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 : // On subsequent retries, wait longer.
385 : // Enable callers with a 25 second request timeout to reliably get a response
386 : const MAX_WAIT: Duration = Duration::from_secs(25);
387 : const MAX_RETRY_PERIOD: Duration = Duration::from_secs(5);
388 :
389 0 : let started_at = Instant::now();
390 0 :
391 0 : // To keep deletion reasonably snappy for small tenants, initially check after 1 second if deletion
392 0 : // completed.
393 0 : let mut retry_period = Duration::from_secs(1);
394 :
395 : loop {
396 0 : let status = f(service.clone()).await?;
397 0 : match status {
398 : StatusCode::ACCEPTED => {
399 0 : tracing::info!("Deletion accepted, waiting to try again...");
400 0 : tokio::time::sleep(retry_period).await;
401 0 : retry_period = MAX_RETRY_PERIOD;
402 : }
403 : StatusCode::CONFLICT => {
404 0 : tracing::info!("Deletion already in progress, waiting to try again...");
405 0 : tokio::time::sleep(retry_period).await;
406 : }
407 : StatusCode::NOT_FOUND => {
408 0 : tracing::info!("Deletion complete");
409 0 : return json_response(StatusCode::OK, ());
410 : }
411 : _ => {
412 0 : tracing::warn!("Unexpected status {status}");
413 0 : return json_response(status, ());
414 : }
415 : }
416 :
417 0 : let now = Instant::now();
418 0 : if now + retry_period > started_at + MAX_WAIT {
419 0 : tracing::info!("Deletion timed out waiting for 404");
420 : // REQUEST_TIMEOUT would be more appropriate, but CONFLICT is already part of
421 : // the pageserver's swagger definition for this endpoint, and has the same desired
422 : // effect of causing the control plane to retry later.
423 0 : return json_response(StatusCode::CONFLICT, ());
424 0 : }
425 : }
426 0 : }
427 :
428 0 : deletion_wrapper(service, move |service| async move {
429 0 : service
430 0 : .tenant_timeline_delete(tenant_id, timeline_id)
431 0 : .await
432 0 : .and_then(map_reqwest_hyper_status)
433 0 : })
434 0 : .await
435 0 : }
436 :
437 0 : async fn handle_tenant_timeline_archival_config(
438 0 : service: Arc<Service>,
439 0 : req: Request<Body>,
440 0 : ) -> Result<Response<Body>, ApiError> {
441 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
442 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
443 :
444 0 : check_permissions(&req, Scope::PageServerApi)?;
445 :
446 0 : let mut req = match maybe_forward(req).await {
447 0 : ForwardOutcome::Forwarded(res) => {
448 0 : return res;
449 : }
450 0 : ForwardOutcome::NotForwarded(req) => req,
451 : };
452 :
453 0 : let create_req = json_request::<TimelineArchivalConfigRequest>(&mut req).await?;
454 :
455 0 : service
456 0 : .tenant_timeline_archival_config(tenant_id, timeline_id, create_req)
457 0 : .await?;
458 :
459 0 : json_response(StatusCode::OK, ())
460 0 : }
461 :
462 0 : async fn handle_tenant_timeline_detach_ancestor(
463 0 : service: Arc<Service>,
464 0 : req: Request<Body>,
465 0 : ) -> Result<Response<Body>, ApiError> {
466 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
467 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
468 :
469 0 : check_permissions(&req, Scope::PageServerApi)?;
470 :
471 0 : match maybe_forward(req).await {
472 0 : ForwardOutcome::Forwarded(res) => {
473 0 : return res;
474 : }
475 0 : ForwardOutcome::NotForwarded(_req) => {}
476 : };
477 :
478 0 : let res = service
479 0 : .tenant_timeline_detach_ancestor(tenant_id, timeline_id)
480 0 : .await?;
481 :
482 0 : json_response(StatusCode::OK, res)
483 0 : }
484 :
485 0 : async fn handle_tenant_timeline_block_unblock_gc(
486 0 : service: Arc<Service>,
487 0 : req: Request<Body>,
488 0 : dir: BlockUnblock,
489 0 : ) -> Result<Response<Body>, ApiError> {
490 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
491 0 : check_permissions(&req, Scope::PageServerApi)?;
492 :
493 0 : let timeline_id: TimelineId = parse_request_param(&req, "timeline_id")?;
494 :
495 0 : service
496 0 : .tenant_timeline_block_unblock_gc(tenant_id, timeline_id, dir)
497 0 : .await?;
498 :
499 0 : json_response(StatusCode::OK, ())
500 0 : }
501 :
502 0 : async fn handle_tenant_timeline_passthrough(
503 0 : service: Arc<Service>,
504 0 : req: Request<Body>,
505 0 : ) -> Result<Response<Body>, ApiError> {
506 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
507 0 : check_permissions(&req, Scope::PageServerApi)?;
508 :
509 0 : let req = match maybe_forward(req).await {
510 0 : ForwardOutcome::Forwarded(res) => {
511 0 : return res;
512 : }
513 0 : ForwardOutcome::NotForwarded(req) => req,
514 : };
515 :
516 0 : let Some(path) = req.uri().path_and_query() else {
517 : // This should never happen, our request router only calls us if there is a path
518 0 : return Err(ApiError::BadRequest(anyhow::anyhow!("Missing path")));
519 : };
520 :
521 0 : tracing::info!("Proxying request for tenant {} ({})", tenant_id, path);
522 :
523 : // Find the node that holds shard zero
524 0 : let (node, tenant_shard_id) = service.tenant_shard0_node(tenant_id).await?;
525 :
526 : // Callers will always pass an unsharded tenant ID. Before proxying, we must
527 : // rewrite this to a shard-aware shard zero ID.
528 0 : let path = format!("{}", path);
529 0 : let tenant_str = tenant_id.to_string();
530 0 : let tenant_shard_str = format!("{}", tenant_shard_id);
531 0 : let path = path.replace(&tenant_str, &tenant_shard_str);
532 0 :
533 0 : let latency = &METRICS_REGISTRY
534 0 : .metrics_group
535 0 : .storage_controller_passthrough_request_latency;
536 0 :
537 0 : // This is a bit awkward. We remove the param from the request
538 0 : // and join the words by '_' to get a label for the request.
539 0 : let just_path = path.replace(&tenant_shard_str, "");
540 0 : let path_label = just_path
541 0 : .split('/')
542 0 : .filter(|token| !token.is_empty())
543 0 : .collect::<Vec<_>>()
544 0 : .join("_");
545 0 : let labels = PageserverRequestLabelGroup {
546 0 : pageserver_id: &node.get_id().to_string(),
547 0 : path: &path_label,
548 0 : method: crate::metrics::Method::Get,
549 0 : };
550 0 :
551 0 : let _timer = latency.start_timer(labels.clone());
552 0 :
553 0 : let client = mgmt_api::Client::new(node.base_url(), service.get_config().jwt_token.as_deref());
554 0 : let resp = client.get_raw(path).await.map_err(|e|
555 : // We return 503 here because if we can't successfully send a request to the pageserver,
556 : // either we aren't available or the pageserver is unavailable.
557 0 : ApiError::ResourceUnavailable(format!("Error sending pageserver API request to {node}: {e}").into()))?;
558 :
559 0 : if !resp.status().is_success() {
560 0 : let error_counter = &METRICS_REGISTRY
561 0 : .metrics_group
562 0 : .storage_controller_passthrough_request_error;
563 0 : error_counter.inc(labels);
564 0 : }
565 :
566 : // Transform 404 into 503 if we raced with a migration
567 0 : if resp.status() == reqwest::StatusCode::NOT_FOUND {
568 : // Look up node again: if we migrated it will be different
569 0 : let (new_node, _tenant_shard_id) = service.tenant_shard0_node(tenant_id).await?;
570 0 : if new_node.get_id() != node.get_id() {
571 : // Rather than retry here, send the client a 503 to prompt a retry: this matches
572 : // the pageserver's use of 503, and all clients calling this API should retry on 503.
573 0 : return Err(ApiError::ResourceUnavailable(
574 0 : format!("Pageserver {node} returned 404, was migrated to {new_node}").into(),
575 0 : ));
576 0 : }
577 0 : }
578 :
579 : // We have a reqest::Response, would like a http::Response
580 0 : let mut builder = hyper::Response::builder().status(map_reqwest_hyper_status(resp.status())?);
581 0 : for (k, v) in resp.headers() {
582 0 : builder = builder.header(k.as_str(), v.as_bytes());
583 0 : }
584 :
585 0 : let response = builder
586 0 : .body(Body::wrap_stream(resp.bytes_stream()))
587 0 : .map_err(|e| ApiError::InternalServerError(e.into()))?;
588 :
589 0 : Ok(response)
590 0 : }
591 :
592 0 : async fn handle_tenant_locate(
593 0 : service: Arc<Service>,
594 0 : req: Request<Body>,
595 0 : ) -> Result<Response<Body>, ApiError> {
596 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
597 :
598 0 : check_permissions(&req, Scope::Admin)?;
599 :
600 0 : match maybe_forward(req).await {
601 0 : ForwardOutcome::Forwarded(res) => {
602 0 : return res;
603 : }
604 0 : ForwardOutcome::NotForwarded(_req) => {}
605 0 : };
606 0 :
607 0 : json_response(StatusCode::OK, service.tenant_locate(tenant_id)?)
608 0 : }
609 :
610 0 : async fn handle_tenant_describe(
611 0 : service: Arc<Service>,
612 0 : req: Request<Body>,
613 0 : ) -> Result<Response<Body>, ApiError> {
614 0 : check_permissions(&req, Scope::Scrubber)?;
615 :
616 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
617 :
618 0 : match maybe_forward(req).await {
619 0 : ForwardOutcome::Forwarded(res) => {
620 0 : return res;
621 : }
622 0 : ForwardOutcome::NotForwarded(_req) => {}
623 0 : };
624 0 :
625 0 : json_response(StatusCode::OK, service.tenant_describe(tenant_id)?)
626 0 : }
627 :
628 0 : async fn handle_tenant_list(
629 0 : service: Arc<Service>,
630 0 : req: Request<Body>,
631 0 : ) -> Result<Response<Body>, ApiError> {
632 0 : check_permissions(&req, Scope::Admin)?;
633 :
634 0 : match maybe_forward(req).await {
635 0 : ForwardOutcome::Forwarded(res) => {
636 0 : return res;
637 : }
638 0 : ForwardOutcome::NotForwarded(_req) => {}
639 0 : };
640 0 :
641 0 : json_response(StatusCode::OK, service.tenant_list())
642 0 : }
643 :
644 0 : async fn handle_node_register(req: Request<Body>) -> Result<Response<Body>, ApiError> {
645 0 : check_permissions(&req, Scope::Infra)?;
646 :
647 0 : let mut req = match maybe_forward(req).await {
648 0 : ForwardOutcome::Forwarded(res) => {
649 0 : return res;
650 : }
651 0 : ForwardOutcome::NotForwarded(req) => req,
652 : };
653 :
654 0 : let register_req = json_request::<NodeRegisterRequest>(&mut req).await?;
655 0 : let state = get_state(&req);
656 0 : state.service.node_register(register_req).await?;
657 0 : json_response(StatusCode::OK, ())
658 0 : }
659 :
660 0 : async fn handle_node_list(req: Request<Body>) -> Result<Response<Body>, ApiError> {
661 0 : check_permissions(&req, Scope::Infra)?;
662 :
663 0 : let req = match maybe_forward(req).await {
664 0 : ForwardOutcome::Forwarded(res) => {
665 0 : return res;
666 : }
667 0 : ForwardOutcome::NotForwarded(req) => req,
668 0 : };
669 0 :
670 0 : let state = get_state(&req);
671 0 : let nodes = state.service.node_list().await?;
672 0 : let api_nodes = nodes.into_iter().map(|n| n.describe()).collect::<Vec<_>>();
673 0 :
674 0 : json_response(StatusCode::OK, api_nodes)
675 0 : }
676 :
677 0 : async fn handle_node_drop(req: Request<Body>) -> Result<Response<Body>, ApiError> {
678 0 : check_permissions(&req, Scope::Admin)?;
679 :
680 0 : let req = match maybe_forward(req).await {
681 0 : ForwardOutcome::Forwarded(res) => {
682 0 : return res;
683 : }
684 0 : ForwardOutcome::NotForwarded(req) => req,
685 0 : };
686 0 :
687 0 : let state = get_state(&req);
688 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
689 0 : json_response(StatusCode::OK, state.service.node_drop(node_id).await?)
690 0 : }
691 :
692 0 : async fn handle_node_delete(req: Request<Body>) -> Result<Response<Body>, ApiError> {
693 0 : check_permissions(&req, Scope::Admin)?;
694 :
695 0 : let req = match maybe_forward(req).await {
696 0 : ForwardOutcome::Forwarded(res) => {
697 0 : return res;
698 : }
699 0 : ForwardOutcome::NotForwarded(req) => req,
700 0 : };
701 0 :
702 0 : let state = get_state(&req);
703 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
704 0 : json_response(StatusCode::OK, state.service.node_delete(node_id).await?)
705 0 : }
706 :
707 0 : async fn handle_node_configure(req: Request<Body>) -> Result<Response<Body>, ApiError> {
708 0 : check_permissions(&req, Scope::Admin)?;
709 :
710 0 : let mut req = match maybe_forward(req).await {
711 0 : ForwardOutcome::Forwarded(res) => {
712 0 : return res;
713 : }
714 0 : ForwardOutcome::NotForwarded(req) => req,
715 : };
716 :
717 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
718 0 : let config_req = json_request::<NodeConfigureRequest>(&mut req).await?;
719 0 : if node_id != config_req.node_id {
720 0 : return Err(ApiError::BadRequest(anyhow::anyhow!(
721 0 : "Path and body node_id differ"
722 0 : )));
723 0 : }
724 0 : let state = get_state(&req);
725 0 :
726 0 : json_response(
727 0 : StatusCode::OK,
728 0 : state
729 0 : .service
730 0 : .external_node_configure(
731 0 : config_req.node_id,
732 0 : config_req.availability.map(NodeAvailability::from),
733 0 : config_req.scheduling,
734 0 : )
735 0 : .await?,
736 : )
737 0 : }
738 :
739 0 : async fn handle_node_status(req: Request<Body>) -> Result<Response<Body>, ApiError> {
740 0 : check_permissions(&req, Scope::Infra)?;
741 :
742 0 : let req = match maybe_forward(req).await {
743 0 : ForwardOutcome::Forwarded(res) => {
744 0 : return res;
745 : }
746 0 : ForwardOutcome::NotForwarded(req) => req,
747 0 : };
748 0 :
749 0 : let state = get_state(&req);
750 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
751 :
752 0 : let node_status = state.service.get_node(node_id).await?;
753 :
754 0 : json_response(StatusCode::OK, node_status)
755 0 : }
756 :
757 0 : async fn handle_node_shards(req: Request<Body>) -> Result<Response<Body>, ApiError> {
758 0 : check_permissions(&req, Scope::Admin)?;
759 :
760 0 : let state = get_state(&req);
761 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
762 :
763 0 : let node_status = state.service.get_node_shards(node_id).await?;
764 :
765 0 : json_response(StatusCode::OK, node_status)
766 0 : }
767 :
768 0 : async fn handle_get_leader(req: Request<Body>) -> Result<Response<Body>, ApiError> {
769 0 : check_permissions(&req, Scope::Admin)?;
770 :
771 0 : let req = match maybe_forward(req).await {
772 0 : ForwardOutcome::Forwarded(res) => {
773 0 : return res;
774 : }
775 0 : ForwardOutcome::NotForwarded(req) => req,
776 0 : };
777 0 :
778 0 : let state = get_state(&req);
779 0 : let leader = state.service.get_leader().await.map_err(|err| {
780 0 : ApiError::InternalServerError(anyhow::anyhow!(
781 0 : "Failed to read leader from database: {err}"
782 0 : ))
783 0 : })?;
784 :
785 0 : json_response(StatusCode::OK, leader)
786 0 : }
787 :
788 0 : async fn handle_node_drain(req: Request<Body>) -> Result<Response<Body>, ApiError> {
789 0 : check_permissions(&req, Scope::Infra)?;
790 :
791 0 : let req = match maybe_forward(req).await {
792 0 : ForwardOutcome::Forwarded(res) => {
793 0 : return res;
794 : }
795 0 : ForwardOutcome::NotForwarded(req) => req,
796 0 : };
797 0 :
798 0 : let state = get_state(&req);
799 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
800 :
801 0 : state.service.start_node_drain(node_id).await?;
802 :
803 0 : json_response(StatusCode::ACCEPTED, ())
804 0 : }
805 :
806 0 : async fn handle_cancel_node_drain(req: Request<Body>) -> Result<Response<Body>, ApiError> {
807 0 : check_permissions(&req, Scope::Infra)?;
808 :
809 0 : let req = match maybe_forward(req).await {
810 0 : ForwardOutcome::Forwarded(res) => {
811 0 : return res;
812 : }
813 0 : ForwardOutcome::NotForwarded(req) => req,
814 0 : };
815 0 :
816 0 : let state = get_state(&req);
817 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
818 :
819 0 : state.service.cancel_node_drain(node_id).await?;
820 :
821 0 : json_response(StatusCode::ACCEPTED, ())
822 0 : }
823 :
824 0 : async fn handle_node_fill(req: Request<Body>) -> Result<Response<Body>, ApiError> {
825 0 : check_permissions(&req, Scope::Infra)?;
826 :
827 0 : let req = match maybe_forward(req).await {
828 0 : ForwardOutcome::Forwarded(res) => {
829 0 : return res;
830 : }
831 0 : ForwardOutcome::NotForwarded(req) => req,
832 0 : };
833 0 :
834 0 : let state = get_state(&req);
835 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
836 :
837 0 : state.service.start_node_fill(node_id).await?;
838 :
839 0 : json_response(StatusCode::ACCEPTED, ())
840 0 : }
841 :
842 0 : async fn handle_cancel_node_fill(req: Request<Body>) -> Result<Response<Body>, ApiError> {
843 0 : check_permissions(&req, Scope::Infra)?;
844 :
845 0 : let req = match maybe_forward(req).await {
846 0 : ForwardOutcome::Forwarded(res) => {
847 0 : return res;
848 : }
849 0 : ForwardOutcome::NotForwarded(req) => req,
850 0 : };
851 0 :
852 0 : let state = get_state(&req);
853 0 : let node_id: NodeId = parse_request_param(&req, "node_id")?;
854 :
855 0 : state.service.cancel_node_fill(node_id).await?;
856 :
857 0 : json_response(StatusCode::ACCEPTED, ())
858 0 : }
859 :
860 0 : async fn handle_metadata_health_update(req: Request<Body>) -> Result<Response<Body>, ApiError> {
861 0 : check_permissions(&req, Scope::Scrubber)?;
862 :
863 0 : let mut req = match maybe_forward(req).await {
864 0 : ForwardOutcome::Forwarded(res) => {
865 0 : return res;
866 : }
867 0 : ForwardOutcome::NotForwarded(req) => req,
868 : };
869 :
870 0 : let update_req = json_request::<MetadataHealthUpdateRequest>(&mut req).await?;
871 0 : let state = get_state(&req);
872 0 :
873 0 : state.service.metadata_health_update(update_req).await?;
874 :
875 0 : json_response(StatusCode::OK, MetadataHealthUpdateResponse {})
876 0 : }
877 :
878 0 : async fn handle_metadata_health_list_unhealthy(
879 0 : req: Request<Body>,
880 0 : ) -> Result<Response<Body>, ApiError> {
881 0 : check_permissions(&req, Scope::Admin)?;
882 :
883 0 : let req = match maybe_forward(req).await {
884 0 : ForwardOutcome::Forwarded(res) => {
885 0 : return res;
886 : }
887 0 : ForwardOutcome::NotForwarded(req) => req,
888 0 : };
889 0 :
890 0 : let state = get_state(&req);
891 0 : let unhealthy_tenant_shards = state.service.metadata_health_list_unhealthy().await?;
892 :
893 0 : json_response(
894 0 : StatusCode::OK,
895 0 : MetadataHealthListUnhealthyResponse {
896 0 : unhealthy_tenant_shards,
897 0 : },
898 0 : )
899 0 : }
900 :
901 0 : async fn handle_metadata_health_list_outdated(
902 0 : req: Request<Body>,
903 0 : ) -> Result<Response<Body>, ApiError> {
904 0 : check_permissions(&req, Scope::Admin)?;
905 :
906 0 : let mut req = match maybe_forward(req).await {
907 0 : ForwardOutcome::Forwarded(res) => {
908 0 : return res;
909 : }
910 0 : ForwardOutcome::NotForwarded(req) => req,
911 : };
912 :
913 0 : let list_outdated_req = json_request::<MetadataHealthListOutdatedRequest>(&mut req).await?;
914 0 : let state = get_state(&req);
915 0 : let health_records = state
916 0 : .service
917 0 : .metadata_health_list_outdated(list_outdated_req.not_scrubbed_for)
918 0 : .await?;
919 :
920 0 : json_response(
921 0 : StatusCode::OK,
922 0 : MetadataHealthListOutdatedResponse { health_records },
923 0 : )
924 0 : }
925 :
926 0 : async fn handle_tenant_shard_split(
927 0 : service: Arc<Service>,
928 0 : req: Request<Body>,
929 0 : ) -> Result<Response<Body>, ApiError> {
930 0 : check_permissions(&req, Scope::Admin)?;
931 :
932 0 : let mut req = match maybe_forward(req).await {
933 0 : ForwardOutcome::Forwarded(res) => {
934 0 : return res;
935 : }
936 0 : ForwardOutcome::NotForwarded(req) => req,
937 : };
938 :
939 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
940 0 : let split_req = json_request::<TenantShardSplitRequest>(&mut req).await?;
941 :
942 : json_response(
943 : StatusCode::OK,
944 0 : service.tenant_shard_split(tenant_id, split_req).await?,
945 : )
946 0 : }
947 :
948 0 : async fn handle_tenant_shard_migrate(
949 0 : service: Arc<Service>,
950 0 : req: Request<Body>,
951 0 : ) -> Result<Response<Body>, ApiError> {
952 0 : check_permissions(&req, Scope::Admin)?;
953 :
954 0 : let mut req = match maybe_forward(req).await {
955 0 : ForwardOutcome::Forwarded(res) => {
956 0 : return res;
957 : }
958 0 : ForwardOutcome::NotForwarded(req) => req,
959 : };
960 :
961 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
962 0 : let migrate_req = json_request::<TenantShardMigrateRequest>(&mut req).await?;
963 : json_response(
964 : StatusCode::OK,
965 0 : service
966 0 : .tenant_shard_migrate(tenant_shard_id, migrate_req)
967 0 : .await?,
968 : )
969 0 : }
970 :
971 0 : async fn handle_tenant_shard_cancel_reconcile(
972 0 : service: Arc<Service>,
973 0 : req: Request<Body>,
974 0 : ) -> Result<Response<Body>, ApiError> {
975 0 : check_permissions(&req, Scope::Admin)?;
976 :
977 0 : let req = match maybe_forward(req).await {
978 0 : ForwardOutcome::Forwarded(res) => {
979 0 : return res;
980 : }
981 0 : ForwardOutcome::NotForwarded(req) => req,
982 : };
983 :
984 0 : let tenant_shard_id: TenantShardId = parse_request_param(&req, "tenant_shard_id")?;
985 : json_response(
986 : StatusCode::OK,
987 0 : service
988 0 : .tenant_shard_cancel_reconcile(tenant_shard_id)
989 0 : .await?,
990 : )
991 0 : }
992 :
993 0 : async fn handle_tenant_update_policy(req: Request<Body>) -> Result<Response<Body>, ApiError> {
994 0 : check_permissions(&req, Scope::Admin)?;
995 :
996 0 : let mut req = match maybe_forward(req).await {
997 0 : ForwardOutcome::Forwarded(res) => {
998 0 : return res;
999 : }
1000 0 : ForwardOutcome::NotForwarded(req) => req,
1001 : };
1002 :
1003 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1004 0 : let update_req = json_request::<TenantPolicyRequest>(&mut req).await?;
1005 0 : let state = get_state(&req);
1006 0 :
1007 0 : json_response(
1008 0 : StatusCode::OK,
1009 0 : state
1010 0 : .service
1011 0 : .tenant_update_policy(tenant_id, update_req)
1012 0 : .await?,
1013 : )
1014 0 : }
1015 :
1016 0 : async fn handle_update_preferred_azs(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1017 0 : check_permissions(&req, Scope::Admin)?;
1018 :
1019 0 : let mut req = match maybe_forward(req).await {
1020 0 : ForwardOutcome::Forwarded(res) => {
1021 0 : return res;
1022 : }
1023 0 : ForwardOutcome::NotForwarded(req) => req,
1024 : };
1025 :
1026 0 : let azs_req = json_request::<ShardsPreferredAzsRequest>(&mut req).await?;
1027 0 : let state = get_state(&req);
1028 0 :
1029 0 : json_response(
1030 0 : StatusCode::OK,
1031 0 : state.service.update_shards_preferred_azs(azs_req).await?,
1032 : )
1033 0 : }
1034 :
1035 0 : async fn handle_step_down(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1036 0 : check_permissions(&req, Scope::ControllerPeer)?;
1037 :
1038 0 : let req = match maybe_forward(req).await {
1039 0 : ForwardOutcome::Forwarded(res) => {
1040 0 : return res;
1041 : }
1042 0 : ForwardOutcome::NotForwarded(req) => req,
1043 0 : };
1044 0 :
1045 0 : let state = get_state(&req);
1046 0 : json_response(StatusCode::OK, state.service.step_down().await)
1047 0 : }
1048 :
1049 0 : async fn handle_tenant_drop(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1050 0 : check_permissions(&req, Scope::PageServerApi)?;
1051 :
1052 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1053 :
1054 0 : let req = match maybe_forward(req).await {
1055 0 : ForwardOutcome::Forwarded(res) => {
1056 0 : return res;
1057 : }
1058 0 : ForwardOutcome::NotForwarded(req) => req,
1059 0 : };
1060 0 :
1061 0 : let state = get_state(&req);
1062 0 :
1063 0 : json_response(StatusCode::OK, state.service.tenant_drop(tenant_id).await?)
1064 0 : }
1065 :
1066 0 : async fn handle_tenant_import(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1067 0 : check_permissions(&req, Scope::PageServerApi)?;
1068 :
1069 0 : let tenant_id: TenantId = parse_request_param(&req, "tenant_id")?;
1070 :
1071 0 : let req = match maybe_forward(req).await {
1072 0 : ForwardOutcome::Forwarded(res) => {
1073 0 : return res;
1074 : }
1075 0 : ForwardOutcome::NotForwarded(req) => req,
1076 0 : };
1077 0 :
1078 0 : let state = get_state(&req);
1079 0 :
1080 0 : json_response(
1081 0 : StatusCode::OK,
1082 0 : state.service.tenant_import(tenant_id).await?,
1083 : )
1084 0 : }
1085 :
1086 0 : async fn handle_tenants_dump(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 : state.service.tenants_dump()
1098 0 : }
1099 :
1100 0 : async fn handle_scheduler_dump(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1101 0 : check_permissions(&req, Scope::Admin)?;
1102 :
1103 0 : let req = match maybe_forward(req).await {
1104 0 : ForwardOutcome::Forwarded(res) => {
1105 0 : return res;
1106 : }
1107 0 : ForwardOutcome::NotForwarded(req) => req,
1108 0 : };
1109 0 :
1110 0 : let state = get_state(&req);
1111 0 : state.service.scheduler_dump()
1112 0 : }
1113 :
1114 0 : async fn handle_consistency_check(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1115 0 : check_permissions(&req, Scope::Admin)?;
1116 :
1117 0 : let req = match maybe_forward(req).await {
1118 0 : ForwardOutcome::Forwarded(res) => {
1119 0 : return res;
1120 : }
1121 0 : ForwardOutcome::NotForwarded(req) => req,
1122 0 : };
1123 0 :
1124 0 : let state = get_state(&req);
1125 0 :
1126 0 : json_response(StatusCode::OK, state.service.consistency_check().await?)
1127 0 : }
1128 :
1129 0 : async fn handle_reconcile_all(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1130 0 : check_permissions(&req, Scope::Admin)?;
1131 :
1132 0 : let req = match maybe_forward(req).await {
1133 0 : ForwardOutcome::Forwarded(res) => {
1134 0 : return res;
1135 : }
1136 0 : ForwardOutcome::NotForwarded(req) => req,
1137 0 : };
1138 0 :
1139 0 : let state = get_state(&req);
1140 0 :
1141 0 : json_response(StatusCode::OK, state.service.reconcile_all_now().await?)
1142 0 : }
1143 :
1144 : /// Status endpoint is just used for checking that our HTTP listener is up
1145 0 : async fn handle_status(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1146 0 : match maybe_forward(req).await {
1147 0 : ForwardOutcome::Forwarded(res) => {
1148 0 : return res;
1149 : }
1150 0 : ForwardOutcome::NotForwarded(_req) => {}
1151 0 : };
1152 0 :
1153 0 : json_response(StatusCode::OK, ())
1154 0 : }
1155 :
1156 : /// Readiness endpoint indicates when we're done doing startup I/O (e.g. reconciling
1157 : /// with remote pageserver nodes). This is intended for use as a kubernetes readiness probe.
1158 0 : async fn handle_ready(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1159 0 : let req = match maybe_forward(req).await {
1160 0 : ForwardOutcome::Forwarded(res) => {
1161 0 : return res;
1162 : }
1163 0 : ForwardOutcome::NotForwarded(req) => req,
1164 0 : };
1165 0 :
1166 0 : let state = get_state(&req);
1167 0 : if state.service.startup_complete.is_ready() {
1168 0 : json_response(StatusCode::OK, ())
1169 : } else {
1170 0 : json_response(StatusCode::SERVICE_UNAVAILABLE, ())
1171 : }
1172 0 : }
1173 :
1174 : impl From<ReconcileError> for ApiError {
1175 0 : fn from(value: ReconcileError) -> Self {
1176 0 : ApiError::Conflict(format!("Reconciliation error: {}", value))
1177 0 : }
1178 : }
1179 :
1180 : /// Return the safekeeper record by instance id, or 404.
1181 : ///
1182 : /// Not used by anything except manual testing.
1183 0 : async fn handle_get_safekeeper(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1184 0 : check_permissions(&req, Scope::Admin)?;
1185 :
1186 0 : let id = parse_request_param::<i64>(&req, "id")?;
1187 :
1188 0 : let req = match maybe_forward(req).await {
1189 0 : ForwardOutcome::Forwarded(res) => {
1190 0 : return res;
1191 : }
1192 0 : ForwardOutcome::NotForwarded(req) => req,
1193 0 : };
1194 0 :
1195 0 : let state = get_state(&req);
1196 :
1197 0 : let res = state.service.get_safekeeper(id).await;
1198 :
1199 0 : match res {
1200 0 : Ok(b) => json_response(StatusCode::OK, b),
1201 : Err(crate::persistence::DatabaseError::Query(diesel::result::Error::NotFound)) => {
1202 0 : Err(ApiError::NotFound("unknown instance_id".into()))
1203 : }
1204 0 : Err(other) => Err(other.into()),
1205 : }
1206 0 : }
1207 :
1208 : /// Used as part of deployment scripts.
1209 : ///
1210 : /// Assumes information is only relayed to storage controller after first selecting an unique id on
1211 : /// control plane database, which means we have an id field in the request and payload.
1212 0 : async fn handle_upsert_safekeeper(mut req: Request<Body>) -> Result<Response<Body>, ApiError> {
1213 0 : check_permissions(&req, Scope::Infra)?;
1214 :
1215 0 : let body = json_request::<SafekeeperPersistence>(&mut req).await?;
1216 0 : let id = parse_request_param::<i64>(&req, "id")?;
1217 :
1218 0 : if id != body.id {
1219 : // it should be repeated
1220 0 : return Err(ApiError::BadRequest(anyhow::anyhow!(
1221 0 : "id mismatch: url={id:?}, body={:?}",
1222 0 : body.id
1223 0 : )));
1224 0 : }
1225 :
1226 0 : let req = match maybe_forward(req).await {
1227 0 : ForwardOutcome::Forwarded(res) => {
1228 0 : return res;
1229 : }
1230 0 : ForwardOutcome::NotForwarded(req) => req,
1231 0 : };
1232 0 :
1233 0 : let state = get_state(&req);
1234 0 :
1235 0 : state.service.upsert_safekeeper(body).await?;
1236 :
1237 0 : Ok(Response::builder()
1238 0 : .status(StatusCode::NO_CONTENT)
1239 0 : .body(Body::empty())
1240 0 : .unwrap())
1241 0 : }
1242 :
1243 : /// Common wrapper for request handlers that call into Service and will operate on tenants: they must only
1244 : /// be allowed to run if Service has finished its initial reconciliation.
1245 0 : async fn tenant_service_handler<R, H>(
1246 0 : request: Request<Body>,
1247 0 : handler: H,
1248 0 : request_name: RequestName,
1249 0 : ) -> R::Output
1250 0 : where
1251 0 : R: std::future::Future<Output = Result<Response<Body>, ApiError>> + Send + 'static,
1252 0 : H: FnOnce(Arc<Service>, Request<Body>) -> R + Send + Sync + 'static,
1253 0 : {
1254 0 : let state = get_state(&request);
1255 0 : let service = state.service.clone();
1256 0 :
1257 0 : let startup_complete = service.startup_complete.clone();
1258 0 : if tokio::time::timeout(STARTUP_RECONCILE_TIMEOUT, startup_complete.wait())
1259 0 : .await
1260 0 : .is_err()
1261 : {
1262 : // This shouldn't happen: it is the responsibilty of [`Service::startup_reconcile`] to use appropriate
1263 : // timeouts around its remote calls, to bound its runtime.
1264 0 : return Err(ApiError::Timeout(
1265 0 : "Timed out waiting for service readiness".into(),
1266 0 : ));
1267 0 : }
1268 0 :
1269 0 : named_request_span(
1270 0 : request,
1271 0 : |request| async move { handler(service, request).await },
1272 0 : request_name,
1273 0 : )
1274 0 : .await
1275 0 : }
1276 :
1277 : /// Check if the required scope is held in the request's token, or if the request has
1278 : /// a token with 'admin' scope then always permit it.
1279 0 : fn check_permissions(request: &Request<Body>, required_scope: Scope) -> Result<(), ApiError> {
1280 0 : check_permission_with(request, |claims| {
1281 0 : match crate::auth::check_permission(claims, required_scope) {
1282 0 : Err(e) => match crate::auth::check_permission(claims, Scope::Admin) {
1283 0 : Ok(()) => Ok(()),
1284 0 : Err(_) => Err(e),
1285 : },
1286 0 : Ok(()) => Ok(()),
1287 : }
1288 0 : })
1289 0 : }
1290 :
1291 : #[derive(Clone, Debug)]
1292 : struct RequestMeta {
1293 : method: hyper::http::Method,
1294 : at: Instant,
1295 : }
1296 :
1297 0 : pub fn prologue_leadership_status_check_middleware<
1298 0 : B: hyper::body::HttpBody + Send + Sync + 'static,
1299 0 : >() -> Middleware<B, ApiError> {
1300 0 : Middleware::pre(move |req| async move {
1301 0 : let state = get_state(&req);
1302 0 : let leadership_status = state.service.get_leadership_status();
1303 :
1304 : enum AllowedRoutes<'a> {
1305 : All,
1306 : Some(Vec<&'a str>),
1307 : }
1308 :
1309 0 : let allowed_routes = match leadership_status {
1310 0 : LeadershipStatus::Leader => AllowedRoutes::All,
1311 0 : LeadershipStatus::SteppedDown => AllowedRoutes::All,
1312 : LeadershipStatus::Candidate => {
1313 0 : AllowedRoutes::Some(["/ready", "/status", "/metrics"].to_vec())
1314 : }
1315 : };
1316 :
1317 0 : let uri = req.uri().to_string();
1318 0 : match allowed_routes {
1319 0 : AllowedRoutes::All => Ok(req),
1320 0 : AllowedRoutes::Some(allowed) if allowed.contains(&uri.as_str()) => Ok(req),
1321 : _ => {
1322 0 : tracing::info!(
1323 0 : "Request {} not allowed due to current leadership state",
1324 0 : req.uri()
1325 : );
1326 :
1327 0 : Err(ApiError::ResourceUnavailable(
1328 0 : format!("Current leadership status is {leadership_status}").into(),
1329 0 : ))
1330 : }
1331 : }
1332 0 : })
1333 0 : }
1334 :
1335 0 : fn prologue_metrics_middleware<B: hyper::body::HttpBody + Send + Sync + 'static>(
1336 0 : ) -> Middleware<B, ApiError> {
1337 0 : Middleware::pre(move |req| async move {
1338 0 : let meta = RequestMeta {
1339 0 : method: req.method().clone(),
1340 0 : at: Instant::now(),
1341 0 : };
1342 0 :
1343 0 : req.set_context(meta);
1344 0 :
1345 0 : Ok(req)
1346 0 : })
1347 0 : }
1348 :
1349 0 : fn epilogue_metrics_middleware<B: hyper::body::HttpBody + Send + Sync + 'static>(
1350 0 : ) -> Middleware<B, ApiError> {
1351 0 : Middleware::post_with_info(move |resp, req_info| async move {
1352 0 : let request_name = match req_info.context::<RequestName>() {
1353 0 : Some(name) => name,
1354 : None => {
1355 0 : return Ok(resp);
1356 : }
1357 : };
1358 :
1359 0 : if let Some(meta) = req_info.context::<RequestMeta>() {
1360 0 : let status = &crate::metrics::METRICS_REGISTRY
1361 0 : .metrics_group
1362 0 : .storage_controller_http_request_status;
1363 0 : let latency = &crate::metrics::METRICS_REGISTRY
1364 0 : .metrics_group
1365 0 : .storage_controller_http_request_latency;
1366 0 :
1367 0 : status.inc(HttpRequestStatusLabelGroup {
1368 0 : path: request_name.0,
1369 0 : method: meta.method.clone().into(),
1370 0 : status: crate::metrics::StatusCode(resp.status()),
1371 0 : });
1372 0 :
1373 0 : latency.observe(
1374 0 : HttpRequestLatencyLabelGroup {
1375 0 : path: request_name.0,
1376 0 : method: meta.method.into(),
1377 0 : },
1378 0 : meta.at.elapsed().as_secs_f64(),
1379 0 : );
1380 0 : }
1381 0 : Ok(resp)
1382 0 : })
1383 0 : }
1384 :
1385 0 : pub async fn measured_metrics_handler(req: Request<Body>) -> Result<Response<Body>, ApiError> {
1386 : pub const TEXT_FORMAT: &str = "text/plain; version=0.0.4";
1387 :
1388 0 : let req = match maybe_forward(req).await {
1389 0 : ForwardOutcome::Forwarded(res) => {
1390 0 : return res;
1391 : }
1392 0 : ForwardOutcome::NotForwarded(req) => req,
1393 0 : };
1394 0 :
1395 0 : let state = get_state(&req);
1396 0 : let payload = crate::metrics::METRICS_REGISTRY.encode(&state.neon_metrics);
1397 0 : let response = Response::builder()
1398 0 : .status(200)
1399 0 : .header(CONTENT_TYPE, TEXT_FORMAT)
1400 0 : .body(payload.into())
1401 0 : .unwrap();
1402 0 :
1403 0 : Ok(response)
1404 0 : }
1405 :
1406 : #[derive(Clone)]
1407 : struct RequestName(&'static str);
1408 :
1409 0 : async fn named_request_span<R, H>(
1410 0 : request: Request<Body>,
1411 0 : handler: H,
1412 0 : name: RequestName,
1413 0 : ) -> R::Output
1414 0 : where
1415 0 : R: Future<Output = Result<Response<Body>, ApiError>> + Send + 'static,
1416 0 : H: FnOnce(Request<Body>) -> R + Send + Sync + 'static,
1417 0 : {
1418 0 : request.set_context(name);
1419 0 : request_span(request, handler).await
1420 0 : }
1421 :
1422 : enum ForwardOutcome {
1423 : Forwarded(Result<Response<Body>, ApiError>),
1424 : NotForwarded(Request<Body>),
1425 : }
1426 :
1427 : /// Potentially forward the request to the current storage controler leader.
1428 : /// More specifically we forward when:
1429 : /// 1. Request is not one of ["/control/v1/step_down", "/status", "/ready", "/metrics"]
1430 : /// 2. Current instance is in [`LeadershipStatus::SteppedDown`] state
1431 : /// 3. There is a leader in the database to forward to
1432 : /// 4. Leader from step (3) is not the current instance
1433 : ///
1434 : /// Why forward?
1435 : /// It turns out that we can't rely on external orchestration to promptly route trafic to the
1436 : /// new leader. This is downtime inducing. Forwarding provides a safe way out.
1437 : ///
1438 : /// Why is it safe?
1439 : /// If a storcon instance is persisted in the database, then we know that it is the current leader.
1440 : /// There's one exception: time between handling step-down request and the new leader updating the
1441 : /// database.
1442 : ///
1443 : /// Let's treat the happy case first. The stepped down node does not produce any side effects,
1444 : /// since all request handling happens on the leader.
1445 : ///
1446 : /// As for the edge case, we are guaranteed to always have a maximum of two running instances.
1447 : /// Hence, if we are in the edge case scenario the leader persisted in the database is the
1448 : /// stepped down instance that received the request. Condition (4) above covers this scenario.
1449 0 : async fn maybe_forward(req: Request<Body>) -> ForwardOutcome {
1450 : const NOT_FOR_FORWARD: [&str; 4] = ["/control/v1/step_down", "/status", "/ready", "/metrics"];
1451 :
1452 0 : let uri = req.uri().to_string();
1453 0 : let uri_for_forward = !NOT_FOR_FORWARD.contains(&uri.as_str());
1454 0 :
1455 0 : let state = get_state(&req);
1456 0 : let leadership_status = state.service.get_leadership_status();
1457 0 :
1458 0 : if leadership_status != LeadershipStatus::SteppedDown || !uri_for_forward {
1459 0 : return ForwardOutcome::NotForwarded(req);
1460 0 : }
1461 :
1462 0 : let leader = state.service.get_leader().await;
1463 0 : let leader = {
1464 0 : match leader {
1465 0 : Ok(Some(leader)) => leader,
1466 : Ok(None) => {
1467 0 : return ForwardOutcome::Forwarded(Err(ApiError::ResourceUnavailable(
1468 0 : "No leader to forward to while in stepped down state".into(),
1469 0 : )));
1470 : }
1471 0 : Err(err) => {
1472 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(
1473 0 : anyhow::anyhow!(
1474 0 : "Failed to get leader for forwarding while in stepped down state: {err}"
1475 0 : ),
1476 0 : )));
1477 : }
1478 : }
1479 : };
1480 :
1481 0 : let cfg = state.service.get_config();
1482 0 : if let Some(ref self_addr) = cfg.address_for_peers {
1483 0 : let leader_addr = match Uri::from_str(leader.address.as_str()) {
1484 0 : Ok(uri) => uri,
1485 0 : Err(err) => {
1486 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(
1487 0 : anyhow::anyhow!(
1488 0 : "Failed to parse leader uri for forwarding while in stepped down state: {err}"
1489 0 : ),
1490 0 : )));
1491 : }
1492 : };
1493 :
1494 0 : if *self_addr == leader_addr {
1495 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(anyhow::anyhow!(
1496 0 : "Leader is stepped down instance"
1497 0 : ))));
1498 0 : }
1499 0 : }
1500 :
1501 0 : tracing::info!("Forwarding {} to leader at {}", uri, leader.address);
1502 :
1503 : // Use [`RECONCILE_TIMEOUT`] as the max amount of time a request should block for and
1504 : // include some leeway to get the timeout for proxied requests.
1505 : const PROXIED_REQUEST_TIMEOUT: Duration = Duration::from_secs(RECONCILE_TIMEOUT.as_secs() + 10);
1506 0 : let client = reqwest::ClientBuilder::new()
1507 0 : .timeout(PROXIED_REQUEST_TIMEOUT)
1508 0 : .build();
1509 0 : let client = match client {
1510 0 : Ok(client) => client,
1511 0 : Err(err) => {
1512 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(anyhow::anyhow!(
1513 0 : "Failed to build leader client for forwarding while in stepped down state: {err}"
1514 0 : ))));
1515 : }
1516 : };
1517 :
1518 0 : let request: reqwest::Request = match convert_request(req, &client, leader.address).await {
1519 0 : Ok(r) => r,
1520 0 : Err(err) => {
1521 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(anyhow::anyhow!(
1522 0 : "Failed to convert request for forwarding while in stepped down state: {err}"
1523 0 : ))));
1524 : }
1525 : };
1526 :
1527 0 : let response = match client.execute(request).await {
1528 0 : Ok(r) => r,
1529 0 : Err(err) => {
1530 0 : return ForwardOutcome::Forwarded(Err(ApiError::InternalServerError(anyhow::anyhow!(
1531 0 : "Failed to forward while in stepped down state: {err}"
1532 0 : ))));
1533 : }
1534 : };
1535 :
1536 0 : ForwardOutcome::Forwarded(convert_response(response).await)
1537 0 : }
1538 :
1539 : /// Convert a [`reqwest::Response`] to a [hyper::Response`] by passing through
1540 : /// a stable representation (string, bytes or integer)
1541 : ///
1542 : /// Ideally, we would not have to do this since both types use the http crate
1543 : /// under the hood. However, they use different versions of the crate and keeping
1544 : /// second order dependencies in sync is difficult.
1545 0 : async fn convert_response(resp: reqwest::Response) -> Result<hyper::Response<Body>, ApiError> {
1546 : use std::str::FromStr;
1547 :
1548 0 : let mut builder = hyper::Response::builder().status(resp.status().as_u16());
1549 0 : for (key, value) in resp.headers().into_iter() {
1550 0 : let key = hyper::header::HeaderName::from_str(key.as_str()).map_err(|err| {
1551 0 : ApiError::InternalServerError(anyhow::anyhow!("Response conversion failed: {err}"))
1552 0 : })?;
1553 :
1554 0 : let value = hyper::header::HeaderValue::from_bytes(value.as_bytes()).map_err(|err| {
1555 0 : ApiError::InternalServerError(anyhow::anyhow!("Response conversion failed: {err}"))
1556 0 : })?;
1557 :
1558 0 : builder = builder.header(key, value);
1559 : }
1560 :
1561 0 : let body = http::Body::wrap_stream(resp.bytes_stream());
1562 0 :
1563 0 : builder.body(body).map_err(|err| {
1564 0 : ApiError::InternalServerError(anyhow::anyhow!("Response conversion failed: {err}"))
1565 0 : })
1566 0 : }
1567 :
1568 : /// Convert a [`reqwest::Request`] to a [hyper::Request`] by passing through
1569 : /// a stable representation (string, bytes or integer)
1570 : ///
1571 : /// See [`convert_response`] for why we are doing it this way.
1572 0 : async fn convert_request(
1573 0 : req: hyper::Request<Body>,
1574 0 : client: &reqwest::Client,
1575 0 : to_address: String,
1576 0 : ) -> Result<reqwest::Request, ApiError> {
1577 : use std::str::FromStr;
1578 :
1579 0 : let (parts, body) = req.into_parts();
1580 0 : let method = reqwest::Method::from_str(parts.method.as_str()).map_err(|err| {
1581 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1582 0 : })?;
1583 :
1584 0 : let path_and_query = parts.uri.path_and_query().ok_or_else(|| {
1585 0 : ApiError::InternalServerError(anyhow::anyhow!(
1586 0 : "Request conversion failed: no path and query"
1587 0 : ))
1588 0 : })?;
1589 :
1590 0 : let uri = reqwest::Url::from_str(
1591 0 : format!(
1592 0 : "{}{}",
1593 0 : to_address.trim_end_matches("/"),
1594 0 : path_and_query.as_str()
1595 0 : )
1596 0 : .as_str(),
1597 0 : )
1598 0 : .map_err(|err| {
1599 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1600 0 : })?;
1601 :
1602 0 : let mut headers = reqwest::header::HeaderMap::new();
1603 0 : for (key, value) in parts.headers.into_iter() {
1604 0 : let key = match key {
1605 0 : Some(k) => k,
1606 : None => {
1607 0 : continue;
1608 : }
1609 : };
1610 :
1611 0 : let key = reqwest::header::HeaderName::from_str(key.as_str()).map_err(|err| {
1612 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1613 0 : })?;
1614 :
1615 0 : let value = reqwest::header::HeaderValue::from_bytes(value.as_bytes()).map_err(|err| {
1616 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1617 0 : })?;
1618 :
1619 0 : headers.insert(key, value);
1620 : }
1621 :
1622 0 : let body = hyper::body::to_bytes(body).await.map_err(|err| {
1623 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1624 0 : })?;
1625 :
1626 0 : client
1627 0 : .request(method, uri)
1628 0 : .headers(headers)
1629 0 : .body(body)
1630 0 : .build()
1631 0 : .map_err(|err| {
1632 0 : ApiError::InternalServerError(anyhow::anyhow!("Request conversion failed: {err}"))
1633 0 : })
1634 0 : }
1635 :
1636 0 : pub fn make_router(
1637 0 : service: Arc<Service>,
1638 0 : auth: Option<Arc<SwappableJwtAuth>>,
1639 0 : build_info: BuildInfo,
1640 0 : ) -> RouterBuilder<hyper::Body, ApiError> {
1641 0 : let mut router = endpoint::make_router()
1642 0 : .middleware(prologue_leadership_status_check_middleware())
1643 0 : .middleware(prologue_metrics_middleware())
1644 0 : .middleware(epilogue_metrics_middleware());
1645 0 : if auth.is_some() {
1646 0 : router = router.middleware(auth_middleware(|request| {
1647 0 : let state = get_state(request);
1648 0 : if state.allowlist_routes.contains(request.uri()) {
1649 0 : None
1650 : } else {
1651 0 : state.auth.as_deref()
1652 : }
1653 0 : }));
1654 0 : }
1655 :
1656 0 : router
1657 0 : .data(Arc::new(HttpState::new(service, auth, build_info)))
1658 0 : .get("/metrics", |r| {
1659 0 : named_request_span(r, measured_metrics_handler, RequestName("metrics"))
1660 0 : })
1661 0 : // Non-prefixed generic endpoints (status, metrics)
1662 0 : .get("/status", |r| {
1663 0 : named_request_span(r, handle_status, RequestName("status"))
1664 0 : })
1665 0 : .get("/ready", |r| {
1666 0 : named_request_span(r, handle_ready, RequestName("ready"))
1667 0 : })
1668 0 : // Upcalls for the pageserver: point the pageserver's `control_plane_api` config to this prefix
1669 0 : .post("/upcall/v1/re-attach", |r| {
1670 0 : named_request_span(r, handle_re_attach, RequestName("upcall_v1_reattach"))
1671 0 : })
1672 0 : .post("/upcall/v1/validate", |r| {
1673 0 : named_request_span(r, handle_validate, RequestName("upcall_v1_validate"))
1674 0 : })
1675 0 : // Test/dev/debug endpoints
1676 0 : .post("/debug/v1/attach-hook", |r| {
1677 0 : named_request_span(r, handle_attach_hook, RequestName("debug_v1_attach_hook"))
1678 0 : })
1679 0 : .post("/debug/v1/inspect", |r| {
1680 0 : named_request_span(r, handle_inspect, RequestName("debug_v1_inspect"))
1681 0 : })
1682 0 : .post("/debug/v1/tenant/:tenant_id/drop", |r| {
1683 0 : named_request_span(r, handle_tenant_drop, RequestName("debug_v1_tenant_drop"))
1684 0 : })
1685 0 : .post("/debug/v1/node/:node_id/drop", |r| {
1686 0 : named_request_span(r, handle_node_drop, RequestName("debug_v1_node_drop"))
1687 0 : })
1688 0 : .post("/debug/v1/tenant/:tenant_id/import", |r| {
1689 0 : named_request_span(
1690 0 : r,
1691 0 : handle_tenant_import,
1692 0 : RequestName("debug_v1_tenant_import"),
1693 0 : )
1694 0 : })
1695 0 : .get("/debug/v1/tenant", |r| {
1696 0 : named_request_span(r, handle_tenants_dump, RequestName("debug_v1_tenant"))
1697 0 : })
1698 0 : .get("/debug/v1/tenant/:tenant_id/locate", |r| {
1699 0 : tenant_service_handler(
1700 0 : r,
1701 0 : handle_tenant_locate,
1702 0 : RequestName("debug_v1_tenant_locate"),
1703 0 : )
1704 0 : })
1705 0 : .get("/debug/v1/scheduler", |r| {
1706 0 : named_request_span(r, handle_scheduler_dump, RequestName("debug_v1_scheduler"))
1707 0 : })
1708 0 : .post("/debug/v1/consistency_check", |r| {
1709 0 : named_request_span(
1710 0 : r,
1711 0 : handle_consistency_check,
1712 0 : RequestName("debug_v1_consistency_check"),
1713 0 : )
1714 0 : })
1715 0 : .post("/debug/v1/reconcile_all", |r| {
1716 0 : request_span(r, handle_reconcile_all)
1717 0 : })
1718 0 : .put("/debug/v1/failpoints", |r| {
1719 0 : request_span(r, |r| failpoints_handler(r, CancellationToken::new()))
1720 0 : })
1721 0 : // Node operations
1722 0 : .post("/control/v1/node", |r| {
1723 0 : named_request_span(r, handle_node_register, RequestName("control_v1_node"))
1724 0 : })
1725 0 : .delete("/control/v1/node/:node_id", |r| {
1726 0 : named_request_span(r, handle_node_delete, RequestName("control_v1_node_delete"))
1727 0 : })
1728 0 : .get("/control/v1/node", |r| {
1729 0 : named_request_span(r, handle_node_list, RequestName("control_v1_node"))
1730 0 : })
1731 0 : .put("/control/v1/node/:node_id/config", |r| {
1732 0 : named_request_span(
1733 0 : r,
1734 0 : handle_node_configure,
1735 0 : RequestName("control_v1_node_config"),
1736 0 : )
1737 0 : })
1738 0 : .get("/control/v1/node/:node_id", |r| {
1739 0 : named_request_span(r, handle_node_status, RequestName("control_v1_node_status"))
1740 0 : })
1741 0 : .get("/control/v1/node/:node_id/shards", |r| {
1742 0 : named_request_span(
1743 0 : r,
1744 0 : handle_node_shards,
1745 0 : RequestName("control_v1_node_describe"),
1746 0 : )
1747 0 : })
1748 0 : .get("/control/v1/leader", |r| {
1749 0 : named_request_span(r, handle_get_leader, RequestName("control_v1_get_leader"))
1750 0 : })
1751 0 : .put("/control/v1/node/:node_id/drain", |r| {
1752 0 : named_request_span(r, handle_node_drain, RequestName("control_v1_node_drain"))
1753 0 : })
1754 0 : .delete("/control/v1/node/:node_id/drain", |r| {
1755 0 : named_request_span(
1756 0 : r,
1757 0 : handle_cancel_node_drain,
1758 0 : RequestName("control_v1_cancel_node_drain"),
1759 0 : )
1760 0 : })
1761 0 : .put("/control/v1/node/:node_id/fill", |r| {
1762 0 : named_request_span(r, handle_node_fill, RequestName("control_v1_node_fill"))
1763 0 : })
1764 0 : .delete("/control/v1/node/:node_id/fill", |r| {
1765 0 : named_request_span(
1766 0 : r,
1767 0 : handle_cancel_node_fill,
1768 0 : RequestName("control_v1_cancel_node_fill"),
1769 0 : )
1770 0 : })
1771 0 : // Metadata health operations
1772 0 : .post("/control/v1/metadata_health/update", |r| {
1773 0 : named_request_span(
1774 0 : r,
1775 0 : handle_metadata_health_update,
1776 0 : RequestName("control_v1_metadata_health_update"),
1777 0 : )
1778 0 : })
1779 0 : .get("/control/v1/metadata_health/unhealthy", |r| {
1780 0 : named_request_span(
1781 0 : r,
1782 0 : handle_metadata_health_list_unhealthy,
1783 0 : RequestName("control_v1_metadata_health_list_unhealthy"),
1784 0 : )
1785 0 : })
1786 0 : .post("/control/v1/metadata_health/outdated", |r| {
1787 0 : named_request_span(
1788 0 : r,
1789 0 : handle_metadata_health_list_outdated,
1790 0 : RequestName("control_v1_metadata_health_list_outdated"),
1791 0 : )
1792 0 : })
1793 0 : // Tenant Shard operations
1794 0 : .put("/control/v1/tenant/:tenant_shard_id/migrate", |r| {
1795 0 : tenant_service_handler(
1796 0 : r,
1797 0 : handle_tenant_shard_migrate,
1798 0 : RequestName("control_v1_tenant_migrate"),
1799 0 : )
1800 0 : })
1801 0 : .put(
1802 0 : "/control/v1/tenant/:tenant_shard_id/cancel_reconcile",
1803 0 : |r| {
1804 0 : tenant_service_handler(
1805 0 : r,
1806 0 : handle_tenant_shard_cancel_reconcile,
1807 0 : RequestName("control_v1_tenant_cancel_reconcile"),
1808 0 : )
1809 0 : },
1810 0 : )
1811 0 : .put("/control/v1/tenant/:tenant_id/shard_split", |r| {
1812 0 : tenant_service_handler(
1813 0 : r,
1814 0 : handle_tenant_shard_split,
1815 0 : RequestName("control_v1_tenant_shard_split"),
1816 0 : )
1817 0 : })
1818 0 : .get("/control/v1/tenant/:tenant_id", |r| {
1819 0 : tenant_service_handler(
1820 0 : r,
1821 0 : handle_tenant_describe,
1822 0 : RequestName("control_v1_tenant_describe"),
1823 0 : )
1824 0 : })
1825 0 : .get("/control/v1/tenant", |r| {
1826 0 : tenant_service_handler(r, handle_tenant_list, RequestName("control_v1_tenant_list"))
1827 0 : })
1828 0 : .put("/control/v1/tenant/:tenant_id/policy", |r| {
1829 0 : named_request_span(
1830 0 : r,
1831 0 : handle_tenant_update_policy,
1832 0 : RequestName("control_v1_tenant_policy"),
1833 0 : )
1834 0 : })
1835 0 : .put("/control/v1/preferred_azs", |r| {
1836 0 : named_request_span(
1837 0 : r,
1838 0 : handle_update_preferred_azs,
1839 0 : RequestName("control_v1_preferred_azs"),
1840 0 : )
1841 0 : })
1842 0 : .put("/control/v1/step_down", |r| {
1843 0 : named_request_span(r, handle_step_down, RequestName("control_v1_step_down"))
1844 0 : })
1845 0 : .get("/control/v1/safekeeper/:id", |r| {
1846 0 : named_request_span(r, handle_get_safekeeper, RequestName("v1_safekeeper"))
1847 0 : })
1848 0 : .post("/control/v1/safekeeper/:id", |r| {
1849 0 : // id is in the body
1850 0 : named_request_span(r, handle_upsert_safekeeper, RequestName("v1_safekeeper"))
1851 0 : })
1852 0 : // Tenant operations
1853 0 : // The ^/v1/ endpoints act as a "Virtual Pageserver", enabling shard-naive clients to call into
1854 0 : // this service to manage tenants that actually consist of many tenant shards, as if they are a single entity.
1855 0 : .post("/v1/tenant", |r| {
1856 0 : tenant_service_handler(r, handle_tenant_create, RequestName("v1_tenant"))
1857 0 : })
1858 0 : .delete("/v1/tenant/:tenant_id", |r| {
1859 0 : tenant_service_handler(r, handle_tenant_delete, RequestName("v1_tenant"))
1860 0 : })
1861 0 : .put("/v1/tenant/config", |r| {
1862 0 : tenant_service_handler(r, handle_tenant_config_set, RequestName("v1_tenant_config"))
1863 0 : })
1864 0 : .get("/v1/tenant/:tenant_id/config", |r| {
1865 0 : tenant_service_handler(r, handle_tenant_config_get, RequestName("v1_tenant_config"))
1866 0 : })
1867 0 : .put("/v1/tenant/:tenant_shard_id/location_config", |r| {
1868 0 : tenant_service_handler(
1869 0 : r,
1870 0 : handle_tenant_location_config,
1871 0 : RequestName("v1_tenant_location_config"),
1872 0 : )
1873 0 : })
1874 0 : .put("/v1/tenant/:tenant_id/time_travel_remote_storage", |r| {
1875 0 : tenant_service_handler(
1876 0 : r,
1877 0 : handle_tenant_time_travel_remote_storage,
1878 0 : RequestName("v1_tenant_time_travel_remote_storage"),
1879 0 : )
1880 0 : })
1881 0 : .post("/v1/tenant/:tenant_id/secondary/download", |r| {
1882 0 : tenant_service_handler(
1883 0 : r,
1884 0 : handle_tenant_secondary_download,
1885 0 : RequestName("v1_tenant_secondary_download"),
1886 0 : )
1887 0 : })
1888 0 : // Timeline operations
1889 0 : .delete("/v1/tenant/:tenant_id/timeline/:timeline_id", |r| {
1890 0 : tenant_service_handler(
1891 0 : r,
1892 0 : handle_tenant_timeline_delete,
1893 0 : RequestName("v1_tenant_timeline"),
1894 0 : )
1895 0 : })
1896 0 : .post("/v1/tenant/:tenant_id/timeline", |r| {
1897 0 : tenant_service_handler(
1898 0 : r,
1899 0 : handle_tenant_timeline_create,
1900 0 : RequestName("v1_tenant_timeline"),
1901 0 : )
1902 0 : })
1903 0 : .put(
1904 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/archival_config",
1905 0 : |r| {
1906 0 : tenant_service_handler(
1907 0 : r,
1908 0 : handle_tenant_timeline_archival_config,
1909 0 : RequestName("v1_tenant_timeline_archival_config"),
1910 0 : )
1911 0 : },
1912 0 : )
1913 0 : .put(
1914 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/detach_ancestor",
1915 0 : |r| {
1916 0 : tenant_service_handler(
1917 0 : r,
1918 0 : handle_tenant_timeline_detach_ancestor,
1919 0 : RequestName("v1_tenant_timeline_detach_ancestor"),
1920 0 : )
1921 0 : },
1922 0 : )
1923 0 : .post(
1924 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/block_gc",
1925 0 : |r| {
1926 0 : tenant_service_handler(
1927 0 : r,
1928 0 : |s, r| handle_tenant_timeline_block_unblock_gc(s, r, BlockUnblock::Block),
1929 0 : RequestName("v1_tenant_timeline_block_unblock_gc"),
1930 0 : )
1931 0 : },
1932 0 : )
1933 0 : .post(
1934 0 : "/v1/tenant/:tenant_id/timeline/:timeline_id/unblock_gc",
1935 0 : |r| {
1936 0 : tenant_service_handler(
1937 0 : r,
1938 0 : |s, r| handle_tenant_timeline_block_unblock_gc(s, r, BlockUnblock::Unblock),
1939 0 : RequestName("v1_tenant_timeline_block_unblock_gc"),
1940 0 : )
1941 0 : },
1942 0 : )
1943 0 : // Tenant detail GET passthrough to shard zero:
1944 0 : .get("/v1/tenant/:tenant_id", |r| {
1945 0 : tenant_service_handler(
1946 0 : r,
1947 0 : handle_tenant_timeline_passthrough,
1948 0 : RequestName("v1_tenant_passthrough"),
1949 0 : )
1950 0 : })
1951 0 : // The `*` in the URL is a wildcard: any tenant/timeline GET APIs on the pageserver
1952 0 : // are implicitly exposed here. This must be last in the list to avoid
1953 0 : // taking precedence over other GET methods we might implement by hand.
1954 0 : .get("/v1/tenant/:tenant_id/*", |r| {
1955 0 : tenant_service_handler(
1956 0 : r,
1957 0 : handle_tenant_timeline_passthrough,
1958 0 : RequestName("v1_tenant_passthrough"),
1959 0 : )
1960 0 : })
1961 0 : }
|
