1281 0 :
1282 0 : let state = get_state(&req);
1283 0 :
1284 0 : json_response(
1285 0 : StatusCode::OK,
1286 0 : state.service.tenant_import(tenant_id).await?,
1287 : )
1288 0 : }
1289 :
1290 0 : async fn handle_timeline_import(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1291 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1292 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
1293 0 : check_permissions(&req, Scope::PageServerApi)?;
1294 0 : maybe_rate_limit(&req, tenant_id).await;
1295 :
1296 0 : let mut req = match maybe_forward(req).await {
1297 0 : ForwardOutcome::Forwarded(res) => {
1298 0 : return res;
1299 : }
1300 0 : ForwardOutcome::NotForwarded(req) => req,
1301 : };
1302 :
1303 0 : let import_req = json_request::<TimelineImportRequest>(&mut req).await?;
1304 :
1305 0 : let state = get_state(&req);
1306 0 :
1307 0 : if import_req.tenant_id != tenant_id || import_req.timeline_id != timeline_id {
1308 0 : return Err(ApiError::BadRequest(anyhow::anyhow!(
1309 0 : "tenant id or timeline id mismatch: url={tenant_id}/{timeline_id}, body={}/{}",
1310 0 : import_req.tenant_id,
1311 0 : import_req.timeline_id
1312 0 : )));
1313 0 : }
1314 0 :
1315 0 : json_response(
1316 0 : StatusCode::OK,
1317 0 : state.service.timeline_import(import_req).await?,
1318 : )
1319 0 : }
1320 :
1321 0 : async fn handle_tenants_dump(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1322 0 : check_permissions(&req, Scope::Admin)?;
1323 :
1324 0 : let req = match maybe_forward(req).await {
1325 0 : ForwardOutcome::Forwarded(res) => {
1326 0 : return res;
1327 : }
1328 0 : ForwardOutcome::NotForwarded(req) => req,
1329 0 : };
1330 0 :
1331 0 : let state = get_state(&req);
1332 0 : state.service.tenants_dump()
1333 0 : }
1334 :
1335 0 : async fn handle_scheduler_dump(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1336 0 : check_permissions(&req, Scope::Admin)?;
1337 :
1338 0 : let req = match maybe_forward(req).await {
1339 0 : ForwardOutcome::Forwarded(res) => {
1340 0 : return res;
1341 : }
1342 0 : ForwardOutcome::NotForwarded(req) => req,
1343 0 : };
1344 0 :
1345 0 : let state = get_state(&req);
1346 0 : state.service.scheduler_dump()
1347 0 : }
1348 :
1349 0 : async fn handle_consistency_check(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1350 0 : check_permissions(&req, Scope::Admin)?;
1351 :
1352 0 : let req = match maybe_forward(req).await {
1353 0 : ForwardOutcome::Forwarded(res) => {
1354 0 : return res;
1355 : }
1356 0 : ForwardOutcome::NotForwarded(req) => req,
1357 0 : };
1358 0 :
1359 0 : let state = get_state(&req);
1360 0 :
1361 0 : json_response(StatusCode::OK, state.service.consistency_check().await?)
1362 0 : }
1363 :
1364 0 : async fn handle_reconcile_all(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1365 0 : check_permissions(&req, Scope::Admin)?;
1366 :
1367 0 : let req = match maybe_forward(req).await {
1368 0 : ForwardOutcome::Forwarded(res) => {
1369 0 : return res;
1370 : }
1371 0 : ForwardOutcome::NotForwarded(req) => req,
1372 0 : };
1373 0 :
1374 0 : let state = get_state(&req);
1375 0 :
1376 0 : json_response(StatusCode::OK, state.service.reconcile_all_now().await?)
1377 0 : }
1378 :
1379 : /// Status endpoint is just used for checking that our HTTP listener is up
1380 0 : async fn handle_status(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1381 0 : match maybe_forward(req).await {
1382 0 : ForwardOutcome::Forwarded(res) => {
1383 0 : return res;
1384 : }
1385 0 : ForwardOutcome::NotForwarded(_req) => {}
1386 0 : };
1387 0 :
1388 0 : json_response(StatusCode::OK, ())
1389 0 : }
1390 :
1391 : /// Readiness endpoint indicates when we're done doing startup I/O (e.g. reconciling
1392 : /// with remote pageserver nodes). This is intended for use as a kubernetes readiness probe.
1393 0 : async fn handle_ready(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1394 0 : let req = match maybe_forward(req).await {
1395 0 : ForwardOutcome::Forwarded(res) => {
1396 0 : return res;
1397 : }
1398 0 : ForwardOutcome::NotForwarded(req) => req,
1399 0 : };
1400 0 :
1401 0 : let state = get_state(&req);
1402 0 : if state.service.startup_complete.is_ready() {
1403 0 : json_response(StatusCode::OK, ())
1404 : } else {
1405 0 : json_response(StatusCode::SERVICE_UNAVAILABLE, ())
1406 : }
1407 0 : }
1408 :
1409 : impl From<ReconcileError> for ApiError {
1410 0 : fn from(value: ReconcileError) -> Self {
1411 0 : ApiError::Conflict(format!("Reconciliation error: {}", value))
1412 0 : }
1413 : }
1414 :
1415 : /// Return the safekeeper record by instance id, or 404.
1416 : ///
1417 : /// Not used by anything except manual testing.
1418 0 : async fn handle_get_safekeeper(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1419 0 : check_permissions(&req, Scope::Infra)?;
1420 :
1421 0 : let id = parse_request_param::<i64>(&req, "id")?;
1422 :
1423 0 : let req = match maybe_forward(req).await {
1424 0 : ForwardOutcome::Forwarded(res) => {
1425 0 : return res;
1426 : }
1427 0 : ForwardOutcome::NotForwarded(req) => req,
1428 0 : };
1429 0 :
1430 0 : let state = get_state(&req);
1431 :
1432 0 : let res = state.service.get_safekeeper(id).await;
1433 :
1434 0 : match res {
1435 0 : Ok(b) => json_response(StatusCode::OK, b),
1436 : Err(crate::persistence::DatabaseError::Query(diesel::result::Error::NotFound)) => {
1437 0 : Err(ApiError::NotFound("unknown instance id".into()))
1438 : }
1439 0 : Err(other) => Err(other.into()),
1440 : }
1441 0 : }
1442 :
1443 : /// Used as part of deployment scripts.
1444 : ///
1445 : /// Assumes information is only relayed to storage controller after first selecting an unique id on
1446 : /// control plane database, which means we have an id field in the request and payload.
1447 0 : async fn handle_upsert_safekeeper(mut req: Request<Body>) -> Result<Response<Body>, ApiError> {
1448 0 : check_permissions(&req, Scope::Infra)?;
1449 :
1450 0 : let body = json_request::<SafekeeperUpsert>(&mut req).await?;
1451 0 : let id = parse_request_param::<i64>(&req, "id")?;
1452 :
1453 0 : if id != body.id {
1454 : // it should be repeated
1455 0 : return Err(ApiError::BadRequest(anyhow::anyhow!(
1456 0 : "id mismatch: url={id:?}, body={:?}",
1457 0 : body.id
1458 0 : )));
1459 0 : }
1460 0 :
1461 0 : if id <= 0 {
1462 0 : return Err(ApiError::BadRequest(anyhow::anyhow!(
1463 0 : "id not allowed to be zero or negative: {id}"
1464 0 : )));
1465 0 : }
1466 :
1467 0 : let req = match maybe_forward(req).await {
1468 0 : ForwardOutcome::Forwarded(res) => {
1469 0 : return res;
1470 : }
1471 0 : ForwardOutcome::NotForwarded(req) => req,
1472 0 : };
1473 0 :
1474 0 : let state = get_state(&req);
1475 0 :
1476 0 : state.service.upsert_safekeeper(body).await?;
1477 :
1478 0 : Ok(Response::builder()
1479 0 : .status(StatusCode::NO_CONTENT)
1480 0 : .body(Body::empty())
1481 0 : .unwrap())
1482 0 : }
1483 :
1484 : /// Sets the scheduling policy of the specified safekeeper
1485 0 : async fn handle_safekeeper_scheduling_policy(
1486 0 : mut req: Request<Body>,
1487 0 : ) -> Result<Response<Body>, ApiError> {
1488 0 : check_permissions(&req, Scope::Admin)?;
1489 :
1490 0 : let body = json_request::<SafekeeperSchedulingPolicyRequest>(&mut req).await?;
1491 0 : let id = parse_request_param::<i64>(&req, "id")?;
1492 :
1493 0 : let req = match maybe_forward(req).await {
1494 0 : ForwardOutcome::Forwarded(res) => {
1495 0 : return res;
1496 : }
1497 0 : ForwardOutcome::NotForwarded(req) => req,
1498 0 : };
1499 0 :
1500 0 : let state = get_state(&req);
1501 0 :
1502 0 : state
1503 0 : .service
1504 0 : .set_safekeeper_scheduling_policy(id, body.scheduling_policy)
1505 0 : .await?;
1506 :
1507 0 : json_response(StatusCode::OK, ())
1508 0 : }
1509 :
1510 : /// Common wrapper for request handlers that call into Service and will operate on tenants: they must only
1511 : /// be allowed to run if Service has finished its initial reconciliation.
1512 0 : async fn tenant_service_handler<R, H>(
1513 0 : request: Request<Body>,
1514 0 : handler: H,
1515 0 : request_name: RequestName,
1516 0 : ) -> R::Output
1517 0 : where
1518 0 : R: std::future::Future<Output = Result<Response<Body>, ApiError>> + Send + 'static,
1519 0 : H: FnOnce(Arc<Service>, Request<Body>) -> R + Send + Sync + 'static,
1520 0 : {
1521 0 : let state = get_state(&request);
1522 0 : let service = state.service.clone();
1523 0 :
1524 0 : let startup_complete = service.startup_complete.clone();
1525 0 : if tokio::time::timeout(STARTUP_RECONCILE_TIMEOUT, startup_complete.wait())
1526 0 : .await
1527 0 : .is_err()
1528 : {
1529 : // This shouldn't happen: it is the responsibilty of [`Service::startup_reconcile`] to use appropriate
1530 : // timeouts around its remote calls, to bound its runtime.
1531 0 : return Err(ApiError::Timeout(
1532 0 : "Timed out waiting for service readiness".into(),
1533 0 : ));
1534 0 : }
1535 0 :
1536 0 : named_request_span(
1537 0 : request,
1538 0 : |request| async move { handler(service, request).await },
1539 0 : request_name,
1540 0 : )
1541 0 : .await
1542 0 : }
1543 :
1544 : /// Check if the required scope is held in the request's token, or if the request has
1545 : /// a token with 'admin' scope then always permit it.
1546 0 : fn check_permissions(request: &Request<Body>, required_scope: Scope) -> Result<(), ApiError> {
1547 0 : check_permission_with(request, |claims| {
1548 0 : match crate::auth::check_permission(claims, required_scope) {
1549 0 : Err(e) => match crate::auth::check_permission(claims, Scope::Admin) {
1550 0 : Ok(()) => Ok(()),
1551 0 : Err(_) => Err(e),
1552 : },
1553 0 : Ok(()) => Ok(()),
1554 : }
1555 0 : })
1556 0 : }
1557 :
1558 : #[derive(Clone, Debug)]
1559 : struct RequestMeta {
1560 : method: hyper::http::Method,
1561 : at: Instant,
1562 : }
1563 :
1564 0 : pub fn prologue_leadership_status_check_middleware<
1565 0 : B: hyper::body::HttpBody + Send + Sync + 'static,
1566 0 : >() -> Middleware<B, ApiError> {
1567 0 : Middleware::pre(move |req| async move {
1568 0 : let state = get_state(&req);
1569 0 : let leadership_status = state.service.get_leadership_status();
1570 :
1571 : enum AllowedRoutes {
1572 : All,
1573 : Some(&'static [&'static str]),
1574 : }
1575 :
1576 0 : let allowed_routes = match leadership_status {
1577 0 : LeadershipStatus::Leader => AllowedRoutes::All,
1578 0 : LeadershipStatus::SteppedDown => AllowedRoutes::All,
1579 0 : LeadershipStatus::Candidate => AllowedRoutes::Some(&[
1580 0 : "/ready",
1581 0 : "/status",
1582 0 : "/metrics",
1583 0 : "/profile/cpu",
1584 0 : "/profile/heap",
1585 0 : ]),
1586 : };
1587 :
1588 0 : match allowed_routes {
1589 0 : AllowedRoutes::All => Ok(req),
1590 0 : AllowedRoutes::Some(allowed) if allowed.contains(&req.uri().path()) => Ok(req),
1591 : _ => {
1592 0 : tracing::info!(
1593 0 : "Request {} not allowed due to current leadership state",
1594 0 : req.uri()
1595 : );
1596 :
1597 0 : Err(ApiError::ResourceUnavailable(
1598 0 : format!("Current leadership status is {leadership_status}").into(),
1599 0 : ))
1600 : }
1601 : }
1602 0 : })
1603 0 : }
1604 :
1605 0 : fn prologue_metrics_middleware<B: hyper::body::HttpBody + Send + Sync + 'static>()
1606 0 : -> Middleware<B, ApiError> {
1607 0 : Middleware::pre(move |req| async move {
1608 0 : let meta = RequestMeta {
1609 0 : method: req.method().clone(),
1610 0 : at: Instant::now(),
1611 0 : };
1612 0 :
1613 0 : req.set_context(meta);
1614 0 :
1615 0 : Ok(req)
1616 0 : })
1617 0 : }
1618 :
1619 0 : fn epilogue_metrics_middleware<B: hyper::body::HttpBody + Send + Sync + 'static>()
1620 0 : -> Middleware<B, ApiError> {
1621 0 : Middleware::post_with_info(move |resp, req_info| async move {
1622 0 : let request_name = match req_info.context::<RequestName>() {
1623 0 : Some(name) => name,
1624 : None => {
1625 0 : return Ok(resp);
1626 : }
1627 : };
1628 :
1629 0 : if let Some(meta) = req_info.context::<RequestMeta>() {
1630 0 : let status = &crate::metrics::METRICS_REGISTRY
1631 0 : .metrics_group
1632 0 : .storage_controller_http_request_status;
1633 0 : let latency = &crate::metrics::METRICS_REGISTRY
1634 0 : .metrics_group
1635 0 : .storage_controller_http_request_latency;
1636 0 :
1637 0 : status.inc(HttpRequestStatusLabelGroup {
1638 0 : path: request_name.0,
1639 0 : method: meta.method.clone().into(),
1640 0 : status: crate::metrics::StatusCode(resp.status()),
1641 0 : });
1642 0 :
1643 0 : latency.observe(
1644 0 : HttpRequestLatencyLabelGroup {
1645 0 : path: request_name.0,
1646 0 : method: meta.method.into(),
1647 0 : },
1648 0 : meta.at.elapsed().as_secs_f64(),
1649 0 : );
1650 0 : }
1651 0 : Ok(resp)
1652 0 : })
1653 0 : }
1654 :
1655 0 : pub async fn measured_metrics_handler(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1656 : pub const TEXT_FORMAT: &str = "text/plain; version=0.0.4";
1657 :
1658 0 : let req = match maybe_forward(req).await {
1659 0 : ForwardOutcome::Forwarded(res) => {
1660 0 : return res;
1661 : }
1662 0 : ForwardOutcome::NotForwarded(req) => req,
1663 0 : };
1664 0 :
1665 0 : let state = get_state(&req);
1666 0 : let payload = crate::metrics::METRICS_REGISTRY.encode(&state.neon_metrics);
1667 0 : let response = Response::builder()
1668 0 : .status(200)
1669 0 : .header(CONTENT_TYPE, TEXT_FORMAT)
1670 0 : .body(payload.into())
1671 0 : .unwrap();
1672 0 :
1673 0 : Ok(response)
1674 0 : }
1675 :
1676 : #[derive(Clone)]
1677 : struct RequestName(&'static str);
1678 :
1679 0 : async fn named_request_span<R, H>(
1680 0 : request: Request<Body>,
1681 0 : handler: H,
1682 0 : name: RequestName,
1683 0 : ) -> R::Output
1684 0 : where
1685 0 : R: Future<Output = Result<Response<Body>, ApiError>> + Send + 'static,
1686 0 : H: FnOnce(Request<Body>) -> R + Send + Sync + 'static,
1687 0 : {
1688 0 : request.set_context(name);
1689 0 : request_span(request, handler).await
1690 0 : }
1691 :
1692 : enum ForwardOutcome {
1693 : Forwarded(Result<Response<Body>, ApiError>),
1694 : NotForwarded(Request<Body>),
1695 : }
1696 :
1697 : /// Potentially forward the request to the current storage controler leader.
1698 : /// More specifically we forward when:
1699 : /// 1. Request is not one of:
1700 : /// ["/control/v1/step_down", "/status", "/ready", "/metrics", "/profile/cpu", "/profile/heap"]
1701 : /// 2. Current instance is in [`LeadershipStatus::SteppedDown`] state
1702 : /// 3. There is a leader in the database to forward to
1703 : /// 4. Leader from step (3) is not the current instance
1704 : ///
1705 : /// Why forward?
1706 : /// It turns out that we can't rely on external orchestration to promptly route trafic to the
1707 : /// new leader. This is downtime inducing. Forwarding provides a safe way out.
1708 : ///
1709 : /// Why is it safe?
1710 : /// If a storcon instance is persisted in the database, then we know that it is the current leader.
1711 : /// There's one exception: time between handling step-down request and the new leader updating the
1712 : /// database.
1713 : ///
1714 : /// Let's treat the happy case first. The stepped down node does not produce any side effects,
1715 : /// since all request handling happens on the leader.
1716 : ///
1717 : /// As for the edge case, we are guaranteed to always have a maximum of two running instances.
1718 : /// Hence, if we are in the edge case scenario the leader persisted in the database is the
1719 : /// stepped down instance that received the request. Condition (4) above covers this scenario.
1720 0 : async fn maybe_forward(req: Request<Body>) -> ForwardOutcome {
1721 : const NOT_FOR_FORWARD: &[&str] = &[
1722 : "/control/v1/step_down",
1723 : "/status",
1724 : "/ready",
1725 : "/metrics",
1726 : "/profile/cpu",
1727 : "/profile/heap",
1728 : ];
1729 :
1730 0 : let uri = req.uri();
1731 0 : let uri_for_forward = !NOT_FOR_FORWARD.contains(&uri.path());
1732 0 :
1733 0 : // Fast return before trying to take any Service locks, if we will never forward anyway
1734 0 : if !uri_for_forward {
1735 0 : return ForwardOutcome::NotForwarded(req);
1736 0 : }
1737 0 :
1738 0 : let state = get_state(&req);
1739 0 : let leadership_status = state.service.get_leadership_status();
1740 0 :
1741 0 : if leadership_status != LeadershipStatus::SteppedDown {
1742 0 : return ForwardOutcome::NotForwarded(req);
1743 0 : }
1744 :
1745 0 : let leader = state.service.get_leader().await;
1746 0 : let leader = {
1747 0 : match leader {
1748 0 : Ok(Some(leader)) => leader,
1749 : Ok(None) => {
1750 0 : return ForwardOutcome::Forwarded(Err(ApiError::ResourceUnavailable(
1751 0 : "No leader to forward to while in stepped down state".into(),
1752 0 : )));
1753 : }
1754 0 : Err(err) => {
1755 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(
1756 0 : anyhow::anyhow!(
1757 0 : "Failed to get leader for forwarding while in stepped down state: {err}"
1758 0 : ),
1759 0 : )));
1760 : }
1761 : }
1762 : };
1763 :
1764 0 : let cfg = state.service.get_config();
1765 0 : if let Some(ref self_addr) = cfg.address_for_peers {
1766 0 : let leader_addr = match Uri::from_str(leader.address.as_str()) {
1767 0 : Ok(uri) => uri,
1768 0 : Err(err) => {
1769 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(
1770 0 : anyhow::anyhow!(
1771 0 : "Failed to parse leader uri for forwarding while in stepped down state: {err}"
1772 0 : ),
1773 0 : )));
1774 : }
1775 : };
1776 :
1777 0 : if *self_addr == leader_addr {
1778 0 : return ForwardOutcome::Forwarded(Err(ApiError::ResourceUnavailable(
1779 0 : "Leader is stepped down instance".into(),
1780 0 : )));
1781 0 : }
1782 0 : }
1783 :
1784 0 : tracing::info!("Forwarding {} to leader at {}", uri, leader.address);
1785 :
1786 : // Use [`RECONCILE_TIMEOUT`] as the max amount of time a request should block for and
1787 : // include some leeway to get the timeout for proxied requests.
1788 : const PROXIED_REQUEST_TIMEOUT: Duration = Duration::from_secs(RECONCILE_TIMEOUT.as_secs() + 10);
1789 :
1790 0 : let client = state.service.get_http_client().clone();
1791 :
1792 0 : let request: reqwest::Request = match convert_request(
1793 0 : req,
1794 0 : &client,
1795 0 : leader.address,
1796 0 : PROXIED_REQUEST_TIMEOUT,
1797 0 : )
1798 0 : .await
1799 : {
1800 0 : Ok(r) => r,
1801 0 : Err(err) => {
1802 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(anyhow::anyhow!(
1803 0 : "Failed to convert request for forwarding while in stepped down state: {err}"
1804 0 : ))));
1805 : }
1806 : };
1807 :
1808 0 : let response = match client.execute(request).await {
1809 0 : Ok(r) => r,
1810 0 : Err(err) => {
1811 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(anyhow::anyhow!(
1812 0 : "Failed to forward while in stepped down state: {err}"
1813 0 : ))));
1814 : }
1815 : };
1816 :
1817 0 : ForwardOutcome::Forwarded(convert_response(response).await)
1818 0 : }
1819 :
1820 : /// Convert a [`reqwest::Response`] to a [hyper::Response`] by passing through
1821 : /// a stable representation (string, bytes or integer)
1822 : ///
1823 : /// Ideally, we would not have to do this since both types use the http crate
1824 : /// under the hood. However, they use different versions of the crate and keeping
1825 : /// second order dependencies in sync is difficult.
1826 0 : async fn convert_response(resp: reqwest::Response) -> Result<hyper::Response<Body>, ApiError> {
1827 : use std::str::FromStr;
1828 :
1829 0 : let mut builder = hyper::Response::builder().status(resp.status().as_u16());
1830 0 : for (key, value) in resp.headers().into_iter() {
1831 0 : let key = hyper::header::HeaderName::from_str(key.as_str()).map_err(|err| {
1832 0 : ApiError::InternalServerError(anyhow::anyhow!("Response conversion failed: {err}"))
1833 0 : })?;
1834 :
1835 0 : let value = hyper::header::HeaderValue::from_bytes(value.as_bytes()).map_err(|err| {
1836 0 : ApiError::InternalServerError(anyhow::anyhow!("Response conversion failed: {err}"))
1837 0 : })?;
1838 :
1839 0 : builder = builder.header(key, value);
1840 : }
1841 :
1842 0 : let body = http::Body::wrap_stream(resp.bytes_stream());
1843 0 :
1844 0 : builder.body(body).map_err(|err| {
1845 0 : ApiError::InternalServerError(anyhow::anyhow!("Response conversion failed: {err}"))
1846 0 : })
1847 0 : }
1848 :
1849 : /// Convert a [`reqwest::Request`] to a [hyper::Request`] by passing through
1850 : /// a stable representation (string, bytes or integer)
1851 : ///
1852 : /// See [`convert_response`] for why we are doing it this way.
1853 0 : async fn convert_request(
1854 0 : req: hyper::Request<Body>,
1855 0 : client: &reqwest::Client,
1856 0 : to_address: String,
1857 0 : timeout: Duration,
1858 0 : ) -> Result<reqwest::Request, ApiError> {
1859 : use std::str::FromStr;
1860 :
1861 0 : let (parts, body) = req.into_parts();
1862 0 : let method = reqwest::Method::from_str(parts.method.as_str()).map_err(|err| {
1863 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1864 0 : })?;
1865 :
1866 0 : let path_and_query = parts.uri.path_and_query().ok_or_else(|| {
1867 0 : ApiError::InternalServerError(anyhow::anyhow!(
1868 0 : "Request conversion failed: no path and query"
1869 0 : ))
1870 0 : })?;
1871 :
1872 0 : let uri = reqwest::Url::from_str(
1873 0 : format!(
1874 0 : "{}{}",
1875 0 : to_address.trim_end_matches("/"),
1876 0 : path_and_query.as_str()
1877 0 : )
1878 0 : .as_str(),
1879 0 : )
1880 0 : .map_err(|err| {
1881 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1882 0 : })?;
1883 :
1884 0 : let mut headers = reqwest::header::HeaderMap::new();
1885 0 : for (key, value) in parts.headers.into_iter() {
1886 0 : let key = match key {
1887 0 : Some(k) => k,
1888 : None => {
1889 0 : continue;
1890 : }
1891 : };
1892 :
1893 0 : let key = reqwest::header::HeaderName::from_str(key.as_str()).map_err(|err| {
1894 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1895 0 : })?;
1896 :
1897 0 : let value = reqwest::header::HeaderValue::from_bytes(value.as_bytes()).map_err(|err| {
1898 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1899 0 : })?;
1900 :
1901 0 : headers.insert(key, value);
1902 : }
1903 :
1904 0 : let body = hyper::body::to_bytes(body).await.map_err(|err| {
1905 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1906 0 : })?;
1907 :
1908 0 : client
1909 0 : .request(method, uri)
1910 0 : .headers(headers)
1911 0 : .body(body)
1912 0 : .timeout(timeout)
1913 0 : .build()
1914 0 : .map_err(|err| {
1915 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1916 0 : })
1917 0 : }
1918 :
1919 0 : pub fn make_router(
1920 0 : service: Arc<Service>,
1921 0 : auth: Option<Arc<SwappableJwtAuth>>,
1922 0 : build_info: BuildInfo,
1923 0 : ) -> RouterBuilder<hyper::Body, ApiError> {
1924 0 : let mut router = endpoint::make_router()
1925 0 : .middleware(prologue_leadership_status_check_middleware())
1926 0 : .middleware(prologue_metrics_middleware())
1927 0 : .middleware(epilogue_metrics_middleware());
1928 0 : if auth.is_some() {
1929 0 : router = router.middleware(auth_middleware(|request| {
1930 0 : let state = get_state(request);
1931 0 : if state.allowlist_routes.contains(&request.uri().path()) {
1932 0 : None
1933 : } else {
1934 0 : state.auth.as_deref()
1935 : }
1936 0 : }));
1937 0 : }
1938 :
1939 0 : router
1940 0 : .data(Arc::new(HttpState::new(service, auth, build_info)))
1941 0 : .get("/metrics", |r| {
1942 0 : named_request_span(r, measured_metrics_handler, RequestName("metrics"))
1943 0 : })
1944 0 : // Non-prefixed generic endpoints (status, metrics, profiling)
1945 0 : .get("/status", |r| {
1946 0 : named_request_span(r, handle_status, RequestName("status"))
1947 0 : })
1948 0 : .get("/ready", |r| {
1949 0 : named_request_span(r, handle_ready, RequestName("ready"))
1950 0 : })
1951 0 : .get("/profile/cpu", |r| {
1952 0 : named_request_span(r, profile_cpu_handler, RequestName("profile_cpu"))
1953 0 : })
1954 0 : .get("/profile/heap", |r| {
1955 0 : named_request_span(r, profile_heap_handler, RequestName("profile_heap"))
1956 0 : })
1957 0 : // Upcalls for the pageserver: point the pageserver's `control_plane_api` config to this prefix
1958 0 : .post("/upcall/v1/re-attach", |r| {
1959 0 : named_request_span(r, handle_re_attach, RequestName("upcall_v1_reattach"))
1960 0 : })
1961 0 : .post("/upcall/v1/validate", |r| {
1962 0 : named_request_span(r, handle_validate, RequestName("upcall_v1_validate"))
1963 0 : })
1964 0 : // Test/dev/debug endpoints
1965 0 : .post("/debug/v1/attach-hook", |r| {
1966 0 : named_request_span(r, handle_attach_hook, RequestName("debug_v1_attach_hook"))
1967 0 : })
1968 0 : .post("/debug/v1/inspect", |r| {
1969 0 : named_request_span(r, handle_inspect, RequestName("debug_v1_inspect"))
1970 0 : })
1971 0 : .post("/debug/v1/tenant/:tenant_id/drop", |r| {
1972 0 : named_request_span(r, handle_tenant_drop, RequestName("debug_v1_tenant_drop"))
1973 0 : })
1974 0 : .post("/debug/v1/node/:node_id/drop", |r| {
1975 0 : named_request_span(r, handle_node_drop, RequestName("debug_v1_node_drop"))
1976 0 : })
1977 0 : .post("/debug/v1/tenant/:tenant_id/import", |r| {
1978 0 : named_request_span(
1979 0 : r,
1980 0 : handle_tenant_import,
1981 0 : RequestName("debug_v1_tenant_import"),
1982 0 : )
1983 0 : })
1984 0 : .get("/debug/v1/tenant", |r| {
1985 0 : named_request_span(r, handle_tenants_dump, RequestName("debug_v1_tenant"))
1986 0 : })
1987 0 : .get("/debug/v1/tenant/:tenant_id/locate", |r| {
1988 0 : tenant_service_handler(
1989 0 : r,
1990 0 : handle_tenant_locate,
1991 0 : RequestName("debug_v1_tenant_locate"),
1992 0 : )
1993 0 : })
1994 0 : .post(
1995 0 : "/debug/v1/tenant/:tenant_id/timeline/:timeline_id/import",
1996 0 : |r| {
1997 0 : named_request_span(
1998 0 : r,
1999 0 : handle_timeline_import,
2000 0 : RequestName("debug_v1_timeline_import"),
2001 0 : )
2002 0 : },
2003 0 : )
2004 0 : .get("/debug/v1/scheduler", |r| {
2005 0 : named_request_span(r, handle_scheduler_dump, RequestName("debug_v1_scheduler"))
2006 0 : })
2007 0 : .post("/debug/v1/consistency_check", |r| {
2008 0 : named_request_span(
2009 0 : r,
2010 0 : handle_consistency_check,
2011 0 : RequestName("debug_v1_consistency_check"),
2012 0 : )
2013 0 : })
2014 0 : .post("/debug/v1/reconcile_all", |r| {
2015 0 : request_span(r, handle_reconcile_all)
2016 0 : })
2017 0 : .put("/debug/v1/failpoints", |r| {
2018 0 : request_span(r, |r| failpoints_handler(r, CancellationToken::new()))
2019 0 : })
2020 0 : // Node operations
2021 0 : .post("/control/v1/node", |r| {
2022 0 : named_request_span(r, handle_node_register, RequestName("control_v1_node"))
2023 0 : })
2024 0 : .delete("/control/v1/node/:node_id", |r| {
2025 0 : named_request_span(r, handle_node_delete, RequestName("control_v1_node_delete"))
2026 0 : })
2027 0 : .get("/control/v1/node", |r| {
2028 0 : named_request_span(r, handle_node_list, RequestName("control_v1_node"))
2029 0 : })
2030 0 : .put("/control/v1/node/:node_id/config", |r| {
2031 0 : named_request_span(
2032 0 : r,
2033 0 : handle_node_configure,
2034 0 : RequestName("control_v1_node_config"),
2035 0 : )
2036 0 : })
2037 0 : .get("/control/v1/node/:node_id", |r| {
2038 0 : named_request_span(r, handle_node_status, RequestName("control_v1_node_status"))
2039 0 : })
2040 0 : .get("/control/v1/node/:node_id/shards", |r| {
2041 0 : named_request_span(
2042 0 : r,
2043 0 : handle_node_shards,
2044 0 : RequestName("control_v1_node_describe"),
2045 0 : )
2046 0 : })
2047 0 : .get("/control/v1/leader", |r| {
2048 0 : named_request_span(r, handle_get_leader, RequestName("control_v1_get_leader"))
2049 0 : })
2050 0 : .put("/control/v1/node/:node_id/drain", |r| {
2051 0 : named_request_span(r, handle_node_drain, RequestName("control_v1_node_drain"))
2052 0 : })
2053 0 : .delete("/control/v1/node/:node_id/drain", |r| {
2054 0 : named_request_span(
2055 0 : r,
2056 0 : handle_cancel_node_drain,
2057 0 : RequestName("control_v1_cancel_node_drain"),
2058 0 : )
2059 0 : })
2060 0 : .put("/control/v1/node/:node_id/fill", |r| {
2061 0 : named_request_span(r, handle_node_fill, RequestName("control_v1_node_fill"))
2062 0 : })
2063 0 : .delete("/control/v1/node/:node_id/fill", |r| {
2064 0 : named_request_span(
2065 0 : r,
2066 0 : handle_cancel_node_fill,
2067 0 : RequestName("control_v1_cancel_node_fill"),
2068 0 : )
2069 0 : })
2070 0 : // Metadata health operations
2071 0 : .post("/control/v1/metadata_health/update", |r| {
2072 0 : named_request_span(
2073 0 : r,
2074 0 : handle_metadata_health_update,
2075 0 : RequestName("control_v1_metadata_health_update"),
2076 0 : )
2077 0 : })
2078 0 : .get("/control/v1/metadata_health/unhealthy", |r| {
2079 0 : named_request_span(
2080 0 : r,
2081 0 : handle_metadata_health_list_unhealthy,
2082 0 : RequestName("control_v1_metadata_health_list_unhealthy"),
2083 0 : )
2084 0 : })
2085 0 : .post("/control/v1/metadata_health/outdated", |r| {
2086 0 : named_request_span(
2087 0 : r,
2088 0 : handle_metadata_health_list_outdated,
2089 0 : RequestName("control_v1_metadata_health_list_outdated"),
2090 0 : )
2091 0 : })
2092 0 : // Safekeepers
2093 0 : .get("/control/v1/safekeeper", |r| {
2094 0 : named_request_span(
2095 0 : r,
2096 0 : handle_safekeeper_list,
2097 0 : RequestName("control_v1_safekeeper_list"),
2098 0 : )
2099 0 : })
2100 0 : .get("/control/v1/safekeeper/:id", |r| {
2101 0 : named_request_span(r, handle_get_safekeeper, RequestName("v1_safekeeper"))
2102 0 : })
2103 0 : .post("/control/v1/safekeeper/:id", |r| {
2104 0 : // id is in the body
2105 0 : named_request_span(
2106 0 : r,
2107 0 : handle_upsert_safekeeper,
2108 0 : RequestName("v1_safekeeper_post"),
2109 0 : )
2110 0 : })
2111 0 : .post("/control/v1/safekeeper/:id/scheduling_policy", |r| {
2112 0 : named_request_span(
2113 0 : r,
2114 0 : handle_safekeeper_scheduling_policy,
2115 0 : RequestName("v1_safekeeper_status"),
2116 0 : )
2117 0 : })
2118 0 : // Tenant Shard operations
2119 0 : .put("/control/v1/tenant/:tenant_shard_id/migrate", |r| {
2120 0 : tenant_service_handler(
2121 0 : r,
2122 0 : handle_tenant_shard_migrate,
2123 0 : RequestName("control_v1_tenant_migrate"),
2124 0 : )
2125 0 : })
2126 0 : .put(
2127 0 : "/control/v1/tenant/:tenant_shard_id/migrate_secondary",
2128 0 : |r| {
2129 0 : tenant_service_handler(
2130 0 : r,
2131 0 : handle_tenant_shard_migrate_secondary,
2132 0 : RequestName("control_v1_tenant_migrate_secondary"),
2133 0 : )
2134 0 : },
2135 0 : )
2136 0 : .put(
2137 0 : "/control/v1/tenant/:tenant_shard_id/cancel_reconcile",
2138 0 : |r| {
2139 0 : tenant_service_handler(
2140 0 : r,
2141 0 : handle_tenant_shard_cancel_reconcile,
2142 0 : RequestName("control_v1_tenant_cancel_reconcile"),
2143 0 : )
2144 0 : },
2145 0 : )
2146 0 : .put("/control/v1/tenant/:tenant_id/shard_split", |r| {
2147 0 : tenant_service_handler(
2148 0 : r,
2149 0 : handle_tenant_shard_split,
2150 0 : RequestName("control_v1_tenant_shard_split"),
2151 0 : )
2152 0 : })
2153 0 : .get("/control/v1/tenant/:tenant_id", |r| {
2154 0 : tenant_service_handler(
2155 0 : r,
2156 0 : handle_tenant_describe,
2157 0 : RequestName("control_v1_tenant_describe"),
2158 0 : )
2159 0 : })
2160 0 : .get("/control/v1/tenant", |r| {
2161 0 : tenant_service_handler(r, handle_tenant_list, RequestName("control_v1_tenant_list"))
2162 0 : })
2163 0 : .put("/control/v1/tenant/:tenant_id/policy", |r| {
2164 0 : named_request_span(
2165 0 : r,
2166 0 : handle_tenant_update_policy,
2167 0 : RequestName("control_v1_tenant_policy"),
2168 0 : )
2169 0 : })
2170 0 : .put("/control/v1/preferred_azs", |r| {
2171 0 : named_request_span(
2172 0 : r,
2173 0 : handle_update_preferred_azs,
2174 0 : RequestName("control_v1_preferred_azs"),
2175 0 : )
2176 0 : })
2177 0 : .put("/control/v1/step_down", |r| {
2178 0 : named_request_span(r, handle_step_down, RequestName("control_v1_step_down"))
2179 0 : })
2180 0 : // Tenant operations
2181 0 : // The ^/v1/ endpoints act as a "Virtual Pageserver", enabling shard-naive clients to call into
2182 0 : // this service to manage tenants that actually consist of many tenant shards, as if they are a single entity.
2183 0 : .post("/v1/tenant", |r| {
2184 0 : tenant_service_handler(r, handle_tenant_create, RequestName("v1_tenant"))
2185 0 : })
2186 0 : .delete("/v1/tenant/:tenant_id", |r| {
2187 0 : tenant_service_handler(r, handle_tenant_delete, RequestName("v1_tenant"))
2188 0 : })
2189 0 : .patch("/v1/tenant/config", |r| {
2190 0 : tenant_service_handler(
2191 0 : r,
2192 0 : handle_tenant_config_patch,
2193 0 : RequestName("v1_tenant_config"),
2194 0 : )
2195 0 : })
2196 0 : .put("/v1/tenant/config", |r| {
2197 0 : tenant_service_handler(r, handle_tenant_config_set, RequestName("v1_tenant_config"))
2198 0 : })
2199 0 : .get("/v1/tenant/:tenant_id/config", |r| {
2200 0 : tenant_service_handler(r, handle_tenant_config_get, RequestName("v1_tenant_config"))
2201 0 : })
2202 0 : .put("/v1/tenant/:tenant_shard_id/location_config", |r| {
2203 0 : tenant_service_handler(
2204 0 : r,
2205 0 : handle_tenant_location_config,
2206 0 : RequestName("v1_tenant_location_config"),
2207 0 : )
2208 0 : })
2209 0 : .put("/v1/tenant/:tenant_id/time_travel_remote_storage", |r| {
2210 0 : tenant_service_handler(
2211 0 : r,
2212 0 : handle_tenant_time_travel_remote_storage,
2213 0 : RequestName("v1_tenant_time_travel_remote_storage"),
2214 0 : )
2215 0 : })
2216 0 : .post("/v1/tenant/:tenant_id/secondary/download", |r| {
2217 0 : tenant_service_handler(
2218 0 : r,
2219 0 : handle_tenant_secondary_download,
2220 0 : RequestName("v1_tenant_secondary_download"),
2221 0 : )
2222 0 : })
2223 0 : // Timeline operations
2224 0 : .delete("/v1/tenant/:tenant_id/timeline/:timeline_id", |r| {
2225 0 : tenant_service_handler(
2226 0 : r,
2227 0 : handle_tenant_timeline_delete,
2228 0 : RequestName("v1_tenant_timeline"),
2229 0 : )
2230 0 : })
2231 0 : .post("/v1/tenant/:tenant_id/timeline", |r| {
2232 0 : tenant_service_handler(
2233 0 : r,
2234 0 : handle_tenant_timeline_create,
2235 0 : RequestName("v1_tenant_timeline"),
2236 0 : )
2237 0 : })
2238 0 : .put(
2239 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/archival_config",
2240 0 : |r| {
2241 0 : tenant_service_handler(
2242 0 : r,
2243 0 : handle_tenant_timeline_archival_config,
2244 0 : RequestName("v1_tenant_timeline_archival_config"),
2245 0 : )
2246 0 : },
2247 0 : )
2248 0 : .put(
2249 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/detach_ancestor",
2250 0 : |r| {
2251 0 : tenant_service_handler(
2252 0 : r,
2253 0 : handle_tenant_timeline_detach_ancestor,
2254 0 : RequestName("v1_tenant_timeline_detach_ancestor"),
2255 0 : )
2256 0 : },
2257 0 : )
2258 0 : .post(
2259 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/block_gc",
2260 0 : |r| {
2261 0 : tenant_service_handler(
2262 0 : r,
2263 0 : |s, r| handle_tenant_timeline_block_unblock_gc(s, r, BlockUnblock::Block),
2264 0 : RequestName("v1_tenant_timeline_block_unblock_gc"),
2265 0 : )
2266 0 : },
2267 0 : )
2268 0 : .post(
2269 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/unblock_gc",
2270 0 : |r| {
2271 0 : tenant_service_handler(
2272 0 : r,
2273 0 : |s, r| handle_tenant_timeline_block_unblock_gc(s, r, BlockUnblock::Unblock),
2274 0 : RequestName("v1_tenant_timeline_block_unblock_gc"),
2275 0 : )
2276 0 : },
2277 0 : )
2278 0 : .post(
2279 0 : "/v1/tenant/:tenant_shard_id/timeline/:timeline_id/download_heatmap_layers",
2280 0 : |r| {
2281 0 : tenant_service_handler(
2282 0 : r,
2283 0 : handle_tenant_timeline_download_heatmap_layers,
2284 0 : RequestName("v1_tenant_timeline_download_heatmap_layers"),
2285 0 : )
2286 0 : },
2287 0 : )
2288 0 : // LSN lease passthrough to all shards
2289 0 : .post(
2290 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/lsn_lease",
2291 0 : |r| {
2292 0 : tenant_service_handler(
2293 0 : r,
2294 0 : handle_tenant_timeline_lsn_lease,
2295 0 : RequestName("v1_tenant_timeline_lsn_lease"),
2296 0 : )
2297 0 : },
2298 0 : )
2299 0 : // Tenant detail GET passthrough to shard zero:
2300 0 : .get("/v1/tenant/:tenant_id", |r| {
2301 0 : tenant_service_handler(
2302 0 : r,
2303 0 : handle_tenant_timeline_passthrough,
2304 0 : RequestName("v1_tenant_passthrough"),
2305 0 : )
2306 0 : })
2307 0 : // The `*` in the URL is a wildcard: any tenant/timeline GET APIs on the pageserver
2308 0 : // are implicitly exposed here. This must be last in the list to avoid
2309 0 : // taking precedence over other GET methods we might implement by hand.
2310 0 : .get("/v1/tenant/:tenant_id/*", |r| {
2311 0 : tenant_service_handler(
2312 0 : r,
2313 0 : handle_tenant_timeline_passthrough,
2314 0 : RequestName("v1_tenant_passthrough"),
2315 0 : )
2316 0 : })
2317 0 : // Tenant timeline mark_invisible passthrough to shard zero
2318 0 : .put(
2319 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/mark_invisible",
2320 0 : |r| {
2321 0 : tenant_service_handler(
2322 0 : r,
2323 0 : handle_tenant_timeline_passthrough,
2324 0 : RequestName("v1_tenant_timeline_mark_invisible_passthrough"),
2325 0 : )
2326 0 : },
2327 0 : )
2328 0 : }
2329 :
2330 : #[cfg(test)]
2331 : mod test {
2332 :
2333 : use super::path_without_ids;
2334 :
2335 : #[test]
2336 1 : fn test_path_without_ids() {
2337 1 : assert_eq!(
2338 1 : path_without_ids(
2339 1 : "/v1/tenant/1a2b3344556677881122334455667788/timeline/AA223344556677881122334455667788"
2340 1 : ),
2341 1 : "/v1/tenant//timeline/"
2342 1 : );
2343 1 : assert_eq!(
2344 1 : path_without_ids(
2345 1 : "/v1/tenant/1a2b3344556677881122334455667788-0108/timeline/AA223344556677881122334455667788"
2346 1 : ),
2347 1 : "/v1/tenant//timeline/"
2348 1 : );
2349 1 : assert_eq!(
2350 1 : path_without_ids(
2351 1 : "/v1/tenant/1a2b3344556677881122334455667788-0108/timeline/AA223344556677881122334455667788?parameter=foo"
2352 1 : ),
2353 1 : "/v1/tenant//timeline/"
2354 1 : );
2355 1 : }
2356 : }
|
