Line data Source code
1 : //! This module acts as a switchboard to access different repositories managed by this
2 : //! page server.
3 :
4 : use camino::{Utf8DirEntry, Utf8Path, Utf8PathBuf};
5 : use futures::StreamExt;
6 : use itertools::Itertools;
7 : use pageserver_api::key::Key;
8 : use pageserver_api::models::LocationConfigMode;
9 : use pageserver_api::shard::{
10 : ShardCount, ShardIdentity, ShardIndex, ShardNumber, ShardStripeSize, TenantShardId,
11 : };
12 : use pageserver_api::upcall_api::ReAttachResponseTenant;
13 : use rand::{distributions::Alphanumeric, Rng};
14 : use std::borrow::Cow;
15 : use std::cmp::Ordering;
16 : use std::collections::{BTreeMap, HashMap, HashSet};
17 : use std::ops::Deref;
18 : use std::sync::Arc;
19 : use std::time::Duration;
20 : use sysinfo::SystemExt;
21 : use tokio::fs;
22 :
23 : use anyhow::Context;
24 : use once_cell::sync::Lazy;
25 : use tokio::task::JoinSet;
26 : use tokio_util::sync::CancellationToken;
27 : use tracing::*;
28 :
29 : use utils::{backoff, completion, crashsafe};
30 :
31 : use crate::config::PageServerConf;
32 : use crate::context::{DownloadBehavior, RequestContext};
33 : use crate::control_plane_client::{
34 : ControlPlaneClient, ControlPlaneGenerationsApi, RetryForeverError,
35 : };
36 : use crate::deletion_queue::DeletionQueueClient;
37 : use crate::http::routes::ACTIVE_TENANT_TIMEOUT;
38 : use crate::metrics::{TENANT, TENANT_MANAGER as METRICS};
39 : use crate::task_mgr::{TaskKind, BACKGROUND_RUNTIME};
40 : use crate::tenant::config::{
41 : AttachedLocationConfig, AttachmentMode, LocationConf, LocationMode, SecondaryLocationConfig,
42 : };
43 : use crate::tenant::span::debug_assert_current_span_has_tenant_id;
44 : use crate::tenant::storage_layer::inmemory_layer;
45 : use crate::tenant::timeline::ShutdownMode;
46 : use crate::tenant::{AttachedTenantConf, GcError, LoadConfigError, SpawnMode, Tenant, TenantState};
47 : use crate::virtual_file::MaybeFatalIo;
48 : use crate::{InitializationOrder, TEMP_FILE_SUFFIX};
49 :
50 : use utils::crashsafe::path_with_suffix_extension;
51 : use utils::fs_ext::PathExt;
52 : use utils::generation::Generation;
53 : use utils::id::{TenantId, TimelineId};
54 :
55 : use super::remote_timeline_client::remote_tenant_path;
56 : use super::secondary::SecondaryTenant;
57 : use super::timeline::detach_ancestor::{self, PreparedTimelineDetach};
58 : use super::{GlobalShutDown, TenantSharedResources};
59 :
60 : /// For a tenant that appears in TenantsMap, it may either be
61 : /// - `Attached`: has a full Tenant object, is elegible to service
62 : /// reads and ingest WAL.
63 : /// - `Secondary`: is only keeping a local cache warm.
64 : ///
65 : /// Secondary is a totally distinct state rather than being a mode of a `Tenant`, because
66 : /// that way we avoid having to carefully switch a tenant's ingestion etc on and off during
67 : /// its lifetime, and we can preserve some important safety invariants like `Tenant` always
68 : /// having a properly acquired generation (Secondary doesn't need a generation)
69 : #[derive(Clone)]
70 : pub(crate) enum TenantSlot {
71 : Attached(Arc<Tenant>),
72 : Secondary(Arc<SecondaryTenant>),
73 : /// In this state, other administrative operations acting on the TenantId should
74 : /// block, or return a retry indicator equivalent to HTTP 503.
75 : InProgress(utils::completion::Barrier),
76 : }
77 :
78 : impl std::fmt::Debug for TenantSlot {
79 0 : fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
80 0 : match self {
81 0 : Self::Attached(tenant) => write!(f, "Attached({})", tenant.current_state()),
82 0 : Self::Secondary(_) => write!(f, "Secondary"),
83 0 : Self::InProgress(_) => write!(f, "InProgress"),
84 : }
85 0 : }
86 : }
87 :
88 : impl TenantSlot {
89 : /// Return the `Tenant` in this slot if attached, else None
90 0 : fn get_attached(&self) -> Option<&Arc<Tenant>> {
91 0 : match self {
92 0 : Self::Attached(t) => Some(t),
93 0 : Self::Secondary(_) => None,
94 0 : Self::InProgress(_) => None,
95 : }
96 0 : }
97 : }
98 :
99 : /// The tenants known to the pageserver.
100 : /// The enum variants are used to distinguish the different states that the pageserver can be in.
101 : pub(crate) enum TenantsMap {
102 : /// [`init_tenant_mgr`] is not done yet.
103 : Initializing,
104 : /// [`init_tenant_mgr`] is done, all on-disk tenants have been loaded.
105 : /// New tenants can be added using [`tenant_map_acquire_slot`].
106 : Open(BTreeMap<TenantShardId, TenantSlot>),
107 : /// The pageserver has entered shutdown mode via [`TenantManager::shutdown`].
108 : /// Existing tenants are still accessible, but no new tenants can be created.
109 : ShuttingDown(BTreeMap<TenantShardId, TenantSlot>),
110 : }
111 :
112 : /// When resolving a TenantId to a shard, we may be looking for the 0th
113 : /// shard, or we might be looking for whichever shard holds a particular page.
114 : #[derive(Copy, Clone)]
115 : pub(crate) enum ShardSelector {
116 : /// Only return the 0th shard, if it is present. If a non-0th shard is present,
117 : /// ignore it.
118 : Zero,
119 : /// Pick the shard that holds this key
120 : Page(Key),
121 : /// The shard ID is known: pick the given shard
122 : Known(ShardIndex),
123 : }
124 :
125 : /// A convenience for use with the re_attach ControlPlaneClient function: rather
126 : /// than the serializable struct, we build this enum that encapsulates
127 : /// the invariant that attached tenants always have generations.
128 : ///
129 : /// This represents the subset of a LocationConfig that we receive during re-attach.
130 : pub(crate) enum TenantStartupMode {
131 : Attached((AttachmentMode, Generation)),
132 : Secondary,
133 : }
134 :
135 : impl TenantStartupMode {
136 : /// Return the generation & mode that should be used when starting
137 : /// this tenant.
138 : ///
139 : /// If this returns None, the re-attach struct is in an invalid state and
140 : /// should be ignored in the response.
141 0 : fn from_reattach_tenant(rart: ReAttachResponseTenant) -> Option<Self> {
142 0 : match (rart.mode, rart.gen) {
143 0 : (LocationConfigMode::Detached, _) => None,
144 0 : (LocationConfigMode::Secondary, _) => Some(Self::Secondary),
145 0 : (LocationConfigMode::AttachedMulti, Some(g)) => {
146 0 : Some(Self::Attached((AttachmentMode::Multi, Generation::new(g))))
147 : }
148 0 : (LocationConfigMode::AttachedSingle, Some(g)) => {
149 0 : Some(Self::Attached((AttachmentMode::Single, Generation::new(g))))
150 : }
151 0 : (LocationConfigMode::AttachedStale, Some(g)) => {
152 0 : Some(Self::Attached((AttachmentMode::Stale, Generation::new(g))))
153 : }
154 : _ => {
155 0 : tracing::warn!(
156 0 : "Received invalid re-attach state for tenant {}: {rart:?}",
157 : rart.id
158 : );
159 0 : None
160 : }
161 : }
162 0 : }
163 : }
164 :
165 : /// Result type for looking up a TenantId to a specific shard
166 : pub(crate) enum ShardResolveResult {
167 : NotFound,
168 : Found(Arc<Tenant>),
169 : // Wait for this barrrier, then query again
170 : InProgress(utils::completion::Barrier),
171 : }
172 :
173 : impl TenantsMap {
174 : /// Convenience function for typical usage, where we want to get a `Tenant` object, for
175 : /// working with attached tenants. If the TenantId is in the map but in Secondary state,
176 : /// None is returned.
177 0 : pub(crate) fn get(&self, tenant_shard_id: &TenantShardId) -> Option<&Arc<Tenant>> {
178 0 : match self {
179 0 : TenantsMap::Initializing => None,
180 0 : TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => {
181 0 : m.get(tenant_shard_id).and_then(|slot| slot.get_attached())
182 : }
183 : }
184 0 : }
185 :
186 : #[cfg(all(debug_assertions, not(test)))]
187 0 : pub(crate) fn len(&self) -> usize {
188 0 : match self {
189 0 : TenantsMap::Initializing => 0,
190 0 : TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => m.len(),
191 : }
192 0 : }
193 : }
194 :
195 : /// Precursor to deletion of a tenant dir: we do a fast rename to a tmp path, and then
196 : /// the slower actual deletion in the background.
197 : ///
198 : /// This is "safe" in that that it won't leave behind a partially deleted directory
199 : /// at the original path, because we rename with TEMP_FILE_SUFFIX before starting deleting
200 : /// the contents.
201 : ///
202 : /// This is pageserver-specific, as it relies on future processes after a crash to check
203 : /// for TEMP_FILE_SUFFIX when loading things.
204 0 : async fn safe_rename_tenant_dir(path: impl AsRef<Utf8Path>) -> std::io::Result<Utf8PathBuf> {
205 0 : let parent = path
206 0 : .as_ref()
207 0 : .parent()
208 0 : // It is invalid to call this function with a relative path. Tenant directories
209 0 : // should always have a parent.
210 0 : .ok_or(std::io::Error::new(
211 0 : std::io::ErrorKind::InvalidInput,
212 0 : "Path must be absolute",
213 0 : ))?;
214 0 : let rand_suffix = rand::thread_rng()
215 0 : .sample_iter(&Alphanumeric)
216 0 : .take(8)
217 0 : .map(char::from)
218 0 : .collect::<String>()
219 0 : + TEMP_FILE_SUFFIX;
220 0 : let tmp_path = path_with_suffix_extension(&path, &rand_suffix);
221 0 : fs::rename(path.as_ref(), &tmp_path).await?;
222 0 : fs::File::open(parent).await?.sync_all().await?;
223 0 : Ok(tmp_path)
224 0 : }
225 :
226 : /// See [`Self::spawn`].
227 : #[derive(Clone, Default)]
228 : pub struct BackgroundPurges(tokio_util::task::TaskTracker);
229 :
230 : impl BackgroundPurges {
231 : /// When we have moved a tenant's content to a temporary directory, we may delete it lazily in
232 : /// the background, and thereby avoid blocking any API requests on this deletion completing.
233 : ///
234 : /// Although we are cleaning up the tenant, this task is not meant to be bound by the lifetime of the tenant in memory.
235 : /// Thus the [`BackgroundPurges`] type to keep track of these tasks.
236 0 : pub fn spawn(&self, tmp_path: Utf8PathBuf) {
237 0 : // because on shutdown we close and wait, we are misusing TaskTracker a bit.
238 0 : //
239 0 : // so first acquire a token, then check if the tracker has been closed. the tracker might get closed
240 0 : // right after, but at least the shutdown will wait for what we are spawning next.
241 0 : let token = self.0.token();
242 0 :
243 0 : if self.0.is_closed() {
244 0 : warn!(
245 : %tmp_path,
246 0 : "trying to spawn background purge during shutdown, ignoring"
247 : );
248 0 : return;
249 0 : }
250 :
251 0 : let span = info_span!(parent: None, "background_purge", %tmp_path);
252 :
253 0 : let task = move || {
254 0 : let _token = token;
255 0 : let _entered = span.entered();
256 0 : if let Err(error) = std::fs::remove_dir_all(tmp_path.as_path()) {
257 : // should we fatal_io_error here?
258 0 : warn!(%error, "failed to purge tenant directory");
259 0 : }
260 0 : };
261 :
262 0 : BACKGROUND_RUNTIME.spawn_blocking(task);
263 0 : }
264 :
265 : /// When this future completes, all background purges have completed.
266 : /// The first poll of the future will already lock out new background purges spawned via [`Self::spawn`].
267 : ///
268 : /// Concurrent calls will coalesce.
269 : ///
270 : /// # Cancellation-Safety
271 : ///
272 : /// If this future is dropped before polled to completion, concurrent and subsequent
273 : /// instances of this future will continue to be correct.
274 0 : #[instrument(skip_all)]
275 : pub async fn shutdown(&self) {
276 : // forbid new tasks (can be called many times)
277 : self.0.close();
278 : self.0.wait().await;
279 : }
280 : }
281 :
282 : static TENANTS: Lazy<std::sync::RwLock<TenantsMap>> =
283 6 : Lazy::new(|| std::sync::RwLock::new(TenantsMap::Initializing));
284 :
285 : /// Responsible for storing and mutating the collection of all tenants
286 : /// that this pageserver has state for.
287 : ///
288 : /// Every Tenant and SecondaryTenant instance lives inside the TenantManager.
289 : ///
290 : /// The most important role of the TenantManager is to prevent conflicts: e.g. trying to attach
291 : /// the same tenant twice concurrently, or trying to configure the same tenant into secondary
292 : /// and attached modes concurrently.
293 : pub struct TenantManager {
294 : conf: &'static PageServerConf,
295 : // TODO: currently this is a &'static pointing to TENANTs. When we finish refactoring
296 : // out of that static variable, the TenantManager can own this.
297 : // See https://github.com/neondatabase/neon/issues/5796
298 : tenants: &'static std::sync::RwLock<TenantsMap>,
299 : resources: TenantSharedResources,
300 :
301 : // Long-running operations that happen outside of a [`Tenant`] lifetime should respect this token.
302 : // This is for edge cases like tenant deletion. In normal cases (within a Tenant lifetime),
303 : // tenants have their own cancellation tokens, which we fire individually in [`Self::shutdown`], or
304 : // when the tenant detaches.
305 : cancel: CancellationToken,
306 :
307 : background_purges: BackgroundPurges,
308 : }
309 :
310 0 : fn emergency_generations(
311 0 : tenant_confs: &HashMap<TenantShardId, Result<LocationConf, LoadConfigError>>,
312 0 : ) -> HashMap<TenantShardId, TenantStartupMode> {
313 0 : tenant_confs
314 0 : .iter()
315 0 : .filter_map(|(tid, lc)| {
316 0 : let lc = match lc {
317 0 : Ok(lc) => lc,
318 0 : Err(_) => return None,
319 : };
320 : Some((
321 0 : *tid,
322 0 : match &lc.mode {
323 0 : LocationMode::Attached(alc) => {
324 0 : TenantStartupMode::Attached((alc.attach_mode, alc.generation))
325 : }
326 0 : LocationMode::Secondary(_) => TenantStartupMode::Secondary,
327 : },
328 : ))
329 0 : })
330 0 : .collect()
331 0 : }
332 :
333 0 : async fn init_load_generations(
334 0 : conf: &'static PageServerConf,
335 0 : tenant_confs: &HashMap<TenantShardId, Result<LocationConf, LoadConfigError>>,
336 0 : resources: &TenantSharedResources,
337 0 : cancel: &CancellationToken,
338 0 : ) -> anyhow::Result<Option<HashMap<TenantShardId, TenantStartupMode>>> {
339 0 : let generations = if conf.control_plane_emergency_mode {
340 0 : error!(
341 0 : "Emergency mode! Tenants will be attached unsafely using their last known generation"
342 : );
343 0 : emergency_generations(tenant_confs)
344 0 : } else if let Some(client) = ControlPlaneClient::new(conf, cancel) {
345 0 : info!("Calling control plane API to re-attach tenants");
346 : // If we are configured to use the control plane API, then it is the source of truth for what tenants to load.
347 0 : match client.re_attach(conf).await {
348 0 : Ok(tenants) => tenants
349 0 : .into_iter()
350 0 : .flat_map(|(id, rart)| {
351 0 : TenantStartupMode::from_reattach_tenant(rart).map(|tsm| (id, tsm))
352 0 : })
353 0 : .collect(),
354 : Err(RetryForeverError::ShuttingDown) => {
355 0 : anyhow::bail!("Shut down while waiting for control plane re-attach response")
356 : }
357 : }
358 : } else {
359 0 : info!("Control plane API not configured, tenant generations are disabled");
360 0 : return Ok(None);
361 : };
362 :
363 : // The deletion queue needs to know about the startup attachment state to decide which (if any) stored
364 : // deletion list entries may still be valid. We provide that by pushing a recovery operation into
365 : // the queue. Sequential processing of te queue ensures that recovery is done before any new tenant deletions
366 : // are processed, even though we don't block on recovery completing here.
367 0 : let attached_tenants = generations
368 0 : .iter()
369 0 : .flat_map(|(id, start_mode)| {
370 0 : match start_mode {
371 0 : TenantStartupMode::Attached((_mode, generation)) => Some(generation),
372 0 : TenantStartupMode::Secondary => None,
373 : }
374 0 : .map(|gen| (*id, *gen))
375 0 : })
376 0 : .collect();
377 0 : resources.deletion_queue_client.recover(attached_tenants)?;
378 :
379 0 : Ok(Some(generations))
380 0 : }
381 :
382 : /// Given a directory discovered in the pageserver's tenants/ directory, attempt
383 : /// to load a tenant config from it.
384 : ///
385 : /// If we cleaned up something expected (like an empty dir or a temp dir), return None.
386 0 : fn load_tenant_config(
387 0 : conf: &'static PageServerConf,
388 0 : tenant_shard_id: TenantShardId,
389 0 : dentry: Utf8DirEntry,
390 0 : ) -> Option<Result<LocationConf, LoadConfigError>> {
391 0 : let tenant_dir_path = dentry.path().to_path_buf();
392 0 : if crate::is_temporary(&tenant_dir_path) {
393 0 : info!("Found temporary tenant directory, removing: {tenant_dir_path}");
394 : // No need to use safe_remove_tenant_dir_all because this is already
395 : // a temporary path
396 0 : std::fs::remove_dir_all(&tenant_dir_path).fatal_err("delete temporary tenant dir");
397 0 : return None;
398 0 : }
399 0 :
400 0 : // This case happens if we crash during attachment before writing a config into the dir
401 0 : let is_empty = tenant_dir_path
402 0 : .is_empty_dir()
403 0 : .fatal_err("Checking for empty tenant dir");
404 0 : if is_empty {
405 0 : info!("removing empty tenant directory {tenant_dir_path:?}");
406 0 : std::fs::remove_dir(&tenant_dir_path).fatal_err("delete empty tenant dir");
407 0 : return None;
408 0 : }
409 0 :
410 0 : Some(Tenant::load_tenant_config(conf, &tenant_shard_id))
411 0 : }
412 :
413 : /// Initial stage of load: walk the local tenants directory, clean up any temp files,
414 : /// and load configurations for the tenants we found.
415 : ///
416 : /// Do this in parallel, because we expect 10k+ tenants, so serial execution can take
417 : /// seconds even on reasonably fast drives.
418 0 : async fn init_load_tenant_configs(
419 0 : conf: &'static PageServerConf,
420 0 : ) -> HashMap<TenantShardId, Result<LocationConf, LoadConfigError>> {
421 0 : let tenants_dir = conf.tenants_path();
422 :
423 0 : let dentries = tokio::task::spawn_blocking(move || -> Vec<Utf8DirEntry> {
424 0 : let context = format!("read tenants dir {tenants_dir}");
425 0 : let dir_entries = tenants_dir.read_dir_utf8().fatal_err(&context);
426 0 :
427 0 : dir_entries
428 0 : .collect::<Result<Vec<_>, std::io::Error>>()
429 0 : .fatal_err(&context)
430 0 : })
431 0 : .await
432 0 : .expect("Config load task panicked");
433 0 :
434 0 : let mut configs = HashMap::new();
435 0 :
436 0 : let mut join_set = JoinSet::new();
437 0 : for dentry in dentries {
438 0 : let tenant_shard_id = match dentry.file_name().parse::<TenantShardId>() {
439 0 : Ok(id) => id,
440 : Err(_) => {
441 0 : warn!(
442 0 : "Invalid tenant path (garbage in our repo directory?): '{}'",
443 0 : dentry.file_name()
444 : );
445 0 : continue;
446 : }
447 : };
448 :
449 0 : join_set.spawn_blocking(move || {
450 0 : (
451 0 : tenant_shard_id,
452 0 : load_tenant_config(conf, tenant_shard_id, dentry),
453 0 : )
454 0 : });
455 0 : }
456 :
457 0 : while let Some(r) = join_set.join_next().await {
458 0 : let (tenant_shard_id, tenant_config) = r.expect("Panic in config load task");
459 0 : if let Some(tenant_config) = tenant_config {
460 0 : configs.insert(tenant_shard_id, tenant_config);
461 0 : }
462 : }
463 :
464 0 : configs
465 0 : }
466 :
467 0 : #[derive(Debug, thiserror::Error)]
468 : pub(crate) enum DeleteTenantError {
469 : #[error("Tenant map slot error {0}")]
470 : SlotError(#[from] TenantSlotError),
471 :
472 : #[error("Cancelled")]
473 : Cancelled,
474 :
475 : #[error(transparent)]
476 : Other(#[from] anyhow::Error),
477 : }
478 :
479 : /// Initialize repositories with locally available timelines.
480 : /// Timelines that are only partially available locally (remote storage has more data than this pageserver)
481 : /// are scheduled for download and added to the tenant once download is completed.
482 0 : #[instrument(skip_all)]
483 : pub async fn init_tenant_mgr(
484 : conf: &'static PageServerConf,
485 : background_purges: BackgroundPurges,
486 : resources: TenantSharedResources,
487 : init_order: InitializationOrder,
488 : cancel: CancellationToken,
489 : ) -> anyhow::Result<TenantManager> {
490 : let mut tenants = BTreeMap::new();
491 :
492 : let ctx = RequestContext::todo_child(TaskKind::Startup, DownloadBehavior::Warn);
493 :
494 : // Initialize dynamic limits that depend on system resources
495 : let system_memory =
496 : sysinfo::System::new_with_specifics(sysinfo::RefreshKind::new().with_memory())
497 : .total_memory();
498 : let max_ephemeral_layer_bytes =
499 : conf.ephemeral_bytes_per_memory_kb as u64 * (system_memory / 1024);
500 : tracing::info!("Initialized ephemeral layer size limit to {max_ephemeral_layer_bytes}, for {system_memory} bytes of memory");
501 : inmemory_layer::GLOBAL_RESOURCES.max_dirty_bytes.store(
502 : max_ephemeral_layer_bytes,
503 : std::sync::atomic::Ordering::Relaxed,
504 : );
505 :
506 : // Scan local filesystem for attached tenants
507 : let tenant_configs = init_load_tenant_configs(conf).await;
508 :
509 : // Determine which tenants are to be secondary or attached, and in which generation
510 : let tenant_modes = init_load_generations(conf, &tenant_configs, &resources, &cancel).await?;
511 :
512 : tracing::info!(
513 : "Attaching {} tenants at startup, warming up {} at a time",
514 : tenant_configs.len(),
515 : conf.concurrent_tenant_warmup.initial_permits()
516 : );
517 : TENANT.startup_scheduled.inc_by(tenant_configs.len() as u64);
518 :
519 : // Accumulate futures for writing tenant configs, so that we can execute in parallel
520 : let mut config_write_futs = Vec::new();
521 :
522 : // Update the location configs according to the re-attach response and persist them to disk
523 : tracing::info!("Updating {} location configs", tenant_configs.len());
524 : for (tenant_shard_id, location_conf) in tenant_configs {
525 : let tenant_dir_path = conf.tenant_path(&tenant_shard_id);
526 :
527 : let mut location_conf = match location_conf {
528 : Ok(l) => l,
529 : Err(e) => {
530 : // This should only happen in the case of a serialization bug or critical local I/O error: we cannot load this tenant
531 : error!(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), "Failed to load tenant config, failed to {e:#}");
532 : continue;
533 : }
534 : };
535 :
536 : // FIXME: if we were attached, and get demoted to secondary on re-attach, we
537 : // don't have a place to get a config.
538 : // (https://github.com/neondatabase/neon/issues/5377)
539 : const DEFAULT_SECONDARY_CONF: SecondaryLocationConfig =
540 : SecondaryLocationConfig { warm: true };
541 :
542 : if let Some(tenant_modes) = &tenant_modes {
543 : // We have a generation map: treat it as the authority for whether
544 : // this tenant is really attached.
545 : match tenant_modes.get(&tenant_shard_id) {
546 : None => {
547 : info!(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), "Detaching tenant, control plane omitted it in re-attach response");
548 :
549 : match safe_rename_tenant_dir(&tenant_dir_path).await {
550 : Ok(tmp_path) => {
551 : background_purges.spawn(tmp_path);
552 : }
553 : Err(e) => {
554 : error!(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(),
555 : "Failed to move detached tenant directory '{tenant_dir_path}': {e:?}");
556 : }
557 : };
558 :
559 : // We deleted local content: move on to next tenant, don't try and spawn this one.
560 : continue;
561 : }
562 : Some(TenantStartupMode::Secondary) => {
563 : if !matches!(location_conf.mode, LocationMode::Secondary(_)) {
564 : location_conf.mode = LocationMode::Secondary(DEFAULT_SECONDARY_CONF);
565 : }
566 : }
567 : Some(TenantStartupMode::Attached((attach_mode, generation))) => {
568 : let old_gen_higher = match &location_conf.mode {
569 : LocationMode::Attached(AttachedLocationConfig {
570 : generation: old_generation,
571 : attach_mode: _attach_mode,
572 : }) => {
573 : if old_generation > generation {
574 : Some(old_generation)
575 : } else {
576 : None
577 : }
578 : }
579 : _ => None,
580 : };
581 : if let Some(old_generation) = old_gen_higher {
582 : tracing::error!(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(),
583 : "Control plane gave decreasing generation ({generation:?}) in re-attach response for tenant that was attached in generation {:?}, demoting to secondary",
584 : old_generation
585 : );
586 :
587 : // We cannot safely attach this tenant given a bogus generation number, but let's avoid throwing away
588 : // local disk content: demote to secondary rather than detaching.
589 : location_conf.mode = LocationMode::Secondary(DEFAULT_SECONDARY_CONF);
590 : } else {
591 : location_conf.attach_in_generation(*attach_mode, *generation);
592 : }
593 : }
594 : }
595 : } else {
596 : // Legacy mode: no generation information, any tenant present
597 : // on local disk may activate
598 : info!(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), "Starting tenant in legacy mode, no generation",);
599 : };
600 :
601 : // Presence of a generation number implies attachment: attach the tenant
602 : // if it wasn't already, and apply the generation number.
603 0 : config_write_futs.push(async move {
604 0 : let r = Tenant::persist_tenant_config(conf, &tenant_shard_id, &location_conf).await;
605 0 : (tenant_shard_id, location_conf, r)
606 0 : });
607 : }
608 :
609 : // Execute config writes with concurrency, to avoid bottlenecking on local FS write latency
610 : tracing::info!(
611 : "Writing {} location config files...",
612 : config_write_futs.len()
613 : );
614 : let config_write_results = futures::stream::iter(config_write_futs)
615 : .buffer_unordered(16)
616 : .collect::<Vec<_>>()
617 : .await;
618 :
619 : tracing::info!(
620 : "Spawning {} tenant shard locations...",
621 : config_write_results.len()
622 : );
623 : // For those shards that have live configurations, construct `Tenant` or `SecondaryTenant` objects and start them running
624 : for (tenant_shard_id, location_conf, config_write_result) in config_write_results {
625 : // Writing a config to local disk is foundational to startup up tenants: panic if we can't.
626 : config_write_result.fatal_err("write tenant shard config file");
627 :
628 : let tenant_dir_path = conf.tenant_path(&tenant_shard_id);
629 : let shard_identity = location_conf.shard;
630 : let slot = match location_conf.mode {
631 : LocationMode::Attached(attached_conf) => TenantSlot::Attached(
632 : tenant_spawn(
633 : conf,
634 : tenant_shard_id,
635 : &tenant_dir_path,
636 : resources.clone(),
637 : AttachedTenantConf::new(location_conf.tenant_conf, attached_conf),
638 : shard_identity,
639 : Some(init_order.clone()),
640 : SpawnMode::Lazy,
641 : &ctx,
642 : )
643 : .expect("global shutdown during init_tenant_mgr cannot happen"),
644 : ),
645 : LocationMode::Secondary(secondary_conf) => {
646 : info!(
647 : tenant_id = %tenant_shard_id.tenant_id,
648 : shard_id = %tenant_shard_id.shard_slug(),
649 : "Starting secondary tenant"
650 : );
651 : TenantSlot::Secondary(SecondaryTenant::new(
652 : tenant_shard_id,
653 : shard_identity,
654 : location_conf.tenant_conf,
655 : &secondary_conf,
656 : ))
657 : }
658 : };
659 :
660 : METRICS.slot_inserted(&slot);
661 : tenants.insert(tenant_shard_id, slot);
662 : }
663 :
664 : info!("Processed {} local tenants at startup", tenants.len());
665 :
666 : let mut tenants_map = TENANTS.write().unwrap();
667 : assert!(matches!(&*tenants_map, &TenantsMap::Initializing));
668 :
669 : *tenants_map = TenantsMap::Open(tenants);
670 :
671 : Ok(TenantManager {
672 : conf,
673 : tenants: &TENANTS,
674 : resources,
675 : cancel: CancellationToken::new(),
676 : background_purges,
677 : })
678 : }
679 :
680 : /// Wrapper for Tenant::spawn that checks invariants before running
681 : #[allow(clippy::too_many_arguments)]
682 0 : fn tenant_spawn(
683 0 : conf: &'static PageServerConf,
684 0 : tenant_shard_id: TenantShardId,
685 0 : tenant_path: &Utf8Path,
686 0 : resources: TenantSharedResources,
687 0 : location_conf: AttachedTenantConf,
688 0 : shard_identity: ShardIdentity,
689 0 : init_order: Option<InitializationOrder>,
690 0 : mode: SpawnMode,
691 0 : ctx: &RequestContext,
692 0 : ) -> Result<Arc<Tenant>, GlobalShutDown> {
693 0 : // All these conditions should have been satisfied by our caller: the tenant dir exists, is a well formed
694 0 : // path, and contains a configuration file. Assertions that do synchronous I/O are limited to debug mode
695 0 : // to avoid impacting prod runtime performance.
696 0 : assert!(!crate::is_temporary(tenant_path));
697 0 : debug_assert!(tenant_path.is_dir());
698 0 : debug_assert!(conf
699 0 : .tenant_location_config_path(&tenant_shard_id)
700 0 : .try_exists()
701 0 : .unwrap());
702 :
703 0 : Tenant::spawn(
704 0 : conf,
705 0 : tenant_shard_id,
706 0 : resources,
707 0 : location_conf,
708 0 : shard_identity,
709 0 : init_order,
710 0 : mode,
711 0 : ctx,
712 0 : )
713 0 : }
714 :
715 6 : async fn shutdown_all_tenants0(tenants: &std::sync::RwLock<TenantsMap>) {
716 6 : let mut join_set = JoinSet::new();
717 0 :
718 0 : #[cfg(all(debug_assertions, not(test)))]
719 0 : {
720 0 : // Check that our metrics properly tracked the size of the tenants map. This is a convenient location to check,
721 0 : // as it happens implicitly at the end of tests etc.
722 0 : let m = tenants.read().unwrap();
723 0 : debug_assert_eq!(METRICS.slots_total(), m.len() as u64);
724 : }
725 :
726 : // Atomically, 1. create the shutdown tasks and 2. prevent creation of new tenants.
727 6 : let (total_in_progress, total_attached) = {
728 6 : let mut m = tenants.write().unwrap();
729 6 : match &mut *m {
730 : TenantsMap::Initializing => {
731 0 : *m = TenantsMap::ShuttingDown(BTreeMap::default());
732 0 : info!("tenants map is empty");
733 0 : return;
734 : }
735 6 : TenantsMap::Open(tenants) => {
736 6 : let mut shutdown_state = BTreeMap::new();
737 6 : let mut total_in_progress = 0;
738 6 : let mut total_attached = 0;
739 :
740 6 : for (tenant_shard_id, v) in std::mem::take(tenants).into_iter() {
741 6 : match v {
742 0 : TenantSlot::Attached(t) => {
743 0 : shutdown_state.insert(tenant_shard_id, TenantSlot::Attached(t.clone()));
744 0 : join_set.spawn(
745 0 : async move {
746 0 : let res = {
747 0 : let (_guard, shutdown_progress) = completion::channel();
748 0 : t.shutdown(shutdown_progress, ShutdownMode::FreezeAndFlush).await
749 : };
750 :
751 0 : if let Err(other_progress) = res {
752 : // join the another shutdown in progress
753 0 : other_progress.wait().await;
754 0 : }
755 :
756 : // we cannot afford per tenant logging here, because if s3 is degraded, we are
757 : // going to log too many lines
758 0 : debug!("tenant successfully stopped");
759 0 : }
760 0 : .instrument(info_span!("shutdown", tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug())),
761 : );
762 :
763 0 : total_attached += 1;
764 : }
765 0 : TenantSlot::Secondary(state) => {
766 0 : // We don't need to wait for this individually per-tenant: the
767 0 : // downloader task will be waited on eventually, this cancel
768 0 : // is just to encourage it to drop out if it is doing work
769 0 : // for this tenant right now.
770 0 : state.cancel.cancel();
771 0 :
772 0 : shutdown_state.insert(tenant_shard_id, TenantSlot::Secondary(state));
773 0 : }
774 6 : TenantSlot::InProgress(notify) => {
775 6 : // InProgress tenants are not visible in TenantsMap::ShuttingDown: we will
776 6 : // wait for their notifications to fire in this function.
777 6 : join_set.spawn(async move {
778 6 : notify.wait().await;
779 6 : });
780 6 :
781 6 : total_in_progress += 1;
782 6 : }
783 : }
784 : }
785 6 : *m = TenantsMap::ShuttingDown(shutdown_state);
786 6 : (total_in_progress, total_attached)
787 : }
788 : TenantsMap::ShuttingDown(_) => {
789 0 : error!("already shutting down, this function isn't supposed to be called more than once");
790 0 : return;
791 : }
792 : }
793 : };
794 :
795 6 : let started_at = std::time::Instant::now();
796 6 :
797 6 : info!(
798 0 : "Waiting for {} InProgress tenants and {} Attached tenants to shut down",
799 : total_in_progress, total_attached
800 : );
801 :
802 6 : let total = join_set.len();
803 6 : let mut panicked = 0;
804 6 : let mut buffering = true;
805 : const BUFFER_FOR: std::time::Duration = std::time::Duration::from_millis(500);
806 6 : let mut buffered = std::pin::pin!(tokio::time::sleep(BUFFER_FOR));
807 :
808 18 : while !join_set.is_empty() {
809 12 : tokio::select! {
810 12 : Some(joined) = join_set.join_next() => {
811 0 : match joined {
812 6 : Ok(()) => {},
813 0 : Err(join_error) if join_error.is_cancelled() => {
814 0 : unreachable!("we are not cancelling any of the tasks");
815 : }
816 0 : Err(join_error) if join_error.is_panic() => {
817 0 : // cannot really do anything, as this panic is likely a bug
818 0 : panicked += 1;
819 0 : }
820 0 : Err(join_error) => {
821 0 : warn!("unknown kind of JoinError: {join_error}");
822 : }
823 : }
824 6 : if !buffering {
825 6 : // buffer so that every 500ms since the first update (or starting) we'll log
826 6 : // how far away we are; this is because we will get SIGKILL'd at 10s, and we
827 6 : // are not able to log *then*.
828 6 : buffering = true;
829 6 : buffered.as_mut().reset(tokio::time::Instant::now() + BUFFER_FOR);
830 6 : }
831 : },
832 12 : _ = &mut buffered, if buffering => {
833 6 : buffering = false;
834 6 : info!(remaining = join_set.len(), total, elapsed_ms = started_at.elapsed().as_millis(), "waiting for tenants to shutdown");
835 : }
836 : }
837 : }
838 :
839 6 : if panicked > 0 {
840 0 : warn!(
841 : panicked,
842 0 : total, "observed panicks while shutting down tenants"
843 : );
844 6 : }
845 :
846 : // caller will log how long we took
847 6 : }
848 :
849 0 : #[derive(thiserror::Error, Debug)]
850 : pub(crate) enum UpsertLocationError {
851 : #[error("Bad config request: {0}")]
852 : BadRequest(anyhow::Error),
853 :
854 : #[error("Cannot change config in this state: {0}")]
855 : Unavailable(#[from] TenantMapError),
856 :
857 : #[error("Tenant is already being modified")]
858 : InProgress,
859 :
860 : #[error("Failed to flush: {0}")]
861 : Flush(anyhow::Error),
862 :
863 : /// This error variant is for unexpected situations (soft assertions) where the system is in an unexpected state.
864 : #[error("Internal error: {0}")]
865 : InternalError(anyhow::Error),
866 : }
867 :
868 : impl TenantManager {
869 : /// Convenience function so that anyone with a TenantManager can get at the global configuration, without
870 : /// having to pass it around everywhere as a separate object.
871 0 : pub(crate) fn get_conf(&self) -> &'static PageServerConf {
872 0 : self.conf
873 0 : }
874 :
875 : /// Gets the attached tenant from the in-memory data, erroring if it's absent, in secondary mode, or currently
876 : /// undergoing a state change (i.e. slot is InProgress).
877 : ///
878 : /// The return Tenant is not guaranteed to be active: check its status after obtaing it, or
879 : /// use [`Tenant::wait_to_become_active`] before using it if you will do I/O on it.
880 0 : pub(crate) fn get_attached_tenant_shard(
881 0 : &self,
882 0 : tenant_shard_id: TenantShardId,
883 0 : ) -> Result<Arc<Tenant>, GetTenantError> {
884 0 : let locked = self.tenants.read().unwrap();
885 :
886 0 : let peek_slot = tenant_map_peek_slot(&locked, &tenant_shard_id, TenantSlotPeekMode::Read)?;
887 :
888 0 : match peek_slot {
889 0 : Some(TenantSlot::Attached(tenant)) => Ok(Arc::clone(tenant)),
890 0 : Some(TenantSlot::InProgress(_)) => Err(GetTenantError::NotActive(tenant_shard_id)),
891 : None | Some(TenantSlot::Secondary(_)) => {
892 0 : Err(GetTenantError::NotFound(tenant_shard_id.tenant_id))
893 : }
894 : }
895 0 : }
896 :
897 0 : pub(crate) fn get_secondary_tenant_shard(
898 0 : &self,
899 0 : tenant_shard_id: TenantShardId,
900 0 : ) -> Option<Arc<SecondaryTenant>> {
901 0 : let locked = self.tenants.read().unwrap();
902 0 :
903 0 : let peek_slot = tenant_map_peek_slot(&locked, &tenant_shard_id, TenantSlotPeekMode::Read)
904 0 : .ok()
905 0 : .flatten();
906 :
907 0 : match peek_slot {
908 0 : Some(TenantSlot::Secondary(s)) => Some(s.clone()),
909 0 : _ => None,
910 : }
911 0 : }
912 :
913 : /// Whether the `TenantManager` is responsible for the tenant shard
914 0 : pub(crate) fn manages_tenant_shard(&self, tenant_shard_id: TenantShardId) -> bool {
915 0 : let locked = self.tenants.read().unwrap();
916 0 :
917 0 : let peek_slot = tenant_map_peek_slot(&locked, &tenant_shard_id, TenantSlotPeekMode::Read)
918 0 : .ok()
919 0 : .flatten();
920 0 :
921 0 : peek_slot.is_some()
922 0 : }
923 :
924 0 : #[instrument(skip_all, fields(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug()))]
925 : pub(crate) async fn upsert_location(
926 : &self,
927 : tenant_shard_id: TenantShardId,
928 : new_location_config: LocationConf,
929 : flush: Option<Duration>,
930 : mut spawn_mode: SpawnMode,
931 : ctx: &RequestContext,
932 : ) -> Result<Option<Arc<Tenant>>, UpsertLocationError> {
933 : debug_assert_current_span_has_tenant_id();
934 : info!("configuring tenant location to state {new_location_config:?}");
935 :
936 : enum FastPathModified {
937 : Attached(Arc<Tenant>),
938 : Secondary(Arc<SecondaryTenant>),
939 : }
940 :
941 : // Special case fast-path for updates to existing slots: if our upsert is only updating configuration,
942 : // then we do not need to set the slot to InProgress, we can just call into the
943 : // existng tenant.
944 : let fast_path_taken = {
945 : let locked = self.tenants.read().unwrap();
946 : let peek_slot =
947 : tenant_map_peek_slot(&locked, &tenant_shard_id, TenantSlotPeekMode::Write)?;
948 : match (&new_location_config.mode, peek_slot) {
949 : (LocationMode::Attached(attach_conf), Some(TenantSlot::Attached(tenant))) => {
950 : match attach_conf.generation.cmp(&tenant.generation) {
951 : Ordering::Equal => {
952 : if attach_conf.attach_mode == AttachmentMode::Single {
953 : tenant
954 : .gc_block
955 : .set_lsn_lease_deadline(tenant.get_lsn_lease_length());
956 : }
957 :
958 : // A transition from Attached to Attached in the same generation, we may
959 : // take our fast path and just provide the updated configuration
960 : // to the tenant.
961 : tenant.set_new_location_config(
962 : AttachedTenantConf::try_from(new_location_config.clone())
963 : .map_err(UpsertLocationError::BadRequest)?,
964 : );
965 :
966 : Some(FastPathModified::Attached(tenant.clone()))
967 : }
968 : Ordering::Less => {
969 : return Err(UpsertLocationError::BadRequest(anyhow::anyhow!(
970 : "Generation {:?} is less than existing {:?}",
971 : attach_conf.generation,
972 : tenant.generation
973 : )));
974 : }
975 : Ordering::Greater => {
976 : // Generation advanced, fall through to general case of replacing `Tenant` object
977 : None
978 : }
979 : }
980 : }
981 : (
982 : LocationMode::Secondary(secondary_conf),
983 : Some(TenantSlot::Secondary(secondary_tenant)),
984 : ) => {
985 : secondary_tenant.set_config(secondary_conf);
986 : secondary_tenant.set_tenant_conf(&new_location_config.tenant_conf);
987 : Some(FastPathModified::Secondary(secondary_tenant.clone()))
988 : }
989 : _ => {
990 : // Not an Attached->Attached transition, fall through to general case
991 : None
992 : }
993 : }
994 : };
995 :
996 : // Fast-path continued: having dropped out of the self.tenants lock, do the async
997 : // phase of writing config and/or waiting for flush, before returning.
998 : match fast_path_taken {
999 : Some(FastPathModified::Attached(tenant)) => {
1000 : Tenant::persist_tenant_config(self.conf, &tenant_shard_id, &new_location_config)
1001 : .await
1002 : .fatal_err("write tenant shard config");
1003 :
1004 : // Transition to AttachedStale means we may well hold a valid generation
1005 : // still, and have been requested to go stale as part of a migration. If
1006 : // the caller set `flush`, then flush to remote storage.
1007 : if let LocationMode::Attached(AttachedLocationConfig {
1008 : generation: _,
1009 : attach_mode: AttachmentMode::Stale,
1010 : }) = &new_location_config.mode
1011 : {
1012 : if let Some(flush_timeout) = flush {
1013 : match tokio::time::timeout(flush_timeout, tenant.flush_remote()).await {
1014 : Ok(Err(e)) => {
1015 : return Err(UpsertLocationError::Flush(e));
1016 : }
1017 : Ok(Ok(_)) => return Ok(Some(tenant)),
1018 : Err(_) => {
1019 : tracing::warn!(
1020 : timeout_ms = flush_timeout.as_millis(),
1021 : "Timed out waiting for flush to remote storage, proceeding anyway."
1022 : )
1023 : }
1024 : }
1025 : }
1026 : }
1027 :
1028 : return Ok(Some(tenant));
1029 : }
1030 : Some(FastPathModified::Secondary(_secondary_tenant)) => {
1031 : Tenant::persist_tenant_config(self.conf, &tenant_shard_id, &new_location_config)
1032 : .await
1033 : .fatal_err("write tenant shard config");
1034 :
1035 : return Ok(None);
1036 : }
1037 : None => {
1038 : // Proceed with the general case procedure, where we will shutdown & remove any existing
1039 : // slot contents and replace with a fresh one
1040 : }
1041 : };
1042 :
1043 : // General case for upserts to TenantsMap, excluding the case above: we will substitute an
1044 : // InProgress value to the slot while we make whatever changes are required. The state for
1045 : // the tenant is inaccessible to the outside world while we are doing this, but that is sensible:
1046 : // the state is ill-defined while we're in transition. Transitions are async, but fast: we do
1047 : // not do significant I/O, and shutdowns should be prompt via cancellation tokens.
1048 : let mut slot_guard = tenant_map_acquire_slot(&tenant_shard_id, TenantSlotAcquireMode::Any)
1049 0 : .map_err(|e| match e {
1050 : TenantSlotError::NotFound(_) => {
1051 0 : unreachable!("Called with mode Any")
1052 : }
1053 0 : TenantSlotError::InProgress => UpsertLocationError::InProgress,
1054 0 : TenantSlotError::MapState(s) => UpsertLocationError::Unavailable(s),
1055 0 : })?;
1056 :
1057 : match slot_guard.get_old_value() {
1058 : Some(TenantSlot::Attached(tenant)) => {
1059 : // The case where we keep a Tenant alive was covered above in the special case
1060 : // for Attached->Attached transitions in the same generation. By this point,
1061 : // if we see an attached tenant we know it will be discarded and should be
1062 : // shut down.
1063 : let (_guard, progress) = utils::completion::channel();
1064 :
1065 : match tenant.get_attach_mode() {
1066 : AttachmentMode::Single | AttachmentMode::Multi => {
1067 : // Before we leave our state as the presumed holder of the latest generation,
1068 : // flush any outstanding deletions to reduce the risk of leaking objects.
1069 : self.resources.deletion_queue_client.flush_advisory()
1070 : }
1071 : AttachmentMode::Stale => {
1072 : // If we're stale there's not point trying to flush deletions
1073 : }
1074 : };
1075 :
1076 : info!("Shutting down attached tenant");
1077 : match tenant.shutdown(progress, ShutdownMode::Hard).await {
1078 : Ok(()) => {}
1079 : Err(barrier) => {
1080 : info!("Shutdown already in progress, waiting for it to complete");
1081 : barrier.wait().await;
1082 : }
1083 : }
1084 : slot_guard.drop_old_value().expect("We just shut it down");
1085 :
1086 : // Edge case: if we were called with SpawnMode::Create, but a Tenant already existed, then
1087 : // the caller thinks they're creating but the tenant already existed. We must switch to
1088 : // Eager mode so that when starting this Tenant we properly probe remote storage for timelines,
1089 : // rather than assuming it to be empty.
1090 : spawn_mode = SpawnMode::Eager;
1091 : }
1092 : Some(TenantSlot::Secondary(state)) => {
1093 : info!("Shutting down secondary tenant");
1094 : state.shutdown().await;
1095 : }
1096 : Some(TenantSlot::InProgress(_)) => {
1097 : // This should never happen: acquire_slot should error out
1098 : // if the contents of a slot were InProgress.
1099 : return Err(UpsertLocationError::InternalError(anyhow::anyhow!(
1100 : "Acquired an InProgress slot, this is a bug."
1101 : )));
1102 : }
1103 : None => {
1104 : // Slot was vacant, nothing needs shutting down.
1105 : }
1106 : }
1107 :
1108 : let tenant_path = self.conf.tenant_path(&tenant_shard_id);
1109 : let timelines_path = self.conf.timelines_path(&tenant_shard_id);
1110 :
1111 : // Directory structure is the same for attached and secondary modes:
1112 : // create it if it doesn't exist. Timeline load/creation expects the
1113 : // timelines/ subdir to already exist.
1114 : //
1115 : // Does not need to be fsync'd because local storage is just a cache.
1116 : tokio::fs::create_dir_all(&timelines_path)
1117 : .await
1118 : .fatal_err("create timelines/ dir");
1119 :
1120 : // Before activating either secondary or attached mode, persist the
1121 : // configuration, so that on restart we will re-attach (or re-start
1122 : // secondary) on the tenant.
1123 : Tenant::persist_tenant_config(self.conf, &tenant_shard_id, &new_location_config)
1124 : .await
1125 : .fatal_err("write tenant shard config");
1126 :
1127 : let new_slot = match &new_location_config.mode {
1128 : LocationMode::Secondary(secondary_config) => {
1129 : let shard_identity = new_location_config.shard;
1130 : TenantSlot::Secondary(SecondaryTenant::new(
1131 : tenant_shard_id,
1132 : shard_identity,
1133 : new_location_config.tenant_conf,
1134 : secondary_config,
1135 : ))
1136 : }
1137 : LocationMode::Attached(_attach_config) => {
1138 : let shard_identity = new_location_config.shard;
1139 :
1140 : // Testing hack: if we are configured with no control plane, then drop the generation
1141 : // from upserts. This enables creating generation-less tenants even though neon_local
1142 : // always uses generations when calling the location conf API.
1143 : let attached_conf = if cfg!(feature = "testing") {
1144 : let mut conf = AttachedTenantConf::try_from(new_location_config)
1145 : .map_err(UpsertLocationError::BadRequest)?;
1146 : if self.conf.control_plane_api.is_none() {
1147 : conf.location.generation = Generation::none();
1148 : }
1149 : conf
1150 : } else {
1151 : AttachedTenantConf::try_from(new_location_config)
1152 : .map_err(UpsertLocationError::BadRequest)?
1153 : };
1154 :
1155 : let tenant = tenant_spawn(
1156 : self.conf,
1157 : tenant_shard_id,
1158 : &tenant_path,
1159 : self.resources.clone(),
1160 : attached_conf,
1161 : shard_identity,
1162 : None,
1163 : spawn_mode,
1164 : ctx,
1165 : )
1166 0 : .map_err(|_: GlobalShutDown| {
1167 0 : UpsertLocationError::Unavailable(TenantMapError::ShuttingDown)
1168 0 : })?;
1169 :
1170 : TenantSlot::Attached(tenant)
1171 : }
1172 : };
1173 :
1174 : let attached_tenant = if let TenantSlot::Attached(tenant) = &new_slot {
1175 : Some(tenant.clone())
1176 : } else {
1177 : None
1178 : };
1179 :
1180 : match slot_guard.upsert(new_slot) {
1181 : Err(TenantSlotUpsertError::InternalError(e)) => {
1182 : Err(UpsertLocationError::InternalError(anyhow::anyhow!(e)))
1183 : }
1184 : Err(TenantSlotUpsertError::MapState(e)) => Err(UpsertLocationError::Unavailable(e)),
1185 : Err(TenantSlotUpsertError::ShuttingDown((new_slot, _completion))) => {
1186 : // If we just called tenant_spawn() on a new tenant, and can't insert it into our map, then
1187 : // we must not leak it: this would violate the invariant that after shutdown_all_tenants, all tenants
1188 : // are shutdown.
1189 : //
1190 : // We must shut it down inline here.
1191 : match new_slot {
1192 : TenantSlot::InProgress(_) => {
1193 : // Unreachable because we never insert an InProgress
1194 : unreachable!()
1195 : }
1196 : TenantSlot::Attached(tenant) => {
1197 : let (_guard, progress) = utils::completion::channel();
1198 : info!("Shutting down just-spawned tenant, because tenant manager is shut down");
1199 : match tenant.shutdown(progress, ShutdownMode::Hard).await {
1200 : Ok(()) => {
1201 : info!("Finished shutting down just-spawned tenant");
1202 : }
1203 : Err(barrier) => {
1204 : info!("Shutdown already in progress, waiting for it to complete");
1205 : barrier.wait().await;
1206 : }
1207 : }
1208 : }
1209 : TenantSlot::Secondary(secondary_tenant) => {
1210 : secondary_tenant.shutdown().await;
1211 : }
1212 : }
1213 :
1214 : Err(UpsertLocationError::Unavailable(
1215 : TenantMapError::ShuttingDown,
1216 : ))
1217 : }
1218 : Ok(()) => Ok(attached_tenant),
1219 : }
1220 : }
1221 :
1222 : /// Resetting a tenant is equivalent to detaching it, then attaching it again with the same
1223 : /// LocationConf that was last used to attach it. Optionally, the local file cache may be
1224 : /// dropped before re-attaching.
1225 : ///
1226 : /// This is not part of a tenant's normal lifecycle: it is used for debug/support, in situations
1227 : /// where an issue is identified that would go away with a restart of the tenant.
1228 : ///
1229 : /// This does not have any special "force" shutdown of a tenant: it relies on the tenant's tasks
1230 : /// to respect the cancellation tokens used in normal shutdown().
1231 0 : #[instrument(skip_all, fields(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %drop_cache))]
1232 : pub(crate) async fn reset_tenant(
1233 : &self,
1234 : tenant_shard_id: TenantShardId,
1235 : drop_cache: bool,
1236 : ctx: &RequestContext,
1237 : ) -> anyhow::Result<()> {
1238 : let mut slot_guard = tenant_map_acquire_slot(&tenant_shard_id, TenantSlotAcquireMode::Any)?;
1239 : let Some(old_slot) = slot_guard.get_old_value() else {
1240 : anyhow::bail!("Tenant not found when trying to reset");
1241 : };
1242 :
1243 : let Some(tenant) = old_slot.get_attached() else {
1244 : slot_guard.revert();
1245 : anyhow::bail!("Tenant is not in attached state");
1246 : };
1247 :
1248 : let (_guard, progress) = utils::completion::channel();
1249 : match tenant.shutdown(progress, ShutdownMode::Hard).await {
1250 : Ok(()) => {
1251 : slot_guard.drop_old_value()?;
1252 : }
1253 : Err(_barrier) => {
1254 : slot_guard.revert();
1255 : anyhow::bail!("Cannot reset Tenant, already shutting down");
1256 : }
1257 : }
1258 :
1259 : let tenant_path = self.conf.tenant_path(&tenant_shard_id);
1260 : let timelines_path = self.conf.timelines_path(&tenant_shard_id);
1261 : let config = Tenant::load_tenant_config(self.conf, &tenant_shard_id)?;
1262 :
1263 : if drop_cache {
1264 : tracing::info!("Dropping local file cache");
1265 :
1266 : match tokio::fs::read_dir(&timelines_path).await {
1267 : Err(e) => {
1268 : tracing::warn!("Failed to list timelines while dropping cache: {}", e);
1269 : }
1270 : Ok(mut entries) => {
1271 : while let Some(entry) = entries.next_entry().await? {
1272 : tokio::fs::remove_dir_all(entry.path()).await?;
1273 : }
1274 : }
1275 : }
1276 : }
1277 :
1278 : let shard_identity = config.shard;
1279 : let tenant = tenant_spawn(
1280 : self.conf,
1281 : tenant_shard_id,
1282 : &tenant_path,
1283 : self.resources.clone(),
1284 : AttachedTenantConf::try_from(config)?,
1285 : shard_identity,
1286 : None,
1287 : SpawnMode::Eager,
1288 : ctx,
1289 : )?;
1290 :
1291 : slot_guard.upsert(TenantSlot::Attached(tenant))?;
1292 :
1293 : Ok(())
1294 : }
1295 :
1296 0 : pub(crate) fn get_attached_active_tenant_shards(&self) -> Vec<Arc<Tenant>> {
1297 0 : let locked = self.tenants.read().unwrap();
1298 0 : match &*locked {
1299 0 : TenantsMap::Initializing => Vec::new(),
1300 0 : TenantsMap::Open(map) | TenantsMap::ShuttingDown(map) => map
1301 0 : .values()
1302 0 : .filter_map(|slot| {
1303 0 : slot.get_attached()
1304 0 : .and_then(|t| if t.is_active() { Some(t.clone()) } else { None })
1305 0 : })
1306 0 : .collect(),
1307 : }
1308 0 : }
1309 : // Do some synchronous work for all tenant slots in Secondary state. The provided
1310 : // callback should be small and fast, as it will be called inside the global
1311 : // TenantsMap lock.
1312 0 : pub(crate) fn foreach_secondary_tenants<F>(&self, mut func: F)
1313 0 : where
1314 0 : // TODO: let the callback return a hint to drop out of the loop early
1315 0 : F: FnMut(&TenantShardId, &Arc<SecondaryTenant>),
1316 0 : {
1317 0 : let locked = self.tenants.read().unwrap();
1318 :
1319 0 : let map = match &*locked {
1320 0 : TenantsMap::Initializing | TenantsMap::ShuttingDown(_) => return,
1321 0 : TenantsMap::Open(m) => m,
1322 : };
1323 :
1324 0 : for (tenant_id, slot) in map {
1325 0 : if let TenantSlot::Secondary(state) = slot {
1326 : // Only expose secondary tenants that are not currently shutting down
1327 0 : if !state.cancel.is_cancelled() {
1328 0 : func(tenant_id, state)
1329 0 : }
1330 0 : }
1331 : }
1332 0 : }
1333 :
1334 : /// Total list of all tenant slots: this includes attached, secondary, and InProgress.
1335 0 : pub(crate) fn list(&self) -> Vec<(TenantShardId, TenantSlot)> {
1336 0 : let locked = self.tenants.read().unwrap();
1337 0 : match &*locked {
1338 0 : TenantsMap::Initializing => Vec::new(),
1339 0 : TenantsMap::Open(map) | TenantsMap::ShuttingDown(map) => {
1340 0 : map.iter().map(|(k, v)| (*k, v.clone())).collect()
1341 : }
1342 : }
1343 0 : }
1344 :
1345 0 : pub(crate) fn get(&self, tenant_shard_id: TenantShardId) -> Option<TenantSlot> {
1346 0 : let locked = self.tenants.read().unwrap();
1347 0 : match &*locked {
1348 0 : TenantsMap::Initializing => None,
1349 0 : TenantsMap::Open(map) | TenantsMap::ShuttingDown(map) => {
1350 0 : map.get(&tenant_shard_id).cloned()
1351 : }
1352 : }
1353 0 : }
1354 :
1355 0 : async fn delete_tenant_remote(
1356 0 : &self,
1357 0 : tenant_shard_id: TenantShardId,
1358 0 : ) -> Result<(), DeleteTenantError> {
1359 0 : let remote_path = remote_tenant_path(&tenant_shard_id);
1360 0 : let mut keys_stream = self.resources.remote_storage.list_streaming(
1361 0 : Some(&remote_path),
1362 0 : remote_storage::ListingMode::NoDelimiter,
1363 0 : None,
1364 0 : &self.cancel,
1365 0 : );
1366 0 : while let Some(chunk) = keys_stream.next().await {
1367 0 : let keys = match chunk {
1368 0 : Ok(listing) => listing.keys,
1369 : Err(remote_storage::DownloadError::Cancelled) => {
1370 0 : return Err(DeleteTenantError::Cancelled)
1371 : }
1372 0 : Err(remote_storage::DownloadError::NotFound) => return Ok(()),
1373 0 : Err(other) => return Err(DeleteTenantError::Other(anyhow::anyhow!(other))),
1374 : };
1375 :
1376 0 : if keys.is_empty() {
1377 0 : tracing::info!("Remote storage already deleted");
1378 : } else {
1379 0 : tracing::info!("Deleting {} keys from remote storage", keys.len());
1380 0 : let keys = keys.into_iter().map(|o| o.key).collect::<Vec<_>>();
1381 0 : self.resources
1382 0 : .remote_storage
1383 0 : .delete_objects(&keys, &self.cancel)
1384 0 : .await?;
1385 : }
1386 : }
1387 :
1388 0 : Ok(())
1389 0 : }
1390 :
1391 : /// If a tenant is attached, detach it. Then remove its data from remote storage.
1392 : ///
1393 : /// A tenant is considered deleted once it is gone from remote storage. It is the caller's
1394 : /// responsibility to avoid trying to attach the tenant again or use it any way once deletion
1395 : /// has started: this operation is not atomic, and must be retried until it succeeds.
1396 0 : pub(crate) async fn delete_tenant(
1397 0 : &self,
1398 0 : tenant_shard_id: TenantShardId,
1399 0 : ) -> Result<(), DeleteTenantError> {
1400 0 : super::span::debug_assert_current_span_has_tenant_id();
1401 :
1402 0 : async fn delete_local(
1403 0 : conf: &PageServerConf,
1404 0 : background_purges: &BackgroundPurges,
1405 0 : tenant_shard_id: &TenantShardId,
1406 0 : ) -> anyhow::Result<()> {
1407 0 : let local_tenant_directory = conf.tenant_path(tenant_shard_id);
1408 0 : let tmp_dir = safe_rename_tenant_dir(&local_tenant_directory)
1409 0 : .await
1410 0 : .with_context(|| {
1411 0 : format!("local tenant directory {local_tenant_directory:?} rename")
1412 0 : })?;
1413 0 : background_purges.spawn(tmp_dir);
1414 0 : Ok(())
1415 0 : }
1416 :
1417 0 : let slot_guard = tenant_map_acquire_slot(&tenant_shard_id, TenantSlotAcquireMode::Any)?;
1418 0 : match &slot_guard.old_value {
1419 0 : Some(TenantSlot::Attached(tenant)) => {
1420 0 : // Legacy deletion flow: the tenant remains attached, goes to Stopping state, and
1421 0 : // deletion will be resumed across restarts.
1422 0 : let tenant = tenant.clone();
1423 0 : let (_guard, progress) = utils::completion::channel();
1424 0 : match tenant.shutdown(progress, ShutdownMode::Hard).await {
1425 0 : Ok(()) => {}
1426 0 : Err(barrier) => {
1427 0 : info!("Shutdown already in progress, waiting for it to complete");
1428 0 : barrier.wait().await;
1429 : }
1430 : }
1431 0 : delete_local(self.conf, &self.background_purges, &tenant_shard_id).await?;
1432 : }
1433 0 : Some(TenantSlot::Secondary(secondary_tenant)) => {
1434 0 : secondary_tenant.shutdown().await;
1435 :
1436 0 : delete_local(self.conf, &self.background_purges, &tenant_shard_id).await?;
1437 : }
1438 0 : Some(TenantSlot::InProgress(_)) => unreachable!(),
1439 0 : None => {}
1440 : };
1441 :
1442 : // Fall through: local state for this tenant is no longer present, proceed with remote delete.
1443 : // - We use a retry wrapper here so that common transient S3 errors (e.g. 503, 429) do not result
1444 : // in 500 responses to delete requests.
1445 : // - We keep the `SlotGuard` during this I/O, so that if a concurrent delete request comes in, it will
1446 : // 503/retry, rather than kicking off a wasteful concurrent deletion.
1447 0 : match backoff::retry(
1448 0 : || async move { self.delete_tenant_remote(tenant_shard_id).await },
1449 0 : |e| match e {
1450 0 : DeleteTenantError::Cancelled => true,
1451 : DeleteTenantError::SlotError(_) => {
1452 0 : unreachable!("Remote deletion doesn't touch slots")
1453 : }
1454 0 : _ => false,
1455 0 : },
1456 0 : 1,
1457 0 : 3,
1458 0 : &format!("delete_tenant[tenant_shard_id={tenant_shard_id}]"),
1459 0 : &self.cancel,
1460 0 : )
1461 0 : .await
1462 : {
1463 0 : Some(r) => r,
1464 0 : None => Err(DeleteTenantError::Cancelled),
1465 : }
1466 0 : }
1467 :
1468 0 : #[instrument(skip_all, fields(tenant_id=%tenant.get_tenant_shard_id().tenant_id, shard_id=%tenant.get_tenant_shard_id().shard_slug(), new_shard_count=%new_shard_count.literal()))]
1469 : pub(crate) async fn shard_split(
1470 : &self,
1471 : tenant: Arc<Tenant>,
1472 : new_shard_count: ShardCount,
1473 : new_stripe_size: Option<ShardStripeSize>,
1474 : ctx: &RequestContext,
1475 : ) -> anyhow::Result<Vec<TenantShardId>> {
1476 : let tenant_shard_id = *tenant.get_tenant_shard_id();
1477 : let r = self
1478 : .do_shard_split(tenant, new_shard_count, new_stripe_size, ctx)
1479 : .await;
1480 : if r.is_err() {
1481 : // Shard splitting might have left the original shard in a partially shut down state (it
1482 : // stops the shard's remote timeline client). Reset it to ensure we leave things in
1483 : // a working state.
1484 : if self.get(tenant_shard_id).is_some() {
1485 : tracing::warn!("Resetting after shard split failure");
1486 : if let Err(e) = self.reset_tenant(tenant_shard_id, false, ctx).await {
1487 : // Log this error because our return value will still be the original error, not this one. This is
1488 : // a severe error: if this happens, we might be leaving behind a tenant that is not fully functional
1489 : // (e.g. has uploads disabled). We can't do anything else: if reset fails then shutting the tenant down or
1490 : // setting it broken probably won't help either.
1491 : tracing::error!("Failed to reset: {e}");
1492 : }
1493 : }
1494 : }
1495 :
1496 : r
1497 : }
1498 :
1499 0 : pub(crate) async fn do_shard_split(
1500 0 : &self,
1501 0 : tenant: Arc<Tenant>,
1502 0 : new_shard_count: ShardCount,
1503 0 : new_stripe_size: Option<ShardStripeSize>,
1504 0 : ctx: &RequestContext,
1505 0 : ) -> anyhow::Result<Vec<TenantShardId>> {
1506 0 : let tenant_shard_id = *tenant.get_tenant_shard_id();
1507 0 :
1508 0 : // Validate the incoming request
1509 0 : if new_shard_count.count() <= tenant_shard_id.shard_count.count() {
1510 0 : anyhow::bail!("Requested shard count is not an increase");
1511 0 : }
1512 0 : let expansion_factor = new_shard_count.count() / tenant_shard_id.shard_count.count();
1513 0 : if !expansion_factor.is_power_of_two() {
1514 0 : anyhow::bail!("Requested split is not a power of two");
1515 0 : }
1516 :
1517 0 : if let Some(new_stripe_size) = new_stripe_size {
1518 0 : if tenant.get_shard_stripe_size() != new_stripe_size
1519 0 : && tenant_shard_id.shard_count.count() > 1
1520 : {
1521 : // This tenant already has multiple shards, it is illegal to try and change its stripe size
1522 0 : anyhow::bail!(
1523 0 : "Shard stripe size may not be modified once tenant has multiple shards"
1524 0 : );
1525 0 : }
1526 0 : }
1527 :
1528 : // Plan: identify what the new child shards will be
1529 0 : let child_shards = tenant_shard_id.split(new_shard_count);
1530 0 : tracing::info!(
1531 0 : "Shard {} splits into: {}",
1532 0 : tenant_shard_id.to_index(),
1533 0 : child_shards
1534 0 : .iter()
1535 0 : .map(|id| format!("{}", id.to_index()))
1536 0 : .join(",")
1537 : );
1538 :
1539 0 : fail::fail_point!("shard-split-pre-prepare", |_| Err(anyhow::anyhow!(
1540 0 : "failpoint"
1541 0 : )));
1542 :
1543 0 : let parent_shard_identity = tenant.shard_identity;
1544 0 : let parent_tenant_conf = tenant.get_tenant_conf();
1545 0 : let parent_generation = tenant.generation;
1546 :
1547 : // Phase 1: Write out child shards' remote index files, in the parent tenant's current generation
1548 0 : if let Err(e) = tenant.split_prepare(&child_shards).await {
1549 : // If [`Tenant::split_prepare`] fails, we must reload the tenant, because it might
1550 : // have been left in a partially-shut-down state.
1551 0 : tracing::warn!("Failed to prepare for split: {e}, reloading Tenant before returning");
1552 0 : return Err(e);
1553 0 : }
1554 0 :
1555 0 : fail::fail_point!("shard-split-post-prepare", |_| Err(anyhow::anyhow!(
1556 0 : "failpoint"
1557 0 : )));
1558 :
1559 0 : self.resources.deletion_queue_client.flush_advisory();
1560 0 :
1561 0 : // Phase 2: Put the parent shard to InProgress and grab a reference to the parent Tenant
1562 0 : drop(tenant);
1563 0 : let mut parent_slot_guard =
1564 0 : tenant_map_acquire_slot(&tenant_shard_id, TenantSlotAcquireMode::Any)?;
1565 0 : let parent = match parent_slot_guard.get_old_value() {
1566 0 : Some(TenantSlot::Attached(t)) => t,
1567 0 : Some(TenantSlot::Secondary(_)) => anyhow::bail!("Tenant location in secondary mode"),
1568 : Some(TenantSlot::InProgress(_)) => {
1569 : // tenant_map_acquire_slot never returns InProgress, if a slot was InProgress
1570 : // it would return an error.
1571 0 : unreachable!()
1572 : }
1573 : None => {
1574 : // We don't actually need the parent shard to still be attached to do our work, but it's
1575 : // a weird enough situation that the caller probably didn't want us to continue working
1576 : // if they had detached the tenant they requested the split on.
1577 0 : anyhow::bail!("Detached parent shard in the middle of split!")
1578 : }
1579 : };
1580 0 : fail::fail_point!("shard-split-pre-hardlink", |_| Err(anyhow::anyhow!(
1581 0 : "failpoint"
1582 0 : )));
1583 : // Optimization: hardlink layers from the parent into the children, so that they don't have to
1584 : // re-download & duplicate the data referenced in their initial IndexPart
1585 0 : self.shard_split_hardlink(parent, child_shards.clone())
1586 0 : .await?;
1587 0 : fail::fail_point!("shard-split-post-hardlink", |_| Err(anyhow::anyhow!(
1588 0 : "failpoint"
1589 0 : )));
1590 :
1591 : // Take a snapshot of where the parent's WAL ingest had got to: we will wait for
1592 : // child shards to reach this point.
1593 0 : let mut target_lsns = HashMap::new();
1594 0 : for timeline in parent.timelines.lock().unwrap().clone().values() {
1595 0 : target_lsns.insert(timeline.timeline_id, timeline.get_last_record_lsn());
1596 0 : }
1597 :
1598 : // TODO: we should have the parent shard stop its WAL ingest here, it's a waste of resources
1599 : // and could slow down the children trying to catch up.
1600 :
1601 : // Phase 3: Spawn the child shards
1602 0 : for child_shard in &child_shards {
1603 0 : let mut child_shard_identity = parent_shard_identity;
1604 0 : if let Some(new_stripe_size) = new_stripe_size {
1605 0 : child_shard_identity.stripe_size = new_stripe_size;
1606 0 : }
1607 0 : child_shard_identity.count = child_shard.shard_count;
1608 0 : child_shard_identity.number = child_shard.shard_number;
1609 0 :
1610 0 : let child_location_conf = LocationConf {
1611 0 : mode: LocationMode::Attached(AttachedLocationConfig {
1612 0 : generation: parent_generation,
1613 0 : attach_mode: AttachmentMode::Single,
1614 0 : }),
1615 0 : shard: child_shard_identity,
1616 0 : tenant_conf: parent_tenant_conf.clone(),
1617 0 : };
1618 0 :
1619 0 : self.upsert_location(
1620 0 : *child_shard,
1621 0 : child_location_conf,
1622 0 : None,
1623 0 : SpawnMode::Eager,
1624 0 : ctx,
1625 0 : )
1626 0 : .await?;
1627 : }
1628 :
1629 0 : fail::fail_point!("shard-split-post-child-conf", |_| Err(anyhow::anyhow!(
1630 0 : "failpoint"
1631 0 : )));
1632 :
1633 : // Phase 4: wait for child chards WAL ingest to catch up to target LSN
1634 0 : for child_shard_id in &child_shards {
1635 0 : let child_shard_id = *child_shard_id;
1636 0 : let child_shard = {
1637 0 : let locked = self.tenants.read().unwrap();
1638 0 : let peek_slot =
1639 0 : tenant_map_peek_slot(&locked, &child_shard_id, TenantSlotPeekMode::Read)?;
1640 0 : peek_slot.and_then(|s| s.get_attached()).cloned()
1641 : };
1642 0 : if let Some(t) = child_shard {
1643 : // Wait for the child shard to become active: this should be very quick because it only
1644 : // has to download the index_part that we just uploaded when creating it.
1645 0 : if let Err(e) = t.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await {
1646 : // This is not fatal: we have durably created the child shard. It just makes the
1647 : // split operation less seamless for clients, as we will may detach the parent
1648 : // shard before the child shards are fully ready to serve requests.
1649 0 : tracing::warn!("Failed to wait for shard {child_shard_id} to activate: {e}");
1650 0 : continue;
1651 0 : }
1652 0 :
1653 0 : let timelines = t.timelines.lock().unwrap().clone();
1654 0 : for timeline in timelines.values() {
1655 0 : let Some(target_lsn) = target_lsns.get(&timeline.timeline_id) else {
1656 0 : continue;
1657 : };
1658 :
1659 0 : tracing::info!(
1660 0 : "Waiting for child shard {}/{} to reach target lsn {}...",
1661 0 : child_shard_id,
1662 0 : timeline.timeline_id,
1663 : target_lsn
1664 : );
1665 :
1666 0 : fail::fail_point!("shard-split-lsn-wait", |_| Err(anyhow::anyhow!(
1667 0 : "failpoint"
1668 0 : )));
1669 0 : if let Err(e) = timeline
1670 0 : .wait_lsn(
1671 0 : *target_lsn,
1672 0 : crate::tenant::timeline::WaitLsnWaiter::Tenant,
1673 0 : ctx,
1674 0 : )
1675 0 : .await
1676 : {
1677 : // Failure here might mean shutdown, in any case this part is an optimization
1678 : // and we shouldn't hold up the split operation.
1679 0 : tracing::warn!(
1680 0 : "Failed to wait for timeline {} to reach lsn {target_lsn}: {e}",
1681 0 : timeline.timeline_id
1682 : );
1683 : } else {
1684 0 : tracing::info!(
1685 0 : "Child shard {}/{} reached target lsn {}",
1686 0 : child_shard_id,
1687 0 : timeline.timeline_id,
1688 : target_lsn
1689 : );
1690 : }
1691 : }
1692 0 : }
1693 : }
1694 :
1695 : // Phase 5: Shut down the parent shard, and erase it from disk
1696 0 : let (_guard, progress) = completion::channel();
1697 0 : match parent.shutdown(progress, ShutdownMode::Hard).await {
1698 0 : Ok(()) => {}
1699 0 : Err(other) => {
1700 0 : other.wait().await;
1701 : }
1702 : }
1703 0 : let local_tenant_directory = self.conf.tenant_path(&tenant_shard_id);
1704 0 : let tmp_path = safe_rename_tenant_dir(&local_tenant_directory)
1705 0 : .await
1706 0 : .with_context(|| format!("local tenant directory {local_tenant_directory:?} rename"))?;
1707 0 : self.background_purges.spawn(tmp_path);
1708 0 :
1709 0 : fail::fail_point!("shard-split-pre-finish", |_| Err(anyhow::anyhow!(
1710 0 : "failpoint"
1711 0 : )));
1712 :
1713 0 : parent_slot_guard.drop_old_value()?;
1714 :
1715 : // Phase 6: Release the InProgress on the parent shard
1716 0 : drop(parent_slot_guard);
1717 0 :
1718 0 : Ok(child_shards)
1719 0 : }
1720 :
1721 : /// Part of [`Self::shard_split`]: hard link parent shard layers into child shards, as an optimization
1722 : /// to avoid the children downloading them again.
1723 : ///
1724 : /// For each resident layer in the parent shard, we will hard link it into all of the child shards.
1725 0 : async fn shard_split_hardlink(
1726 0 : &self,
1727 0 : parent_shard: &Tenant,
1728 0 : child_shards: Vec<TenantShardId>,
1729 0 : ) -> anyhow::Result<()> {
1730 0 : debug_assert_current_span_has_tenant_id();
1731 0 :
1732 0 : let parent_path = self.conf.tenant_path(parent_shard.get_tenant_shard_id());
1733 0 : let (parent_timelines, parent_layers) = {
1734 0 : let mut parent_layers = Vec::new();
1735 0 : let timelines = parent_shard.timelines.lock().unwrap().clone();
1736 0 : let parent_timelines = timelines.keys().cloned().collect::<Vec<_>>();
1737 0 : for timeline in timelines.values() {
1738 0 : tracing::info!(timeline_id=%timeline.timeline_id, "Loading list of layers to hardlink");
1739 0 : let layers = timeline.layers.read().await;
1740 :
1741 0 : for layer in layers.likely_resident_layers() {
1742 0 : let relative_path = layer
1743 0 : .local_path()
1744 0 : .strip_prefix(&parent_path)
1745 0 : .context("Removing prefix from parent layer path")?;
1746 0 : parent_layers.push(relative_path.to_owned());
1747 : }
1748 : }
1749 0 : debug_assert!(
1750 0 : !parent_layers.is_empty(),
1751 0 : "shutdown cannot empty the layermap"
1752 : );
1753 0 : (parent_timelines, parent_layers)
1754 0 : };
1755 0 :
1756 0 : let mut child_prefixes = Vec::new();
1757 0 : let mut create_dirs = Vec::new();
1758 :
1759 0 : for child in child_shards {
1760 0 : let child_prefix = self.conf.tenant_path(&child);
1761 0 : create_dirs.push(child_prefix.clone());
1762 0 : create_dirs.extend(
1763 0 : parent_timelines
1764 0 : .iter()
1765 0 : .map(|t| self.conf.timeline_path(&child, t)),
1766 0 : );
1767 0 :
1768 0 : child_prefixes.push(child_prefix);
1769 0 : }
1770 :
1771 : // Since we will do a large number of small filesystem metadata operations, batch them into
1772 : // spawn_blocking calls rather than doing each one as a tokio::fs round-trip.
1773 0 : let span = tracing::Span::current();
1774 0 : let jh = tokio::task::spawn_blocking(move || -> anyhow::Result<usize> {
1775 0 : // Run this synchronous code in the same log context as the outer function that spawned it.
1776 0 : let _span = span.enter();
1777 0 :
1778 0 : tracing::info!("Creating {} directories", create_dirs.len());
1779 0 : for dir in &create_dirs {
1780 0 : if let Err(e) = std::fs::create_dir_all(dir) {
1781 : // Ignore AlreadyExists errors, drop out on all other errors
1782 0 : match e.kind() {
1783 0 : std::io::ErrorKind::AlreadyExists => {}
1784 : _ => {
1785 0 : return Err(anyhow::anyhow!(e).context(format!("Creating {dir}")));
1786 : }
1787 : }
1788 0 : }
1789 : }
1790 :
1791 0 : for child_prefix in child_prefixes {
1792 0 : tracing::info!(
1793 0 : "Hard-linking {} parent layers into child path {}",
1794 0 : parent_layers.len(),
1795 : child_prefix
1796 : );
1797 0 : for relative_layer in &parent_layers {
1798 0 : let parent_path = parent_path.join(relative_layer);
1799 0 : let child_path = child_prefix.join(relative_layer);
1800 0 : if let Err(e) = std::fs::hard_link(&parent_path, &child_path) {
1801 0 : match e.kind() {
1802 0 : std::io::ErrorKind::AlreadyExists => {}
1803 : std::io::ErrorKind::NotFound => {
1804 0 : tracing::info!(
1805 0 : "Layer {} not found during hard-linking, evicted during split?",
1806 : relative_layer
1807 : );
1808 : }
1809 : _ => {
1810 0 : return Err(anyhow::anyhow!(e).context(format!(
1811 0 : "Hard linking {relative_layer} into {child_prefix}"
1812 0 : )))
1813 : }
1814 : }
1815 0 : }
1816 : }
1817 : }
1818 :
1819 : // Durability is not required for correctness, but if we crashed during split and
1820 : // then came restarted with empty timeline dirs, it would be very inefficient to
1821 : // re-populate from remote storage.
1822 0 : tracing::info!("fsyncing {} directories", create_dirs.len());
1823 0 : for dir in create_dirs {
1824 0 : if let Err(e) = crashsafe::fsync(&dir) {
1825 : // Something removed a newly created timeline dir out from underneath us? Extremely
1826 : // unexpected, but not worth panic'ing over as this whole function is just an
1827 : // optimization.
1828 0 : tracing::warn!("Failed to fsync directory {dir}: {e}")
1829 0 : }
1830 : }
1831 :
1832 0 : Ok(parent_layers.len())
1833 0 : });
1834 0 :
1835 0 : match jh.await {
1836 0 : Ok(Ok(layer_count)) => {
1837 0 : tracing::info!(count = layer_count, "Hard linked layers into child shards");
1838 : }
1839 0 : Ok(Err(e)) => {
1840 0 : // This is an optimization, so we tolerate failure.
1841 0 : tracing::warn!("Error hard-linking layers, proceeding anyway: {e}")
1842 : }
1843 0 : Err(e) => {
1844 0 : // This is something totally unexpected like a panic, so bail out.
1845 0 : anyhow::bail!("Error joining hard linking task: {e}");
1846 : }
1847 : }
1848 :
1849 0 : Ok(())
1850 0 : }
1851 :
1852 : ///
1853 : /// Shut down all tenants. This runs as part of pageserver shutdown.
1854 : ///
1855 : /// NB: We leave the tenants in the map, so that they remain accessible through
1856 : /// the management API until we shut it down. If we removed the shut-down tenants
1857 : /// from the tenants map, the management API would return 404 for these tenants,
1858 : /// because TenantsMap::get() now returns `None`.
1859 : /// That could be easily misinterpreted by control plane, the consumer of the
1860 : /// management API. For example, it could attach the tenant on a different pageserver.
1861 : /// We would then be in split-brain once this pageserver restarts.
1862 0 : #[instrument(skip_all)]
1863 : pub(crate) async fn shutdown(&self) {
1864 : self.cancel.cancel();
1865 :
1866 : shutdown_all_tenants0(self.tenants).await
1867 : }
1868 :
1869 0 : pub(crate) async fn detach_tenant(
1870 0 : &self,
1871 0 : conf: &'static PageServerConf,
1872 0 : tenant_shard_id: TenantShardId,
1873 0 : deletion_queue_client: &DeletionQueueClient,
1874 0 : ) -> Result<(), TenantStateError> {
1875 0 : let tmp_path = self
1876 0 : .detach_tenant0(conf, tenant_shard_id, deletion_queue_client)
1877 0 : .await?;
1878 0 : self.background_purges.spawn(tmp_path);
1879 0 :
1880 0 : Ok(())
1881 0 : }
1882 :
1883 0 : async fn detach_tenant0(
1884 0 : &self,
1885 0 : conf: &'static PageServerConf,
1886 0 : tenant_shard_id: TenantShardId,
1887 0 : deletion_queue_client: &DeletionQueueClient,
1888 0 : ) -> Result<Utf8PathBuf, TenantStateError> {
1889 0 : let tenant_dir_rename_operation = |tenant_id_to_clean: TenantShardId| async move {
1890 0 : let local_tenant_directory = conf.tenant_path(&tenant_id_to_clean);
1891 0 : safe_rename_tenant_dir(&local_tenant_directory)
1892 0 : .await
1893 0 : .with_context(|| {
1894 0 : format!("local tenant directory {local_tenant_directory:?} rename")
1895 0 : })
1896 0 : };
1897 :
1898 0 : let removal_result = remove_tenant_from_memory(
1899 0 : self.tenants,
1900 0 : tenant_shard_id,
1901 0 : tenant_dir_rename_operation(tenant_shard_id),
1902 0 : )
1903 0 : .await;
1904 :
1905 : // Flush pending deletions, so that they have a good chance of passing validation
1906 : // before this tenant is potentially re-attached elsewhere.
1907 0 : deletion_queue_client.flush_advisory();
1908 0 :
1909 0 : removal_result
1910 0 : }
1911 :
1912 0 : pub(crate) fn list_tenants(
1913 0 : &self,
1914 0 : ) -> Result<Vec<(TenantShardId, TenantState, Generation)>, TenantMapListError> {
1915 0 : let tenants = self.tenants.read().unwrap();
1916 0 : let m = match &*tenants {
1917 0 : TenantsMap::Initializing => return Err(TenantMapListError::Initializing),
1918 0 : TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => m,
1919 0 : };
1920 0 : Ok(m.iter()
1921 0 : .filter_map(|(id, tenant)| match tenant {
1922 0 : TenantSlot::Attached(tenant) => {
1923 0 : Some((*id, tenant.current_state(), tenant.generation()))
1924 : }
1925 0 : TenantSlot::Secondary(_) => None,
1926 0 : TenantSlot::InProgress(_) => None,
1927 0 : })
1928 0 : .collect())
1929 0 : }
1930 :
1931 : /// Completes an earlier prepared timeline detach ancestor.
1932 0 : pub(crate) async fn complete_detaching_timeline_ancestor(
1933 0 : &self,
1934 0 : tenant_shard_id: TenantShardId,
1935 0 : timeline_id: TimelineId,
1936 0 : prepared: PreparedTimelineDetach,
1937 0 : mut attempt: detach_ancestor::Attempt,
1938 0 : ctx: &RequestContext,
1939 0 : ) -> Result<HashSet<TimelineId>, detach_ancestor::Error> {
1940 : use detach_ancestor::Error;
1941 :
1942 0 : let slot_guard =
1943 0 : tenant_map_acquire_slot(&tenant_shard_id, TenantSlotAcquireMode::MustExist).map_err(
1944 0 : |e| {
1945 : use TenantSlotError::*;
1946 :
1947 0 : match e {
1948 0 : MapState(TenantMapError::ShuttingDown) => Error::ShuttingDown,
1949 0 : NotFound(_) | InProgress | MapState(_) => Error::DetachReparent(e.into()),
1950 : }
1951 0 : },
1952 0 : )?;
1953 :
1954 0 : let tenant = {
1955 0 : let old_slot = slot_guard
1956 0 : .get_old_value()
1957 0 : .as_ref()
1958 0 : .expect("requested MustExist");
1959 :
1960 0 : let Some(tenant) = old_slot.get_attached() else {
1961 0 : return Err(Error::DetachReparent(anyhow::anyhow!(
1962 0 : "Tenant is not in attached state"
1963 0 : )));
1964 : };
1965 :
1966 0 : if !tenant.is_active() {
1967 0 : return Err(Error::DetachReparent(anyhow::anyhow!(
1968 0 : "Tenant is not active"
1969 0 : )));
1970 0 : }
1971 0 :
1972 0 : tenant.clone()
1973 : };
1974 :
1975 0 : let timeline = tenant
1976 0 : .get_timeline(timeline_id, true)
1977 0 : .map_err(Error::NotFound)?;
1978 :
1979 0 : let resp = timeline
1980 0 : .detach_from_ancestor_and_reparent(&tenant, prepared, ctx)
1981 0 : .await?;
1982 :
1983 0 : let mut slot_guard = slot_guard;
1984 :
1985 0 : let tenant = if resp.reset_tenant_required() {
1986 0 : attempt.before_reset_tenant();
1987 0 :
1988 0 : let (_guard, progress) = utils::completion::channel();
1989 0 : match tenant.shutdown(progress, ShutdownMode::Hard).await {
1990 0 : Ok(()) => {
1991 0 : slot_guard.drop_old_value().expect("it was just shutdown");
1992 0 : }
1993 0 : Err(_barrier) => {
1994 0 : slot_guard.revert();
1995 0 : // this really should not happen, at all, unless a shutdown without acquiring
1996 0 : // tenant slot was already going? regardless, on restart the attempt tracking
1997 0 : // will reset to retryable.
1998 0 : return Err(Error::ShuttingDown);
1999 : }
2000 : }
2001 :
2002 0 : let tenant_path = self.conf.tenant_path(&tenant_shard_id);
2003 0 : let config = Tenant::load_tenant_config(self.conf, &tenant_shard_id)
2004 0 : .map_err(|e| Error::DetachReparent(e.into()))?;
2005 :
2006 0 : let shard_identity = config.shard;
2007 0 : let tenant = tenant_spawn(
2008 0 : self.conf,
2009 0 : tenant_shard_id,
2010 0 : &tenant_path,
2011 0 : self.resources.clone(),
2012 0 : AttachedTenantConf::try_from(config).map_err(Error::DetachReparent)?,
2013 0 : shard_identity,
2014 0 : None,
2015 0 : SpawnMode::Eager,
2016 0 : ctx,
2017 0 : )
2018 0 : .map_err(|_| Error::ShuttingDown)?;
2019 :
2020 : {
2021 0 : let mut g = tenant.ongoing_timeline_detach.lock().unwrap();
2022 0 : assert!(
2023 0 : g.is_none(),
2024 0 : "there cannot be any new timeline detach ancestor on newly created tenant"
2025 : );
2026 0 : *g = Some((attempt.timeline_id, attempt.new_barrier()));
2027 0 : }
2028 0 :
2029 0 : // if we bail out here, we will not allow a new attempt, which should be fine.
2030 0 : // pageserver should be shutting down regardless? tenant_reset would help, unless it
2031 0 : // runs into the same problem.
2032 0 : slot_guard
2033 0 : .upsert(TenantSlot::Attached(tenant.clone()))
2034 0 : .map_err(|e| match e {
2035 0 : TenantSlotUpsertError::ShuttingDown(_) => Error::ShuttingDown,
2036 0 : other => Error::DetachReparent(other.into()),
2037 0 : })?;
2038 0 : tenant
2039 : } else {
2040 0 : tracing::info!("skipping tenant_reset as no changes made required it");
2041 0 : tenant
2042 : };
2043 :
2044 0 : if let Some(reparented) = resp.completed() {
2045 : // finally ask the restarted tenant to complete the detach
2046 : //
2047 : // rationale for 9999s: we don't really have a timetable here; if retried, the caller
2048 : // will get an 503.
2049 0 : tenant
2050 0 : .wait_to_become_active(std::time::Duration::from_secs(9999))
2051 0 : .await
2052 0 : .map_err(|e| {
2053 : use pageserver_api::models::TenantState;
2054 : use GetActiveTenantError::{Cancelled, WillNotBecomeActive};
2055 0 : match e {
2056 : Cancelled | WillNotBecomeActive(TenantState::Stopping { .. }) => {
2057 0 : Error::ShuttingDown
2058 : }
2059 0 : other => Error::Complete(other.into()),
2060 : }
2061 0 : })?;
2062 :
2063 0 : utils::pausable_failpoint!(
2064 0 : "timeline-detach-ancestor::after_activating_before_finding-pausable"
2065 0 : );
2066 :
2067 0 : let timeline = tenant
2068 0 : .get_timeline(attempt.timeline_id, true)
2069 0 : .map_err(Error::NotFound)?;
2070 :
2071 0 : timeline
2072 0 : .complete_detaching_timeline_ancestor(&tenant, attempt, ctx)
2073 0 : .await
2074 0 : .map(|()| reparented)
2075 : } else {
2076 : // at least the latest versions have now been downloaded and refreshed; be ready to
2077 : // retry another time.
2078 0 : Err(Error::FailedToReparentAll)
2079 : }
2080 0 : }
2081 :
2082 : /// A page service client sends a TenantId, and to look up the correct Tenant we must
2083 : /// resolve this to a fully qualified TenantShardId.
2084 : ///
2085 : /// During shard splits: we shall see parent shards in InProgress state and skip them, and
2086 : /// instead match on child shards which should appear in Attached state. Very early in a shard
2087 : /// split, or in other cases where a shard is InProgress, we will return our own InProgress result
2088 : /// to instruct the caller to wait for that to finish before querying again.
2089 0 : pub(crate) fn resolve_attached_shard(
2090 0 : &self,
2091 0 : tenant_id: &TenantId,
2092 0 : selector: ShardSelector,
2093 0 : ) -> ShardResolveResult {
2094 0 : let tenants = self.tenants.read().unwrap();
2095 0 : let mut want_shard = None;
2096 0 : let mut any_in_progress = None;
2097 0 :
2098 0 : match &*tenants {
2099 0 : TenantsMap::Initializing => ShardResolveResult::NotFound,
2100 0 : TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => {
2101 0 : for slot in m.range(TenantShardId::tenant_range(*tenant_id)) {
2102 : // Ignore all slots that don't contain an attached tenant
2103 0 : let tenant = match &slot.1 {
2104 0 : TenantSlot::Attached(t) => t,
2105 0 : TenantSlot::InProgress(barrier) => {
2106 0 : // We might still find a usable shard, but in case we don't, remember that
2107 0 : // we saw at least one InProgress slot, so that we can distinguish this case
2108 0 : // from a simple NotFound in our return value.
2109 0 : any_in_progress = Some(barrier.clone());
2110 0 : continue;
2111 : }
2112 0 : _ => continue,
2113 : };
2114 :
2115 0 : match selector {
2116 0 : ShardSelector::Zero if slot.0.shard_number == ShardNumber(0) => {
2117 0 : return ShardResolveResult::Found(tenant.clone())
2118 : }
2119 0 : ShardSelector::Page(key) => {
2120 0 : // First slot we see for this tenant, calculate the expected shard number
2121 0 : // for the key: we will use this for checking if this and subsequent
2122 0 : // slots contain the key, rather than recalculating the hash each time.
2123 0 : if want_shard.is_none() {
2124 0 : want_shard = Some(tenant.shard_identity.get_shard_number(&key));
2125 0 : }
2126 :
2127 0 : if Some(tenant.shard_identity.number) == want_shard {
2128 0 : return ShardResolveResult::Found(tenant.clone());
2129 0 : }
2130 : }
2131 0 : ShardSelector::Known(shard)
2132 0 : if tenant.shard_identity.shard_index() == shard =>
2133 0 : {
2134 0 : return ShardResolveResult::Found(tenant.clone());
2135 : }
2136 0 : _ => continue,
2137 : }
2138 : }
2139 :
2140 : // Fall through: we didn't find a slot that was in Attached state & matched our selector. If
2141 : // we found one or more InProgress slot, indicate to caller that they should retry later. Otherwise
2142 : // this requested shard simply isn't found.
2143 0 : if let Some(barrier) = any_in_progress {
2144 0 : ShardResolveResult::InProgress(barrier)
2145 : } else {
2146 0 : ShardResolveResult::NotFound
2147 : }
2148 : }
2149 : }
2150 0 : }
2151 :
2152 : /// Calculate the tenant shards' contributions to this pageserver's utilization metrics. The
2153 : /// returned values are:
2154 : /// - the number of bytes of local disk space this pageserver's shards are requesting, i.e.
2155 : /// how much space they would use if not impacted by disk usage eviction.
2156 : /// - the number of tenant shards currently on this pageserver, including attached
2157 : /// and secondary.
2158 : ///
2159 : /// This function is quite expensive: callers are expected to cache the result and
2160 : /// limit how often they call it.
2161 0 : pub(crate) fn calculate_utilization(&self) -> Result<(u64, u32), TenantMapListError> {
2162 0 : let tenants = self.tenants.read().unwrap();
2163 0 : let m = match &*tenants {
2164 0 : TenantsMap::Initializing => return Err(TenantMapListError::Initializing),
2165 0 : TenantsMap::Open(m) | TenantsMap::ShuttingDown(m) => m,
2166 0 : };
2167 0 : let shard_count = m.len();
2168 0 : let mut wanted_bytes = 0;
2169 :
2170 0 : for tenant_slot in m.values() {
2171 0 : match tenant_slot {
2172 0 : TenantSlot::InProgress(_barrier) => {
2173 0 : // While a slot is being changed, we can't know how much storage it wants. This
2174 0 : // means this function's output can fluctuate if a lot of changes are going on
2175 0 : // (such as transitions from secondary to attached).
2176 0 : //
2177 0 : // We could wait for the barrier and retry, but it's important that the utilization
2178 0 : // API is responsive, and the data quality impact is not very significant.
2179 0 : continue;
2180 : }
2181 0 : TenantSlot::Attached(tenant) => {
2182 0 : wanted_bytes += tenant.local_storage_wanted();
2183 0 : }
2184 0 : TenantSlot::Secondary(secondary) => {
2185 0 : let progress = secondary.progress.lock().unwrap();
2186 0 : wanted_bytes += if progress.heatmap_mtime.is_some() {
2187 : // If we have heatmap info, then we will 'want' the sum
2188 : // of the size of layers in the heatmap: this is how much space
2189 : // we would use if not doing any eviction.
2190 0 : progress.bytes_total
2191 : } else {
2192 : // In the absence of heatmap info, assume that the secondary location simply
2193 : // needs as much space as it is currently using.
2194 0 : secondary.resident_size_metric.get()
2195 : }
2196 : }
2197 : }
2198 : }
2199 :
2200 0 : Ok((wanted_bytes, shard_count as u32))
2201 0 : }
2202 : }
2203 :
2204 0 : #[derive(Debug, thiserror::Error)]
2205 : pub(crate) enum GetTenantError {
2206 : /// NotFound is a TenantId rather than TenantShardId, because this error type is used from
2207 : /// getters that use a TenantId and a ShardSelector, not just getters that target a specific shard.
2208 : #[error("Tenant {0} not found")]
2209 : NotFound(TenantId),
2210 :
2211 : #[error("Tenant {0} is not active")]
2212 : NotActive(TenantShardId),
2213 :
2214 : // Initializing or shutting down: cannot authoritatively say whether we have this tenant
2215 : #[error("Tenant map is not available: {0}")]
2216 : MapState(#[from] TenantMapError),
2217 : }
2218 :
2219 0 : #[derive(thiserror::Error, Debug)]
2220 : pub(crate) enum GetActiveTenantError {
2221 : /// We may time out either while TenantSlot is InProgress, or while the Tenant
2222 : /// is in a non-Active state
2223 : #[error(
2224 : "Timed out waiting {wait_time:?} for tenant active state. Latest state: {latest_state:?}"
2225 : )]
2226 : WaitForActiveTimeout {
2227 : latest_state: Option<TenantState>,
2228 : wait_time: Duration,
2229 : },
2230 :
2231 : /// The TenantSlot is absent, or in secondary mode
2232 : #[error(transparent)]
2233 : NotFound(#[from] GetTenantError),
2234 :
2235 : /// Cancellation token fired while we were waiting
2236 : #[error("cancelled")]
2237 : Cancelled,
2238 :
2239 : /// Tenant exists, but is in a state that cannot become active (e.g. Stopping, Broken)
2240 : #[error("will not become active. Current state: {0}")]
2241 : WillNotBecomeActive(TenantState),
2242 :
2243 : /// Broken is logically a subset of WillNotBecomeActive, but a distinct error is useful as
2244 : /// WillNotBecomeActive is a permitted error under some circumstances, whereas broken should
2245 : /// never happen.
2246 : #[error("Tenant is broken: {0}")]
2247 : Broken(String),
2248 :
2249 : #[error("reconnect to switch tenant id")]
2250 : SwitchedTenant,
2251 : }
2252 :
2253 0 : #[derive(Debug, thiserror::Error)]
2254 : pub(crate) enum DeleteTimelineError {
2255 : #[error("Tenant {0}")]
2256 : Tenant(#[from] GetTenantError),
2257 :
2258 : #[error("Timeline {0}")]
2259 : Timeline(#[from] crate::tenant::DeleteTimelineError),
2260 : }
2261 :
2262 0 : #[derive(Debug, thiserror::Error)]
2263 : pub(crate) enum TenantStateError {
2264 : #[error("Tenant {0} is stopping")]
2265 : IsStopping(TenantShardId),
2266 : #[error(transparent)]
2267 : SlotError(#[from] TenantSlotError),
2268 : #[error(transparent)]
2269 : SlotUpsertError(#[from] TenantSlotUpsertError),
2270 : #[error(transparent)]
2271 : Other(#[from] anyhow::Error),
2272 : }
2273 :
2274 0 : #[derive(Debug, thiserror::Error)]
2275 : pub(crate) enum TenantMapListError {
2276 : #[error("tenant map is still initiailizing")]
2277 : Initializing,
2278 : }
2279 :
2280 0 : #[derive(Debug, thiserror::Error)]
2281 : pub(crate) enum TenantMapInsertError {
2282 : #[error(transparent)]
2283 : SlotError(#[from] TenantSlotError),
2284 : #[error(transparent)]
2285 : SlotUpsertError(#[from] TenantSlotUpsertError),
2286 : #[error(transparent)]
2287 : Other(#[from] anyhow::Error),
2288 : }
2289 :
2290 : /// Superset of TenantMapError: issues that can occur when acquiring a slot
2291 : /// for a particular tenant ID.
2292 0 : #[derive(Debug, thiserror::Error)]
2293 : pub(crate) enum TenantSlotError {
2294 : /// When acquiring a slot with the expectation that the tenant already exists.
2295 : #[error("Tenant {0} not found")]
2296 : NotFound(TenantShardId),
2297 :
2298 : // Tried to read a slot that is currently being mutated by another administrative
2299 : // operation.
2300 : #[error("tenant has a state change in progress, try again later")]
2301 : InProgress,
2302 :
2303 : #[error(transparent)]
2304 : MapState(#[from] TenantMapError),
2305 : }
2306 :
2307 : /// Superset of TenantMapError: issues that can occur when using a SlotGuard
2308 : /// to insert a new value.
2309 0 : #[derive(thiserror::Error)]
2310 : pub(crate) enum TenantSlotUpsertError {
2311 : /// An error where the slot is in an unexpected state, indicating a code bug
2312 : #[error("Internal error updating Tenant")]
2313 : InternalError(Cow<'static, str>),
2314 :
2315 : #[error(transparent)]
2316 : MapState(TenantMapError),
2317 :
2318 : // If we encounter TenantManager shutdown during upsert, we must carry the Completion
2319 : // from the SlotGuard, so that the caller can hold it while they clean up: otherwise
2320 : // TenantManager shutdown might race ahead before we're done cleaning up any Tenant that
2321 : // was protected by the SlotGuard.
2322 : #[error("Shutting down")]
2323 : ShuttingDown((TenantSlot, utils::completion::Completion)),
2324 : }
2325 :
2326 : impl std::fmt::Debug for TenantSlotUpsertError {
2327 0 : fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
2328 0 : match self {
2329 0 : Self::InternalError(reason) => write!(f, "Internal Error {reason}"),
2330 0 : Self::MapState(map_error) => write!(f, "Tenant map state: {map_error:?}"),
2331 0 : Self::ShuttingDown(_completion) => write!(f, "Tenant map shutting down"),
2332 : }
2333 0 : }
2334 : }
2335 :
2336 0 : #[derive(Debug, thiserror::Error)]
2337 : enum TenantSlotDropError {
2338 : /// It is only legal to drop a TenantSlot if its contents are fully shut down
2339 : #[error("Tenant was not shut down")]
2340 : NotShutdown,
2341 : }
2342 :
2343 : /// Errors that can happen any time we are walking the tenant map to try and acquire
2344 : /// the TenantSlot for a particular tenant.
2345 0 : #[derive(Debug, thiserror::Error)]
2346 : pub enum TenantMapError {
2347 : // Tried to read while initializing
2348 : #[error("tenant map is still initializing")]
2349 : StillInitializing,
2350 :
2351 : // Tried to read while shutting down
2352 : #[error("tenant map is shutting down")]
2353 : ShuttingDown,
2354 : }
2355 :
2356 : /// Guards a particular tenant_id's content in the TenantsMap.
2357 : ///
2358 : /// While this structure exists, the TenantsMap will contain a [`TenantSlot::InProgress`]
2359 : /// for this tenant, which acts as a marker for any operations targeting
2360 : /// this tenant to retry later, or wait for the InProgress state to end.
2361 : ///
2362 : /// This structure enforces the important invariant that we do not have overlapping
2363 : /// tasks that will try use local storage for a the same tenant ID: we enforce that
2364 : /// the previous contents of a slot have been shut down before the slot can be
2365 : /// left empty or used for something else
2366 : ///
2367 : /// Holders of a SlotGuard should explicitly dispose of it, using either `upsert`
2368 : /// to provide a new value, or `revert` to put the slot back into its initial
2369 : /// state. If the SlotGuard is dropped without calling either of these, then
2370 : /// we will leave the slot empty if our `old_value` is already shut down, else
2371 : /// we will replace the slot with `old_value` (equivalent to doing a revert).
2372 : ///
2373 : /// The `old_value` may be dropped before the SlotGuard is dropped, by calling
2374 : /// `drop_old_value`. It is an error to call this without shutting down
2375 : /// the conents of `old_value`.
2376 : pub struct SlotGuard {
2377 : tenant_shard_id: TenantShardId,
2378 : old_value: Option<TenantSlot>,
2379 : upserted: bool,
2380 :
2381 : /// [`TenantSlot::InProgress`] carries the corresponding Barrier: it will
2382 : /// release any waiters as soon as this SlotGuard is dropped.
2383 : completion: utils::completion::Completion,
2384 : }
2385 :
2386 : impl SlotGuard {
2387 6 : fn new(
2388 6 : tenant_shard_id: TenantShardId,
2389 6 : old_value: Option<TenantSlot>,
2390 6 : completion: utils::completion::Completion,
2391 6 : ) -> Self {
2392 6 : Self {
2393 6 : tenant_shard_id,
2394 6 : old_value,
2395 6 : upserted: false,
2396 6 : completion,
2397 6 : }
2398 6 : }
2399 :
2400 : /// Get any value that was present in the slot before we acquired ownership
2401 : /// of it: in state transitions, this will be the old state.
2402 : ///
2403 : // FIXME: get_ prefix
2404 : // FIXME: this should be .as_ref() -- unsure why no clippy
2405 6 : fn get_old_value(&self) -> &Option<TenantSlot> {
2406 6 : &self.old_value
2407 6 : }
2408 :
2409 : /// Emplace a new value in the slot. This consumes the guard, and after
2410 : /// returning, the slot is no longer protected from concurrent changes.
2411 0 : fn upsert(mut self, new_value: TenantSlot) -> Result<(), TenantSlotUpsertError> {
2412 0 : if !self.old_value_is_shutdown() {
2413 : // This is a bug: callers should never try to drop an old value without
2414 : // shutting it down
2415 0 : return Err(TenantSlotUpsertError::InternalError(
2416 0 : "Old TenantSlot value not shut down".into(),
2417 0 : ));
2418 0 : }
2419 :
2420 0 : let replaced = {
2421 0 : let mut locked = TENANTS.write().unwrap();
2422 0 :
2423 0 : if let TenantSlot::InProgress(_) = new_value {
2424 : // It is never expected to try and upsert InProgress via this path: it should
2425 : // only be written via the tenant_map_acquire_slot path. If we hit this it's a bug.
2426 0 : return Err(TenantSlotUpsertError::InternalError(
2427 0 : "Attempt to upsert an InProgress state".into(),
2428 0 : ));
2429 0 : }
2430 :
2431 0 : let m = match &mut *locked {
2432 : TenantsMap::Initializing => {
2433 0 : return Err(TenantSlotUpsertError::MapState(
2434 0 : TenantMapError::StillInitializing,
2435 0 : ))
2436 : }
2437 : TenantsMap::ShuttingDown(_) => {
2438 0 : return Err(TenantSlotUpsertError::ShuttingDown((
2439 0 : new_value,
2440 0 : self.completion.clone(),
2441 0 : )));
2442 : }
2443 0 : TenantsMap::Open(m) => m,
2444 0 : };
2445 0 :
2446 0 : METRICS.slot_inserted(&new_value);
2447 0 :
2448 0 : let replaced = m.insert(self.tenant_shard_id, new_value);
2449 0 : self.upserted = true;
2450 0 : if let Some(replaced) = replaced.as_ref() {
2451 0 : METRICS.slot_removed(replaced);
2452 0 : }
2453 :
2454 0 : replaced
2455 : };
2456 :
2457 : // Sanity check: on an upsert we should always be replacing an InProgress marker
2458 0 : match replaced {
2459 : Some(TenantSlot::InProgress(_)) => {
2460 : // Expected case: we find our InProgress in the map: nothing should have
2461 : // replaced it because the code that acquires slots will not grant another
2462 : // one for the same TenantId.
2463 0 : Ok(())
2464 : }
2465 : None => {
2466 0 : METRICS.unexpected_errors.inc();
2467 0 : error!(
2468 : tenant_shard_id = %self.tenant_shard_id,
2469 0 : "Missing InProgress marker during tenant upsert, this is a bug."
2470 : );
2471 0 : Err(TenantSlotUpsertError::InternalError(
2472 0 : "Missing InProgress marker during tenant upsert".into(),
2473 0 : ))
2474 : }
2475 0 : Some(slot) => {
2476 0 : METRICS.unexpected_errors.inc();
2477 0 : error!(tenant_shard_id=%self.tenant_shard_id, "Unexpected contents of TenantSlot during upsert, this is a bug. Contents: {:?}", slot);
2478 0 : Err(TenantSlotUpsertError::InternalError(
2479 0 : "Unexpected contents of TenantSlot".into(),
2480 0 : ))
2481 : }
2482 : }
2483 0 : }
2484 :
2485 : /// Replace the InProgress slot with whatever was in the guard when we started
2486 0 : fn revert(mut self) {
2487 0 : if let Some(value) = self.old_value.take() {
2488 0 : match self.upsert(value) {
2489 0 : Err(TenantSlotUpsertError::InternalError(_)) => {
2490 0 : // We already logged the error, nothing else we can do.
2491 0 : }
2492 : Err(
2493 : TenantSlotUpsertError::MapState(_) | TenantSlotUpsertError::ShuttingDown(_),
2494 0 : ) => {
2495 0 : // If the map is shutting down, we need not replace anything
2496 0 : }
2497 0 : Ok(()) => {}
2498 : }
2499 0 : }
2500 0 : }
2501 :
2502 : /// We may never drop our old value until it is cleanly shut down: otherwise we might leave
2503 : /// rogue background tasks that would write to the local tenant directory that this guard
2504 : /// is responsible for protecting
2505 6 : fn old_value_is_shutdown(&self) -> bool {
2506 6 : match self.old_value.as_ref() {
2507 6 : Some(TenantSlot::Attached(tenant)) => tenant.gate.close_complete(),
2508 0 : Some(TenantSlot::Secondary(secondary_tenant)) => secondary_tenant.gate.close_complete(),
2509 : Some(TenantSlot::InProgress(_)) => {
2510 : // A SlotGuard cannot be constructed for a slot that was already InProgress
2511 0 : unreachable!()
2512 : }
2513 0 : None => true,
2514 : }
2515 6 : }
2516 :
2517 : /// The guard holder is done with the old value of the slot: they are obliged to already
2518 : /// shut it down before we reach this point.
2519 6 : fn drop_old_value(&mut self) -> Result<(), TenantSlotDropError> {
2520 6 : if !self.old_value_is_shutdown() {
2521 0 : Err(TenantSlotDropError::NotShutdown)
2522 : } else {
2523 6 : self.old_value.take();
2524 6 : Ok(())
2525 : }
2526 6 : }
2527 : }
2528 :
2529 : impl Drop for SlotGuard {
2530 6 : fn drop(&mut self) {
2531 6 : if self.upserted {
2532 0 : return;
2533 6 : }
2534 6 : // Our old value is already shutdown, or it never existed: it is safe
2535 6 : // for us to fully release the TenantSlot back into an empty state
2536 6 :
2537 6 : let mut locked = TENANTS.write().unwrap();
2538 :
2539 6 : let m = match &mut *locked {
2540 : TenantsMap::Initializing => {
2541 : // There is no map, this should never happen.
2542 6 : return;
2543 : }
2544 : TenantsMap::ShuttingDown(_) => {
2545 : // When we transition to shutdown, InProgress elements are removed
2546 : // from the map, so we do not need to clean up our Inprogress marker.
2547 : // See [`shutdown_all_tenants0`]
2548 0 : return;
2549 : }
2550 0 : TenantsMap::Open(m) => m,
2551 : };
2552 :
2553 : use std::collections::btree_map::Entry;
2554 0 : match m.entry(self.tenant_shard_id) {
2555 0 : Entry::Occupied(mut entry) => {
2556 0 : if !matches!(entry.get(), TenantSlot::InProgress(_)) {
2557 0 : METRICS.unexpected_errors.inc();
2558 0 : error!(tenant_shard_id=%self.tenant_shard_id, "Unexpected contents of TenantSlot during drop, this is a bug. Contents: {:?}", entry.get());
2559 0 : }
2560 :
2561 0 : if self.old_value_is_shutdown() {
2562 0 : METRICS.slot_removed(entry.get());
2563 0 : entry.remove();
2564 0 : } else {
2565 0 : let inserting = self.old_value.take().unwrap();
2566 0 : METRICS.slot_inserted(&inserting);
2567 0 : let replaced = entry.insert(inserting);
2568 0 : METRICS.slot_removed(&replaced);
2569 0 : }
2570 : }
2571 : Entry::Vacant(_) => {
2572 0 : METRICS.unexpected_errors.inc();
2573 0 : error!(
2574 : tenant_shard_id = %self.tenant_shard_id,
2575 0 : "Missing InProgress marker during SlotGuard drop, this is a bug."
2576 : );
2577 : }
2578 : }
2579 6 : }
2580 : }
2581 :
2582 : enum TenantSlotPeekMode {
2583 : /// In Read mode, peek will be permitted to see the slots even if the pageserver is shutting down
2584 : Read,
2585 : /// In Write mode, trying to peek at a slot while the pageserver is shutting down is an error
2586 : Write,
2587 : }
2588 :
2589 0 : fn tenant_map_peek_slot<'a>(
2590 0 : tenants: &'a std::sync::RwLockReadGuard<'a, TenantsMap>,
2591 0 : tenant_shard_id: &TenantShardId,
2592 0 : mode: TenantSlotPeekMode,
2593 0 : ) -> Result<Option<&'a TenantSlot>, TenantMapError> {
2594 0 : match tenants.deref() {
2595 0 : TenantsMap::Initializing => Err(TenantMapError::StillInitializing),
2596 0 : TenantsMap::ShuttingDown(m) => match mode {
2597 : TenantSlotPeekMode::Read => Ok(Some(
2598 : // When reading in ShuttingDown state, we must translate None results
2599 : // into a ShuttingDown error, because absence of a tenant shard ID in the map
2600 : // isn't a reliable indicator of the tenant being gone: it might have been
2601 : // InProgress when shutdown started, and cleaned up from that state such
2602 : // that it's now no longer in the map. Callers will have to wait until
2603 : // we next start up to get a proper answer. This avoids incorrect 404 API responses.
2604 0 : m.get(tenant_shard_id).ok_or(TenantMapError::ShuttingDown)?,
2605 : )),
2606 0 : TenantSlotPeekMode::Write => Err(TenantMapError::ShuttingDown),
2607 : },
2608 0 : TenantsMap::Open(m) => Ok(m.get(tenant_shard_id)),
2609 : }
2610 0 : }
2611 :
2612 : enum TenantSlotAcquireMode {
2613 : /// Acquire the slot irrespective of current state, or whether it already exists
2614 : Any,
2615 : /// Return an error if trying to acquire a slot and it doesn't already exist
2616 : MustExist,
2617 : }
2618 :
2619 0 : fn tenant_map_acquire_slot(
2620 0 : tenant_shard_id: &TenantShardId,
2621 0 : mode: TenantSlotAcquireMode,
2622 0 : ) -> Result<SlotGuard, TenantSlotError> {
2623 0 : tenant_map_acquire_slot_impl(tenant_shard_id, &TENANTS, mode)
2624 0 : }
2625 :
2626 6 : fn tenant_map_acquire_slot_impl(
2627 6 : tenant_shard_id: &TenantShardId,
2628 6 : tenants: &std::sync::RwLock<TenantsMap>,
2629 6 : mode: TenantSlotAcquireMode,
2630 6 : ) -> Result<SlotGuard, TenantSlotError> {
2631 : use TenantSlotAcquireMode::*;
2632 6 : METRICS.tenant_slot_writes.inc();
2633 6 :
2634 6 : let mut locked = tenants.write().unwrap();
2635 6 : let span = tracing::info_span!("acquire_slot", tenant_id=%tenant_shard_id.tenant_id, shard_id = %tenant_shard_id.shard_slug());
2636 6 : let _guard = span.enter();
2637 :
2638 6 : let m = match &mut *locked {
2639 0 : TenantsMap::Initializing => return Err(TenantMapError::StillInitializing.into()),
2640 0 : TenantsMap::ShuttingDown(_) => return Err(TenantMapError::ShuttingDown.into()),
2641 6 : TenantsMap::Open(m) => m,
2642 : };
2643 :
2644 : use std::collections::btree_map::Entry;
2645 :
2646 6 : let entry = m.entry(*tenant_shard_id);
2647 6 :
2648 6 : match entry {
2649 0 : Entry::Vacant(v) => match mode {
2650 : MustExist => {
2651 0 : tracing::debug!("Vacant && MustExist: return NotFound");
2652 0 : Err(TenantSlotError::NotFound(*tenant_shard_id))
2653 : }
2654 : _ => {
2655 0 : let (completion, barrier) = utils::completion::channel();
2656 0 : let inserting = TenantSlot::InProgress(barrier);
2657 0 : METRICS.slot_inserted(&inserting);
2658 0 : v.insert(inserting);
2659 0 : tracing::debug!("Vacant, inserted InProgress");
2660 0 : Ok(SlotGuard::new(*tenant_shard_id, None, completion))
2661 : }
2662 : },
2663 6 : Entry::Occupied(mut o) => {
2664 6 : // Apply mode-driven checks
2665 6 : match (o.get(), mode) {
2666 : (TenantSlot::InProgress(_), _) => {
2667 0 : tracing::debug!("Occupied, failing for InProgress");
2668 0 : Err(TenantSlotError::InProgress)
2669 : }
2670 : _ => {
2671 : // Happy case: the slot was not in any state that violated our mode
2672 6 : let (completion, barrier) = utils::completion::channel();
2673 6 : let in_progress = TenantSlot::InProgress(barrier);
2674 6 : METRICS.slot_inserted(&in_progress);
2675 6 : let old_value = o.insert(in_progress);
2676 6 : METRICS.slot_removed(&old_value);
2677 6 : tracing::debug!("Occupied, replaced with InProgress");
2678 6 : Ok(SlotGuard::new(
2679 6 : *tenant_shard_id,
2680 6 : Some(old_value),
2681 6 : completion,
2682 6 : ))
2683 : }
2684 : }
2685 : }
2686 : }
2687 6 : }
2688 :
2689 : /// Stops and removes the tenant from memory, if it's not [`TenantState::Stopping`] already, bails otherwise.
2690 : /// Allows to remove other tenant resources manually, via `tenant_cleanup`.
2691 : /// If the cleanup fails, tenant will stay in memory in [`TenantState::Broken`] state, and another removal
2692 : /// operation would be needed to remove it.
2693 6 : async fn remove_tenant_from_memory<V, F>(
2694 6 : tenants: &std::sync::RwLock<TenantsMap>,
2695 6 : tenant_shard_id: TenantShardId,
2696 6 : tenant_cleanup: F,
2697 6 : ) -> Result<V, TenantStateError>
2698 6 : where
2699 6 : F: std::future::Future<Output = anyhow::Result<V>>,
2700 6 : {
2701 6 : let mut slot_guard =
2702 6 : tenant_map_acquire_slot_impl(&tenant_shard_id, tenants, TenantSlotAcquireMode::MustExist)?;
2703 :
2704 : // allow pageserver shutdown to await for our completion
2705 6 : let (_guard, progress) = completion::channel();
2706 :
2707 : // The SlotGuard allows us to manipulate the Tenant object without fear of some
2708 : // concurrent API request doing something else for the same tenant ID.
2709 6 : let attached_tenant = match slot_guard.get_old_value() {
2710 6 : Some(TenantSlot::Attached(tenant)) => {
2711 6 : // whenever we remove a tenant from memory, we don't want to flush and wait for upload
2712 6 : let shutdown_mode = ShutdownMode::Hard;
2713 6 :
2714 6 : // shutdown is sure to transition tenant to stopping, and wait for all tasks to complete, so
2715 6 : // that we can continue safely to cleanup.
2716 6 : match tenant.shutdown(progress, shutdown_mode).await {
2717 6 : Ok(()) => {}
2718 0 : Err(_other) => {
2719 0 : // if pageserver shutdown or other detach/ignore is already ongoing, we don't want to
2720 0 : // wait for it but return an error right away because these are distinct requests.
2721 0 : slot_guard.revert();
2722 0 : return Err(TenantStateError::IsStopping(tenant_shard_id));
2723 : }
2724 : }
2725 6 : Some(tenant)
2726 : }
2727 0 : Some(TenantSlot::Secondary(secondary_state)) => {
2728 0 : tracing::info!("Shutting down in secondary mode");
2729 0 : secondary_state.shutdown().await;
2730 0 : None
2731 : }
2732 : Some(TenantSlot::InProgress(_)) => {
2733 : // Acquiring a slot guarantees its old value was not InProgress
2734 0 : unreachable!();
2735 : }
2736 0 : None => None,
2737 : };
2738 :
2739 6 : match tenant_cleanup
2740 6 : .await
2741 6 : .with_context(|| format!("Failed to run cleanup for tenant {tenant_shard_id}"))
2742 : {
2743 6 : Ok(hook_value) => {
2744 6 : // Success: drop the old TenantSlot::Attached.
2745 6 : slot_guard
2746 6 : .drop_old_value()
2747 6 : .expect("We just called shutdown");
2748 6 :
2749 6 : Ok(hook_value)
2750 : }
2751 0 : Err(e) => {
2752 : // If we had a Tenant, set it to Broken and put it back in the TenantsMap
2753 0 : if let Some(attached_tenant) = attached_tenant {
2754 0 : attached_tenant.set_broken(e.to_string()).await;
2755 0 : }
2756 : // Leave the broken tenant in the map
2757 0 : slot_guard.revert();
2758 0 :
2759 0 : Err(TenantStateError::Other(e))
2760 : }
2761 : }
2762 6 : }
2763 :
2764 : use {
2765 : crate::repository::GcResult, pageserver_api::models::TimelineGcRequest,
2766 : utils::http::error::ApiError,
2767 : };
2768 :
2769 0 : #[instrument(skip_all, fields(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id))]
2770 : pub(crate) async fn immediate_gc(
2771 : tenant_shard_id: TenantShardId,
2772 : timeline_id: TimelineId,
2773 : gc_req: TimelineGcRequest,
2774 : cancel: CancellationToken,
2775 : ctx: &RequestContext,
2776 : ) -> Result<GcResult, ApiError> {
2777 : let tenant = {
2778 : let guard = TENANTS.read().unwrap();
2779 : guard
2780 : .get(&tenant_shard_id)
2781 : .cloned()
2782 0 : .with_context(|| format!("tenant {tenant_shard_id}"))
2783 0 : .map_err(|e| ApiError::NotFound(e.into()))?
2784 : };
2785 :
2786 0 : let gc_horizon = gc_req.gc_horizon.unwrap_or_else(|| tenant.get_gc_horizon());
2787 : // Use tenant's pitr setting
2788 : let pitr = tenant.get_pitr_interval();
2789 :
2790 : tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
2791 :
2792 : // Run in task_mgr to avoid race with tenant_detach operation
2793 : let ctx: RequestContext =
2794 : ctx.detached_child(TaskKind::GarbageCollector, DownloadBehavior::Download);
2795 :
2796 0 : let _gate_guard = tenant.gate.enter().map_err(|_| ApiError::ShuttingDown)?;
2797 :
2798 : fail::fail_point!("immediate_gc_task_pre");
2799 :
2800 : #[allow(unused_mut)]
2801 : let mut result = tenant
2802 : .gc_iteration(Some(timeline_id), gc_horizon, pitr, &cancel, &ctx)
2803 : .await;
2804 : // FIXME: `gc_iteration` can return an error for multiple reasons; we should handle it
2805 : // better once the types support it.
2806 :
2807 : #[cfg(feature = "testing")]
2808 : {
2809 : // we need to synchronize with drop completion for python tests without polling for
2810 : // log messages
2811 : if let Ok(result) = result.as_mut() {
2812 : let mut js = tokio::task::JoinSet::new();
2813 : for layer in std::mem::take(&mut result.doomed_layers) {
2814 : js.spawn(layer.wait_drop());
2815 : }
2816 : tracing::info!(
2817 : total = js.len(),
2818 : "starting to wait for the gc'd layers to be dropped"
2819 : );
2820 : while let Some(res) = js.join_next().await {
2821 : res.expect("wait_drop should not panic");
2822 : }
2823 : }
2824 :
2825 : let timeline = tenant.get_timeline(timeline_id, false).ok();
2826 0 : let rtc = timeline.as_ref().map(|x| &x.remote_client);
2827 :
2828 : if let Some(rtc) = rtc {
2829 : // layer drops schedule actions on remote timeline client to actually do the
2830 : // deletions; don't care about the shutdown error, just exit fast
2831 : drop(rtc.wait_completion().await);
2832 : }
2833 : }
2834 :
2835 0 : result.map_err(|e| match e {
2836 0 : GcError::TenantCancelled | GcError::TimelineCancelled => ApiError::ShuttingDown,
2837 : GcError::TimelineNotFound => {
2838 0 : ApiError::NotFound(anyhow::anyhow!("Timeline not found").into())
2839 : }
2840 0 : other => ApiError::InternalServerError(anyhow::anyhow!(other)),
2841 0 : })
2842 : }
2843 :
2844 : #[cfg(test)]
2845 : mod tests {
2846 : use std::collections::BTreeMap;
2847 : use std::sync::Arc;
2848 : use tracing::Instrument;
2849 :
2850 : use crate::tenant::mgr::TenantSlot;
2851 :
2852 : use super::{super::harness::TenantHarness, TenantsMap};
2853 :
2854 : #[tokio::test(start_paused = true)]
2855 6 : async fn shutdown_awaits_in_progress_tenant() {
2856 6 : // Test that if an InProgress tenant is in the map during shutdown, the shutdown will gracefully
2857 6 : // wait for it to complete before proceeding.
2858 6 :
2859 6 : let h = TenantHarness::create("shutdown_awaits_in_progress_tenant")
2860 6 : .await
2861 6 : .unwrap();
2862 24 : let (t, _ctx) = h.load().await;
2863 6 :
2864 6 : // harness loads it to active, which is forced and nothing is running on the tenant
2865 6 :
2866 6 : let id = t.tenant_shard_id();
2867 6 :
2868 6 : // tenant harness configures the logging and we cannot escape it
2869 6 : let span = h.span();
2870 6 : let _e = span.enter();
2871 6 :
2872 6 : let tenants = BTreeMap::from([(id, TenantSlot::Attached(t.clone()))]);
2873 6 : let tenants = Arc::new(std::sync::RwLock::new(TenantsMap::Open(tenants)));
2874 6 :
2875 6 : // Invoke remove_tenant_from_memory with a cleanup hook that blocks until we manually
2876 6 : // permit it to proceed: that will stick the tenant in InProgress
2877 6 :
2878 6 : let (until_cleanup_completed, can_complete_cleanup) = utils::completion::channel();
2879 6 : let (until_cleanup_started, cleanup_started) = utils::completion::channel();
2880 6 : let mut remove_tenant_from_memory_task = {
2881 6 : let jh = tokio::spawn({
2882 6 : let tenants = tenants.clone();
2883 6 : async move {
2884 6 : let cleanup = async move {
2885 6 : drop(until_cleanup_started);
2886 6 : can_complete_cleanup.wait().await;
2887 6 : anyhow::Ok(())
2888 6 : };
2889 6 : super::remove_tenant_from_memory(&tenants, id, cleanup).await
2890 6 : }
2891 6 : .instrument(h.span())
2892 6 : });
2893 6 :
2894 6 : // now the long cleanup should be in place, with the stopping state
2895 6 : cleanup_started.wait().await;
2896 6 : jh
2897 6 : };
2898 6 :
2899 6 : let mut shutdown_task = {
2900 6 : let (until_shutdown_started, shutdown_started) = utils::completion::channel();
2901 6 :
2902 6 : let shutdown_task = tokio::spawn(async move {
2903 6 : drop(until_shutdown_started);
2904 12 : super::shutdown_all_tenants0(&tenants).await;
2905 6 : });
2906 6 :
2907 6 : shutdown_started.wait().await;
2908 6 : shutdown_task
2909 6 : };
2910 6 :
2911 6 : let long_time = std::time::Duration::from_secs(15);
2912 6 : tokio::select! {
2913 6 : _ = &mut shutdown_task => unreachable!("shutdown should block on remove_tenant_from_memory completing"),
2914 6 : _ = &mut remove_tenant_from_memory_task => unreachable!("remove_tenant_from_memory_task should not complete until explicitly unblocked"),
2915 6 : _ = tokio::time::sleep(long_time) => {},
2916 6 : }
2917 6 :
2918 6 : drop(until_cleanup_completed);
2919 6 :
2920 6 : // Now that we allow it to proceed, shutdown should complete immediately
2921 6 : remove_tenant_from_memory_task.await.unwrap().unwrap();
2922 6 : shutdown_task.await.unwrap();
2923 6 : }
2924 : }
|
