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