Line data Source code
1 : //!
2 : //! `neon_local` is an executable that can be used to create a local
3 : //! Neon environment, for testing purposes. The local environment is
4 : //! quite different from the cloud environment with Kubernetes, but it
5 : //! easier to work with locally. The python tests in `test_runner`
6 : //! rely on `neon_local` to set up the environment for each test.
7 : //!
8 : use anyhow::{anyhow, bail, Context, Result};
9 : use clap::Parser;
10 : use compute_api::spec::ComputeMode;
11 : use control_plane::endpoint::ComputeControlPlane;
12 : use control_plane::local_env::{
13 : InitForceMode, LocalEnv, NeonBroker, NeonLocalInitConf, NeonLocalInitPageserverConf,
14 : SafekeeperConf,
15 : };
16 : use control_plane::pageserver::PageServerNode;
17 : use control_plane::safekeeper::SafekeeperNode;
18 : use control_plane::storage_controller::{
19 : NeonStorageControllerStartArgs, NeonStorageControllerStopArgs, StorageController,
20 : };
21 : use control_plane::{broker, local_env};
22 : use nix::fcntl::{flock, FlockArg};
23 : use pageserver_api::config::{
24 : DEFAULT_HTTP_LISTEN_PORT as DEFAULT_PAGESERVER_HTTP_PORT,
25 : DEFAULT_PG_LISTEN_PORT as DEFAULT_PAGESERVER_PG_PORT,
26 : };
27 : use pageserver_api::controller_api::{
28 : NodeAvailabilityWrapper, PlacementPolicy, TenantCreateRequest,
29 : };
30 : use pageserver_api::models::{ShardParameters, TimelineCreateRequest, TimelineInfo};
31 : use pageserver_api::shard::{ShardCount, ShardStripeSize, TenantShardId};
32 : use postgres_backend::AuthType;
33 : use postgres_connection::parse_host_port;
34 : use safekeeper_api::{
35 : DEFAULT_HTTP_LISTEN_PORT as DEFAULT_SAFEKEEPER_HTTP_PORT,
36 : DEFAULT_PG_LISTEN_PORT as DEFAULT_SAFEKEEPER_PG_PORT,
37 : };
38 : use std::borrow::Cow;
39 : use std::collections::{BTreeSet, HashMap};
40 : use std::fs::File;
41 : use std::os::fd::AsRawFd;
42 : use std::path::PathBuf;
43 : use std::process::exit;
44 : use std::str::FromStr;
45 : use std::time::Duration;
46 : use storage_broker::DEFAULT_LISTEN_ADDR as DEFAULT_BROKER_ADDR;
47 : use tokio::task::JoinSet;
48 : use url::Host;
49 : use utils::{
50 : auth::{Claims, Scope},
51 : id::{NodeId, TenantId, TenantTimelineId, TimelineId},
52 : lsn::Lsn,
53 : project_git_version,
54 : };
55 :
56 : // Default id of a safekeeper node, if not specified on the command line.
57 : const DEFAULT_SAFEKEEPER_ID: NodeId = NodeId(1);
58 : const DEFAULT_PAGESERVER_ID: NodeId = NodeId(1);
59 : const DEFAULT_BRANCH_NAME: &str = "main";
60 : project_git_version!(GIT_VERSION);
61 :
62 : const DEFAULT_PG_VERSION: u32 = 16;
63 :
64 : const DEFAULT_PAGESERVER_CONTROL_PLANE_API: &str = "http://127.0.0.1:1234/upcall/v1/";
65 :
66 : #[derive(clap::Parser)]
67 : #[command(version = GIT_VERSION, about, name = "Neon CLI")]
68 : struct Cli {
69 : #[command(subcommand)]
70 : command: NeonLocalCmd,
71 : }
72 :
73 : #[derive(clap::Subcommand)]
74 : enum NeonLocalCmd {
75 : Init(InitCmdArgs),
76 :
77 : #[command(subcommand)]
78 : Tenant(TenantCmd),
79 : #[command(subcommand)]
80 : Timeline(TimelineCmd),
81 : #[command(subcommand)]
82 : Pageserver(PageserverCmd),
83 : #[command(subcommand)]
84 : #[clap(alias = "storage_controller")]
85 : StorageController(StorageControllerCmd),
86 : #[command(subcommand)]
87 : #[clap(alias = "storage_broker")]
88 : StorageBroker(StorageBrokerCmd),
89 : #[command(subcommand)]
90 : Safekeeper(SafekeeperCmd),
91 : #[command(subcommand)]
92 : Endpoint(EndpointCmd),
93 : #[command(subcommand)]
94 : Mappings(MappingsCmd),
95 :
96 : Start(StartCmdArgs),
97 : Stop(StopCmdArgs),
98 : }
99 :
100 : #[derive(clap::Args)]
101 : #[clap(about = "Initialize a new Neon repository, preparing configs for services to start with")]
102 : struct InitCmdArgs {
103 : #[clap(long, help("How many pageservers to create (default 1)"))]
104 : num_pageservers: Option<u16>,
105 :
106 : #[clap(long)]
107 : config: Option<PathBuf>,
108 :
109 : #[clap(long, help("Force initialization even if the repository is not empty"))]
110 : #[arg(value_parser)]
111 : #[clap(default_value = "must-not-exist")]
112 0 : force: InitForceMode,
113 : }
114 :
115 : #[derive(clap::Args)]
116 : #[clap(about = "Start pageserver and safekeepers")]
117 : struct StartCmdArgs {
118 : #[clap(long = "start-timeout", default_value = "10s")]
119 0 : timeout: humantime::Duration,
120 : }
121 :
122 : #[derive(clap::Args)]
123 : #[clap(about = "Stop pageserver and safekeepers")]
124 : struct StopCmdArgs {
125 : #[arg(value_enum)]
126 0 : #[clap(long, default_value_t = StopMode::Fast)]
127 0 : mode: StopMode,
128 : }
129 :
130 : #[derive(Clone, Copy, clap::ValueEnum)]
131 : enum StopMode {
132 : Fast,
133 : Immediate,
134 : }
135 :
136 : #[derive(clap::Subcommand)]
137 : #[clap(about = "Manage tenants")]
138 : enum TenantCmd {
139 : List,
140 : Create(TenantCreateCmdArgs),
141 : SetDefault(TenantSetDefaultCmdArgs),
142 : Config(TenantConfigCmdArgs),
143 : Import(TenantImportCmdArgs),
144 : }
145 :
146 : #[derive(clap::Args)]
147 : struct TenantCreateCmdArgs {
148 : #[clap(
149 : long = "tenant-id",
150 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
151 : )]
152 : tenant_id: Option<TenantId>,
153 :
154 : #[clap(
155 : long,
156 : help = "Use a specific timeline id when creating a tenant and its initial timeline"
157 : )]
158 : timeline_id: Option<TimelineId>,
159 :
160 : #[clap(short = 'c')]
161 0 : config: Vec<String>,
162 :
163 0 : #[arg(default_value_t = DEFAULT_PG_VERSION)]
164 : #[clap(long, help = "Postgres version to use for the initial timeline")]
165 0 : pg_version: u32,
166 :
167 : #[clap(
168 : long,
169 : help = "Use this tenant in future CLI commands where tenant_id is needed, but not specified"
170 : )]
171 0 : set_default: bool,
172 :
173 : #[clap(long, help = "Number of shards in the new tenant")]
174 0 : #[arg(default_value_t = 0)]
175 0 : shard_count: u8,
176 : #[clap(long, help = "Sharding stripe size in pages")]
177 : shard_stripe_size: Option<u32>,
178 :
179 : #[clap(long, help = "Placement policy shards in this tenant")]
180 : #[arg(value_parser = parse_placement_policy)]
181 : placement_policy: Option<PlacementPolicy>,
182 : }
183 :
184 0 : fn parse_placement_policy(s: &str) -> anyhow::Result<PlacementPolicy> {
185 0 : Ok(serde_json::from_str::<PlacementPolicy>(s)?)
186 0 : }
187 :
188 : #[derive(clap::Args)]
189 : #[clap(
190 : about = "Set a particular tenant as default in future CLI commands where tenant_id is needed, but not specified"
191 : )]
192 : struct TenantSetDefaultCmdArgs {
193 : #[clap(
194 : long = "tenant-id",
195 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
196 : )]
197 0 : tenant_id: TenantId,
198 : }
199 :
200 : #[derive(clap::Args)]
201 : struct TenantConfigCmdArgs {
202 : #[clap(
203 : long = "tenant-id",
204 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
205 : )]
206 : tenant_id: Option<TenantId>,
207 :
208 : #[clap(short = 'c')]
209 0 : config: Vec<String>,
210 : }
211 :
212 : #[derive(clap::Args)]
213 : #[clap(
214 : about = "Import a tenant that is present in remote storage, and create branches for its timelines"
215 : )]
216 : struct TenantImportCmdArgs {
217 : #[clap(
218 : long = "tenant-id",
219 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
220 : )]
221 0 : tenant_id: TenantId,
222 : }
223 :
224 : #[derive(clap::Subcommand)]
225 : #[clap(about = "Manage timelines")]
226 : enum TimelineCmd {
227 : List(TimelineListCmdArgs),
228 : Branch(TimelineBranchCmdArgs),
229 : Create(TimelineCreateCmdArgs),
230 : Import(TimelineImportCmdArgs),
231 : }
232 :
233 : #[derive(clap::Args)]
234 : #[clap(about = "List all timelines available to this pageserver")]
235 : struct TimelineListCmdArgs {
236 : #[clap(
237 : long = "tenant-id",
238 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
239 : )]
240 : tenant_shard_id: Option<TenantShardId>,
241 : }
242 :
243 : #[derive(clap::Args)]
244 : #[clap(about = "Create a new timeline, branching off from another timeline")]
245 : struct TimelineBranchCmdArgs {
246 : #[clap(
247 : long = "tenant-id",
248 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
249 : )]
250 : tenant_id: Option<TenantId>,
251 :
252 : #[clap(long, help = "New timeline's ID")]
253 : timeline_id: Option<TimelineId>,
254 :
255 : #[clap(long, help = "Human-readable alias for the new timeline")]
256 0 : branch_name: String,
257 :
258 : #[clap(
259 : long,
260 : help = "Use last Lsn of another timeline (and its data) as base when creating the new timeline. The timeline gets resolved by its branch name."
261 : )]
262 : ancestor_branch_name: Option<String>,
263 :
264 : #[clap(
265 : long,
266 : help = "When using another timeline as base, use a specific Lsn in it instead of the latest one"
267 : )]
268 : ancestor_start_lsn: Option<Lsn>,
269 : }
270 :
271 : #[derive(clap::Args)]
272 : #[clap(about = "Create a new blank timeline")]
273 : struct TimelineCreateCmdArgs {
274 : #[clap(
275 : long = "tenant-id",
276 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
277 : )]
278 : tenant_id: Option<TenantId>,
279 :
280 : #[clap(long, help = "New timeline's ID")]
281 : timeline_id: Option<TimelineId>,
282 :
283 : #[clap(long, help = "Human-readable alias for the new timeline")]
284 0 : branch_name: String,
285 :
286 0 : #[arg(default_value_t = DEFAULT_PG_VERSION)]
287 : #[clap(long, help = "Postgres version")]
288 0 : pg_version: u32,
289 : }
290 :
291 : #[derive(clap::Args)]
292 : #[clap(about = "Import timeline from a basebackup directory")]
293 : struct TimelineImportCmdArgs {
294 : #[clap(
295 : long = "tenant-id",
296 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
297 : )]
298 : tenant_id: Option<TenantId>,
299 :
300 : #[clap(long, help = "New timeline's ID")]
301 0 : timeline_id: TimelineId,
302 :
303 : #[clap(long, help = "Human-readable alias for the new timeline")]
304 0 : branch_name: String,
305 :
306 : #[clap(long, help = "Basebackup tarfile to import")]
307 0 : base_tarfile: PathBuf,
308 :
309 : #[clap(long, help = "Lsn the basebackup starts at")]
310 0 : base_lsn: Lsn,
311 :
312 : #[clap(long, help = "Wal to add after base")]
313 : wal_tarfile: Option<PathBuf>,
314 :
315 : #[clap(long, help = "Lsn the basebackup ends at")]
316 : end_lsn: Option<Lsn>,
317 :
318 0 : #[arg(default_value_t = DEFAULT_PG_VERSION)]
319 : #[clap(long, help = "Postgres version of the backup being imported")]
320 0 : pg_version: u32,
321 : }
322 :
323 : #[derive(clap::Subcommand)]
324 : #[clap(about = "Manage pageservers")]
325 : enum PageserverCmd {
326 : Status(PageserverStatusCmdArgs),
327 : Start(PageserverStartCmdArgs),
328 : Stop(PageserverStopCmdArgs),
329 : Restart(PageserverRestartCmdArgs),
330 : }
331 :
332 : #[derive(clap::Args)]
333 : #[clap(about = "Show status of a local pageserver")]
334 : struct PageserverStatusCmdArgs {
335 : #[clap(long = "id", help = "pageserver id")]
336 : pageserver_id: Option<NodeId>,
337 : }
338 :
339 : #[derive(clap::Args)]
340 : #[clap(about = "Start local pageserver")]
341 : struct PageserverStartCmdArgs {
342 : #[clap(long = "id", help = "pageserver id")]
343 : pageserver_id: Option<NodeId>,
344 :
345 : #[clap(short = 't', long, help = "timeout until we fail the command")]
346 : #[arg(default_value = "10s")]
347 0 : start_timeout: humantime::Duration,
348 : }
349 :
350 : #[derive(clap::Args)]
351 : #[clap(about = "Stop local pageserver")]
352 : struct PageserverStopCmdArgs {
353 : #[clap(long = "id", help = "pageserver id")]
354 : pageserver_id: Option<NodeId>,
355 :
356 : #[clap(
357 : short = 'm',
358 : help = "If 'immediate', don't flush repository data at shutdown"
359 : )]
360 : #[arg(value_enum, default_value = "fast")]
361 0 : stop_mode: StopMode,
362 : }
363 :
364 : #[derive(clap::Args)]
365 : #[clap(about = "Restart local pageserver")]
366 : struct PageserverRestartCmdArgs {
367 : #[clap(long = "id", help = "pageserver id")]
368 : pageserver_id: Option<NodeId>,
369 :
370 : #[clap(short = 't', long, help = "timeout until we fail the command")]
371 : #[arg(default_value = "10s")]
372 0 : start_timeout: humantime::Duration,
373 : }
374 :
375 : #[derive(clap::Subcommand)]
376 : #[clap(about = "Manage storage controller")]
377 : enum StorageControllerCmd {
378 : Start(StorageControllerStartCmdArgs),
379 : Stop(StorageControllerStopCmdArgs),
380 : }
381 :
382 : #[derive(clap::Args)]
383 : #[clap(about = "Start storage controller")]
384 : struct StorageControllerStartCmdArgs {
385 : #[clap(short = 't', long, help = "timeout until we fail the command")]
386 : #[arg(default_value = "10s")]
387 0 : start_timeout: humantime::Duration,
388 :
389 : #[clap(
390 : long,
391 : help = "Identifier used to distinguish storage controller instances"
392 : )]
393 0 : #[arg(default_value_t = 1)]
394 0 : instance_id: u8,
395 :
396 : #[clap(
397 : long,
398 : help = "Base port for the storage controller instance idenfified by instance-id (defaults to pageserver cplane api)"
399 : )]
400 : base_port: Option<u16>,
401 : }
402 :
403 : #[derive(clap::Args)]
404 : #[clap(about = "Stop storage controller")]
405 : struct StorageControllerStopCmdArgs {
406 : #[clap(
407 : short = 'm',
408 : help = "If 'immediate', don't flush repository data at shutdown"
409 : )]
410 : #[arg(value_enum, default_value = "fast")]
411 0 : stop_mode: StopMode,
412 :
413 : #[clap(
414 : long,
415 : help = "Identifier used to distinguish storage controller instances"
416 : )]
417 0 : #[arg(default_value_t = 1)]
418 0 : instance_id: u8,
419 : }
420 :
421 : #[derive(clap::Subcommand)]
422 : #[clap(about = "Manage storage broker")]
423 : enum StorageBrokerCmd {
424 : Start(StorageBrokerStartCmdArgs),
425 : Stop(StorageBrokerStopCmdArgs),
426 : }
427 :
428 : #[derive(clap::Args)]
429 : #[clap(about = "Start broker")]
430 : struct StorageBrokerStartCmdArgs {
431 : #[clap(short = 't', long, help = "timeout until we fail the command")]
432 : #[arg(default_value = "10s")]
433 0 : start_timeout: humantime::Duration,
434 : }
435 :
436 : #[derive(clap::Args)]
437 : #[clap(about = "stop broker")]
438 : struct StorageBrokerStopCmdArgs {
439 : #[clap(
440 : short = 'm',
441 : help = "If 'immediate', don't flush repository data at shutdown"
442 : )]
443 : #[arg(value_enum, default_value = "fast")]
444 0 : stop_mode: StopMode,
445 : }
446 :
447 : #[derive(clap::Subcommand)]
448 : #[clap(about = "Manage safekeepers")]
449 : enum SafekeeperCmd {
450 : Start(SafekeeperStartCmdArgs),
451 : Stop(SafekeeperStopCmdArgs),
452 : Restart(SafekeeperRestartCmdArgs),
453 : }
454 :
455 : #[derive(clap::Args)]
456 : #[clap(about = "Start local safekeeper")]
457 : struct SafekeeperStartCmdArgs {
458 : #[clap(help = "safekeeper id")]
459 0 : #[arg(default_value_t = NodeId(1))]
460 0 : id: NodeId,
461 :
462 : #[clap(
463 : short = 'e',
464 : long = "safekeeper-extra-opt",
465 : help = "Additional safekeeper invocation options, e.g. -e=--http-auth-public-key-path=foo"
466 : )]
467 0 : extra_opt: Vec<String>,
468 :
469 : #[clap(short = 't', long, help = "timeout until we fail the command")]
470 : #[arg(default_value = "10s")]
471 0 : start_timeout: humantime::Duration,
472 : }
473 :
474 : #[derive(clap::Args)]
475 : #[clap(about = "Stop local safekeeper")]
476 : struct SafekeeperStopCmdArgs {
477 : #[clap(help = "safekeeper id")]
478 0 : #[arg(default_value_t = NodeId(1))]
479 0 : id: NodeId,
480 :
481 : #[arg(value_enum, default_value = "fast")]
482 : #[clap(
483 : short = 'm',
484 : help = "If 'immediate', don't flush repository data at shutdown"
485 : )]
486 0 : stop_mode: StopMode,
487 : }
488 :
489 : #[derive(clap::Args)]
490 : #[clap(about = "Restart local safekeeper")]
491 : struct SafekeeperRestartCmdArgs {
492 : #[clap(help = "safekeeper id")]
493 0 : #[arg(default_value_t = NodeId(1))]
494 0 : id: NodeId,
495 :
496 : #[arg(value_enum, default_value = "fast")]
497 : #[clap(
498 : short = 'm',
499 : help = "If 'immediate', don't flush repository data at shutdown"
500 : )]
501 0 : stop_mode: StopMode,
502 :
503 : #[clap(
504 : short = 'e',
505 : long = "safekeeper-extra-opt",
506 : help = "Additional safekeeper invocation options, e.g. -e=--http-auth-public-key-path=foo"
507 : )]
508 0 : extra_opt: Vec<String>,
509 :
510 : #[clap(short = 't', long, help = "timeout until we fail the command")]
511 : #[arg(default_value = "10s")]
512 0 : start_timeout: humantime::Duration,
513 : }
514 :
515 : #[derive(clap::Subcommand)]
516 : #[clap(about = "Manage Postgres instances")]
517 : enum EndpointCmd {
518 : List(EndpointListCmdArgs),
519 : Create(EndpointCreateCmdArgs),
520 : Start(EndpointStartCmdArgs),
521 : Reconfigure(EndpointReconfigureCmdArgs),
522 : Stop(EndpointStopCmdArgs),
523 : }
524 :
525 : #[derive(clap::Args)]
526 : #[clap(about = "List endpoints")]
527 : struct EndpointListCmdArgs {
528 : #[clap(
529 : long = "tenant-id",
530 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
531 : )]
532 : tenant_shard_id: Option<TenantShardId>,
533 : }
534 :
535 : #[derive(clap::Args)]
536 : #[clap(about = "Create a compute endpoint")]
537 : struct EndpointCreateCmdArgs {
538 : #[clap(
539 : long = "tenant-id",
540 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
541 : )]
542 : tenant_id: Option<TenantId>,
543 :
544 : #[clap(help = "Postgres endpoint id")]
545 : endpoint_id: Option<String>,
546 : #[clap(long, help = "Name of the branch the endpoint will run on")]
547 : branch_name: Option<String>,
548 : #[clap(
549 : long,
550 : help = "Specify Lsn on the timeline to start from. By default, end of the timeline would be used"
551 : )]
552 : lsn: Option<Lsn>,
553 : #[clap(long)]
554 : pg_port: Option<u16>,
555 : #[clap(long, alias = "http-port")]
556 : external_http_port: Option<u16>,
557 : #[clap(long)]
558 : internal_http_port: Option<u16>,
559 : #[clap(long = "pageserver-id")]
560 : endpoint_pageserver_id: Option<NodeId>,
561 :
562 : #[clap(
563 : long,
564 : help = "Don't do basebackup, create endpoint directory with only config files",
565 : action = clap::ArgAction::Set,
566 0 : default_value_t = false
567 : )]
568 0 : config_only: bool,
569 :
570 0 : #[arg(default_value_t = DEFAULT_PG_VERSION)]
571 : #[clap(long, help = "Postgres version")]
572 0 : pg_version: u32,
573 :
574 : #[clap(
575 : long,
576 : help = "If set, the node will be a hot replica on the specified timeline",
577 : action = clap::ArgAction::Set,
578 0 : default_value_t = false
579 : )]
580 0 : hot_standby: bool,
581 :
582 : #[clap(long, help = "If set, will set up the catalog for neon_superuser")]
583 0 : update_catalog: bool,
584 :
585 : #[clap(
586 : long,
587 : help = "Allow multiple primary endpoints running on the same branch. Shouldn't be used normally, but useful for tests."
588 : )]
589 0 : allow_multiple: bool,
590 : }
591 :
592 : #[derive(clap::Args)]
593 : #[clap(about = "Start postgres. If the endpoint doesn't exist yet, it is created.")]
594 : struct EndpointStartCmdArgs {
595 : #[clap(help = "Postgres endpoint id")]
596 0 : endpoint_id: String,
597 : #[clap(long = "pageserver-id")]
598 : endpoint_pageserver_id: Option<NodeId>,
599 :
600 : #[clap(long)]
601 : safekeepers: Option<String>,
602 :
603 : #[clap(
604 : long,
605 : help = "Configure the remote extensions storage proxy gateway to request for extensions."
606 : )]
607 : remote_ext_config: Option<String>,
608 :
609 : #[clap(
610 : long,
611 : help = "If set, will create test user `user` and `neondb` database. Requires `update-catalog = true`"
612 : )]
613 0 : create_test_user: bool,
614 :
615 : #[clap(
616 : long,
617 : help = "Allow multiple primary endpoints running on the same branch. Shouldn't be used normally, but useful for tests."
618 : )]
619 0 : allow_multiple: bool,
620 :
621 : #[clap(short = 't', long, help = "timeout until we fail the command")]
622 : #[arg(default_value = "10s")]
623 0 : start_timeout: humantime::Duration,
624 : }
625 :
626 : #[derive(clap::Args)]
627 : #[clap(about = "Reconfigure an endpoint")]
628 : struct EndpointReconfigureCmdArgs {
629 : #[clap(
630 : long = "tenant-id",
631 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
632 : )]
633 : tenant_id: Option<TenantId>,
634 :
635 : #[clap(help = "Postgres endpoint id")]
636 0 : endpoint_id: String,
637 : #[clap(long = "pageserver-id")]
638 : endpoint_pageserver_id: Option<NodeId>,
639 :
640 : #[clap(long)]
641 : safekeepers: Option<String>,
642 : }
643 :
644 : #[derive(clap::Args)]
645 : #[clap(about = "Stop an endpoint")]
646 : struct EndpointStopCmdArgs {
647 : #[clap(help = "Postgres endpoint id")]
648 0 : endpoint_id: String,
649 :
650 : #[clap(
651 : long,
652 : help = "Also delete data directory (now optional, should be default in future)"
653 : )]
654 0 : destroy: bool,
655 :
656 : #[clap(long, help = "Postgres shutdown mode, passed to \"pg_ctl -m <mode>\"")]
657 : #[arg(value_parser(["smart", "fast", "immediate"]))]
658 : #[arg(default_value = "fast")]
659 0 : mode: String,
660 : }
661 :
662 : #[derive(clap::Subcommand)]
663 : #[clap(about = "Manage neon_local branch name mappings")]
664 : enum MappingsCmd {
665 : Map(MappingsMapCmdArgs),
666 : }
667 :
668 : #[derive(clap::Args)]
669 : #[clap(about = "Create new mapping which cannot exist already")]
670 : struct MappingsMapCmdArgs {
671 : #[clap(
672 : long,
673 : help = "Tenant id. Represented as a hexadecimal string 32 symbols length"
674 : )]
675 0 : tenant_id: TenantId,
676 : #[clap(
677 : long,
678 : help = "Timeline id. Represented as a hexadecimal string 32 symbols length"
679 : )]
680 0 : timeline_id: TimelineId,
681 : #[clap(long, help = "Branch name to give to the timeline")]
682 0 : branch_name: String,
683 : }
684 :
685 : ///
686 : /// Timelines tree element used as a value in the HashMap.
687 : ///
688 : struct TimelineTreeEl {
689 : /// `TimelineInfo` received from the `pageserver` via the `timeline_list` http API call.
690 : pub info: TimelineInfo,
691 : /// Name, recovered from neon config mappings
692 : pub name: Option<String>,
693 : /// Holds all direct children of this timeline referenced using `timeline_id`.
694 : pub children: BTreeSet<TimelineId>,
695 : }
696 :
697 : /// A flock-based guard over the neon_local repository directory
698 : struct RepoLock {
699 : _file: File,
700 : }
701 :
702 : impl RepoLock {
703 0 : fn new() -> Result<Self> {
704 0 : let repo_dir = File::open(local_env::base_path())?;
705 0 : let repo_dir_fd = repo_dir.as_raw_fd();
706 0 : flock(repo_dir_fd, FlockArg::LockExclusive)?;
707 :
708 0 : Ok(Self { _file: repo_dir })
709 0 : }
710 : }
711 :
712 : // Main entry point for the 'neon_local' CLI utility
713 : //
714 : // This utility helps to manage neon installation. That includes following:
715 : // * Management of local postgres installations running on top of the
716 : // pageserver.
717 : // * Providing CLI api to the pageserver
718 : // * TODO: export/import to/from usual postgres
719 0 : fn main() -> Result<()> {
720 0 : let cli = Cli::parse();
721 :
722 : // Check for 'neon init' command first.
723 0 : let (subcommand_result, _lock) = if let NeonLocalCmd::Init(args) = cli.command {
724 0 : (handle_init(&args).map(|env| Some(Cow::Owned(env))), None)
725 : } else {
726 : // This tool uses a collection of simple files to store its state, and consequently
727 : // it is not generally safe to run multiple commands concurrently. Rather than expect
728 : // all callers to know this, use a lock file to protect against concurrent execution.
729 0 : let _repo_lock = RepoLock::new().unwrap();
730 :
731 : // all other commands need an existing config
732 0 : let env = LocalEnv::load_config(&local_env::base_path()).context("Error loading config")?;
733 0 : let original_env = env.clone();
734 0 : let env = Box::leak(Box::new(env));
735 0 : let rt = tokio::runtime::Builder::new_current_thread()
736 0 : .enable_all()
737 0 : .build()
738 0 : .unwrap();
739 :
740 0 : let subcommand_result = match cli.command {
741 0 : NeonLocalCmd::Init(_) => unreachable!("init was handled earlier already"),
742 0 : NeonLocalCmd::Start(args) => rt.block_on(handle_start_all(&args, env)),
743 0 : NeonLocalCmd::Stop(args) => rt.block_on(handle_stop_all(&args, env)),
744 0 : NeonLocalCmd::Tenant(subcmd) => rt.block_on(handle_tenant(&subcmd, env)),
745 0 : NeonLocalCmd::Timeline(subcmd) => rt.block_on(handle_timeline(&subcmd, env)),
746 0 : NeonLocalCmd::Pageserver(subcmd) => rt.block_on(handle_pageserver(&subcmd, env)),
747 0 : NeonLocalCmd::StorageController(subcmd) => {
748 0 : rt.block_on(handle_storage_controller(&subcmd, env))
749 : }
750 0 : NeonLocalCmd::StorageBroker(subcmd) => rt.block_on(handle_storage_broker(&subcmd, env)),
751 0 : NeonLocalCmd::Safekeeper(subcmd) => rt.block_on(handle_safekeeper(&subcmd, env)),
752 0 : NeonLocalCmd::Endpoint(subcmd) => rt.block_on(handle_endpoint(&subcmd, env)),
753 0 : NeonLocalCmd::Mappings(subcmd) => handle_mappings(&subcmd, env),
754 : };
755 :
756 0 : let subcommand_result = if &original_env != env {
757 0 : subcommand_result.map(|()| Some(Cow::Borrowed(env)))
758 : } else {
759 0 : subcommand_result.map(|()| None)
760 : };
761 0 : (subcommand_result, Some(_repo_lock))
762 : };
763 :
764 0 : match subcommand_result {
765 0 : Ok(Some(updated_env)) => updated_env.persist_config()?,
766 0 : Ok(None) => (),
767 0 : Err(e) => {
768 0 : eprintln!("command failed: {e:?}");
769 0 : exit(1);
770 : }
771 : }
772 0 : Ok(())
773 0 : }
774 :
775 : ///
776 : /// Prints timelines list as a tree-like structure.
777 : ///
778 0 : fn print_timelines_tree(
779 0 : timelines: Vec<TimelineInfo>,
780 0 : mut timeline_name_mappings: HashMap<TenantTimelineId, String>,
781 0 : ) -> Result<()> {
782 0 : let mut timelines_hash = timelines
783 0 : .iter()
784 0 : .map(|t| {
785 0 : (
786 0 : t.timeline_id,
787 0 : TimelineTreeEl {
788 0 : info: t.clone(),
789 0 : children: BTreeSet::new(),
790 0 : name: timeline_name_mappings
791 0 : .remove(&TenantTimelineId::new(t.tenant_id.tenant_id, t.timeline_id)),
792 0 : },
793 0 : )
794 0 : })
795 0 : .collect::<HashMap<_, _>>();
796 :
797 : // Memorize all direct children of each timeline.
798 0 : for timeline in timelines.iter() {
799 0 : if let Some(ancestor_timeline_id) = timeline.ancestor_timeline_id {
800 0 : timelines_hash
801 0 : .get_mut(&ancestor_timeline_id)
802 0 : .context("missing timeline info in the HashMap")?
803 : .children
804 0 : .insert(timeline.timeline_id);
805 0 : }
806 : }
807 :
808 0 : for timeline in timelines_hash.values() {
809 : // Start with root local timelines (no ancestors) first.
810 0 : if timeline.info.ancestor_timeline_id.is_none() {
811 0 : print_timeline(0, &Vec::from([true]), timeline, &timelines_hash)?;
812 0 : }
813 : }
814 :
815 0 : Ok(())
816 0 : }
817 :
818 : ///
819 : /// Recursively prints timeline info with all its children.
820 : ///
821 0 : fn print_timeline(
822 0 : nesting_level: usize,
823 0 : is_last: &[bool],
824 0 : timeline: &TimelineTreeEl,
825 0 : timelines: &HashMap<TimelineId, TimelineTreeEl>,
826 0 : ) -> Result<()> {
827 0 : if nesting_level > 0 {
828 0 : let ancestor_lsn = match timeline.info.ancestor_lsn {
829 0 : Some(lsn) => lsn.to_string(),
830 0 : None => "Unknown Lsn".to_string(),
831 : };
832 :
833 0 : let mut br_sym = "┣━";
834 0 :
835 0 : // Draw each nesting padding with proper style
836 0 : // depending on whether its timeline ended or not.
837 0 : if nesting_level > 1 {
838 0 : for l in &is_last[1..is_last.len() - 1] {
839 0 : if *l {
840 0 : print!(" ");
841 0 : } else {
842 0 : print!("┃ ");
843 0 : }
844 : }
845 0 : }
846 :
847 : // We are the last in this sub-timeline
848 0 : if *is_last.last().unwrap() {
849 0 : br_sym = "┗━";
850 0 : }
851 :
852 0 : print!("{} @{}: ", br_sym, ancestor_lsn);
853 0 : }
854 :
855 : // Finally print a timeline id and name with new line
856 0 : println!(
857 0 : "{} [{}]",
858 0 : timeline.name.as_deref().unwrap_or("_no_name_"),
859 0 : timeline.info.timeline_id
860 0 : );
861 0 :
862 0 : let len = timeline.children.len();
863 0 : let mut i: usize = 0;
864 0 : let mut is_last_new = Vec::from(is_last);
865 0 : is_last_new.push(false);
866 :
867 0 : for child in &timeline.children {
868 0 : i += 1;
869 0 :
870 0 : // Mark that the last padding is the end of the timeline
871 0 : if i == len {
872 0 : if let Some(last) = is_last_new.last_mut() {
873 0 : *last = true;
874 0 : }
875 0 : }
876 :
877 : print_timeline(
878 0 : nesting_level + 1,
879 0 : &is_last_new,
880 0 : timelines
881 0 : .get(child)
882 0 : .context("missing timeline info in the HashMap")?,
883 0 : timelines,
884 0 : )?;
885 : }
886 :
887 0 : Ok(())
888 0 : }
889 :
890 : /// Returns a map of timeline IDs to timeline_id@lsn strings.
891 : /// Connects to the pageserver to query this information.
892 0 : async fn get_timeline_infos(
893 0 : env: &local_env::LocalEnv,
894 0 : tenant_shard_id: &TenantShardId,
895 0 : ) -> Result<HashMap<TimelineId, TimelineInfo>> {
896 0 : Ok(get_default_pageserver(env)
897 0 : .timeline_list(tenant_shard_id)
898 0 : .await?
899 0 : .into_iter()
900 0 : .map(|timeline_info| (timeline_info.timeline_id, timeline_info))
901 0 : .collect())
902 0 : }
903 :
904 : /// Helper function to get tenant id from an optional --tenant_id option or from the config file
905 0 : fn get_tenant_id(
906 0 : tenant_id_arg: Option<TenantId>,
907 0 : env: &local_env::LocalEnv,
908 0 : ) -> anyhow::Result<TenantId> {
909 0 : if let Some(tenant_id_from_arguments) = tenant_id_arg {
910 0 : Ok(tenant_id_from_arguments)
911 0 : } else if let Some(default_id) = env.default_tenant_id {
912 0 : Ok(default_id)
913 : } else {
914 0 : anyhow::bail!("No tenant id. Use --tenant-id, or set a default tenant");
915 : }
916 0 : }
917 :
918 : /// Helper function to get tenant-shard ID from an optional --tenant_id option or from the config file,
919 : /// for commands that accept a shard suffix
920 0 : fn get_tenant_shard_id(
921 0 : tenant_shard_id_arg: Option<TenantShardId>,
922 0 : env: &local_env::LocalEnv,
923 0 : ) -> anyhow::Result<TenantShardId> {
924 0 : if let Some(tenant_id_from_arguments) = tenant_shard_id_arg {
925 0 : Ok(tenant_id_from_arguments)
926 0 : } else if let Some(default_id) = env.default_tenant_id {
927 0 : Ok(TenantShardId::unsharded(default_id))
928 : } else {
929 0 : anyhow::bail!("No tenant shard id. Use --tenant-id, or set a default tenant");
930 : }
931 0 : }
932 :
933 0 : fn handle_init(args: &InitCmdArgs) -> anyhow::Result<LocalEnv> {
934 : // Create the in-memory `LocalEnv` that we'd normally load from disk in `load_config`.
935 0 : let init_conf: NeonLocalInitConf = if let Some(config_path) = &args.config {
936 : // User (likely the Python test suite) provided a description of the environment.
937 0 : if args.num_pageservers.is_some() {
938 0 : bail!("Cannot specify both --num-pageservers and --config, use key `pageservers` in the --config file instead");
939 0 : }
940 : // load and parse the file
941 0 : let contents = std::fs::read_to_string(config_path).with_context(|| {
942 0 : format!(
943 0 : "Could not read configuration file '{}'",
944 0 : config_path.display()
945 0 : )
946 0 : })?;
947 0 : toml_edit::de::from_str(&contents)?
948 : } else {
949 : // User (likely interactive) did not provide a description of the environment, give them the default
950 0 : NeonLocalInitConf {
951 0 : control_plane_api: Some(DEFAULT_PAGESERVER_CONTROL_PLANE_API.parse().unwrap()),
952 0 : broker: NeonBroker {
953 0 : listen_addr: DEFAULT_BROKER_ADDR.parse().unwrap(),
954 0 : },
955 0 : safekeepers: vec![SafekeeperConf {
956 0 : id: DEFAULT_SAFEKEEPER_ID,
957 0 : pg_port: DEFAULT_SAFEKEEPER_PG_PORT,
958 0 : http_port: DEFAULT_SAFEKEEPER_HTTP_PORT,
959 0 : ..Default::default()
960 0 : }],
961 0 : pageservers: (0..args.num_pageservers.unwrap_or(1))
962 0 : .map(|i| {
963 0 : let pageserver_id = NodeId(DEFAULT_PAGESERVER_ID.0 + i as u64);
964 0 : let pg_port = DEFAULT_PAGESERVER_PG_PORT + i;
965 0 : let http_port = DEFAULT_PAGESERVER_HTTP_PORT + i;
966 0 : NeonLocalInitPageserverConf {
967 0 : id: pageserver_id,
968 0 : listen_pg_addr: format!("127.0.0.1:{pg_port}"),
969 0 : listen_http_addr: format!("127.0.0.1:{http_port}"),
970 0 : pg_auth_type: AuthType::Trust,
971 0 : http_auth_type: AuthType::Trust,
972 0 : other: Default::default(),
973 0 : // Typical developer machines use disks with slow fsync, and we don't care
974 0 : // about data integrity: disable disk syncs.
975 0 : no_sync: true,
976 0 : }
977 0 : })
978 0 : .collect(),
979 0 : pg_distrib_dir: None,
980 0 : neon_distrib_dir: None,
981 0 : default_tenant_id: TenantId::from_array(std::array::from_fn(|_| 0)),
982 0 : storage_controller: None,
983 0 : control_plane_compute_hook_api: None,
984 0 : }
985 : };
986 :
987 0 : LocalEnv::init(init_conf, &args.force)
988 0 : .context("materialize initial neon_local environment on disk")?;
989 0 : Ok(LocalEnv::load_config(&local_env::base_path())
990 0 : .expect("freshly written config should be loadable"))
991 0 : }
992 :
993 : /// The default pageserver is the one where CLI tenant/timeline operations are sent by default.
994 : /// For typical interactive use, one would just run with a single pageserver. Scenarios with
995 : /// tenant/timeline placement across multiple pageservers are managed by python test code rather
996 : /// than this CLI.
997 0 : fn get_default_pageserver(env: &local_env::LocalEnv) -> PageServerNode {
998 0 : let ps_conf = env
999 0 : .pageservers
1000 0 : .first()
1001 0 : .expect("Config is validated to contain at least one pageserver");
1002 0 : PageServerNode::from_env(env, ps_conf)
1003 0 : }
1004 :
1005 0 : async fn handle_tenant(subcmd: &TenantCmd, env: &mut local_env::LocalEnv) -> anyhow::Result<()> {
1006 0 : let pageserver = get_default_pageserver(env);
1007 0 : match subcmd {
1008 : TenantCmd::List => {
1009 0 : for t in pageserver.tenant_list().await? {
1010 0 : println!("{} {:?}", t.id, t.state);
1011 0 : }
1012 : }
1013 0 : TenantCmd::Import(args) => {
1014 0 : let tenant_id = args.tenant_id;
1015 0 :
1016 0 : let storage_controller = StorageController::from_env(env);
1017 0 : let create_response = storage_controller.tenant_import(tenant_id).await?;
1018 :
1019 0 : let shard_zero = create_response
1020 0 : .shards
1021 0 : .first()
1022 0 : .expect("Import response omitted shards");
1023 0 :
1024 0 : let attached_pageserver_id = shard_zero.node_id;
1025 0 : let pageserver =
1026 0 : PageServerNode::from_env(env, env.get_pageserver_conf(attached_pageserver_id)?);
1027 :
1028 0 : println!(
1029 0 : "Imported tenant {tenant_id}, attached to pageserver {attached_pageserver_id}"
1030 0 : );
1031 :
1032 0 : let timelines = pageserver
1033 0 : .http_client
1034 0 : .list_timelines(shard_zero.shard_id)
1035 0 : .await?;
1036 :
1037 : // Pick a 'main' timeline that has no ancestors, the rest will get arbitrary names
1038 0 : let main_timeline = timelines
1039 0 : .iter()
1040 0 : .find(|t| t.ancestor_timeline_id.is_none())
1041 0 : .expect("No timelines found")
1042 0 : .timeline_id;
1043 0 :
1044 0 : let mut branch_i = 0;
1045 0 : for timeline in timelines.iter() {
1046 0 : let branch_name = if timeline.timeline_id == main_timeline {
1047 0 : "main".to_string()
1048 : } else {
1049 0 : branch_i += 1;
1050 0 : format!("branch_{branch_i}")
1051 : };
1052 :
1053 0 : println!(
1054 0 : "Importing timeline {tenant_id}/{} as branch {branch_name}",
1055 0 : timeline.timeline_id
1056 0 : );
1057 0 :
1058 0 : env.register_branch_mapping(branch_name, tenant_id, timeline.timeline_id)?;
1059 : }
1060 : }
1061 0 : TenantCmd::Create(args) => {
1062 0 : let tenant_conf: HashMap<_, _> =
1063 0 : args.config.iter().flat_map(|c| c.split_once(':')).collect();
1064 :
1065 0 : let tenant_conf = PageServerNode::parse_config(tenant_conf)?;
1066 :
1067 : // If tenant ID was not specified, generate one
1068 0 : let tenant_id = args.tenant_id.unwrap_or_else(TenantId::generate);
1069 0 :
1070 0 : // We must register the tenant with the storage controller, so
1071 0 : // that when the pageserver restarts, it will be re-attached.
1072 0 : let storage_controller = StorageController::from_env(env);
1073 0 : storage_controller
1074 0 : .tenant_create(TenantCreateRequest {
1075 0 : // Note that ::unsharded here isn't actually because the tenant is unsharded, its because the
1076 0 : // storage controller expects a shard-naive tenant_id in this attribute, and the TenantCreateRequest
1077 0 : // type is used both in the storage controller (for creating tenants) and in the pageserver (for
1078 0 : // creating shards)
1079 0 : new_tenant_id: TenantShardId::unsharded(tenant_id),
1080 0 : generation: None,
1081 0 : shard_parameters: ShardParameters {
1082 0 : count: ShardCount::new(args.shard_count),
1083 0 : stripe_size: args
1084 0 : .shard_stripe_size
1085 0 : .map(ShardStripeSize)
1086 0 : .unwrap_or(ShardParameters::DEFAULT_STRIPE_SIZE),
1087 0 : },
1088 0 : placement_policy: args.placement_policy.clone(),
1089 0 : config: tenant_conf,
1090 0 : })
1091 0 : .await?;
1092 0 : println!("tenant {tenant_id} successfully created on the pageserver");
1093 0 :
1094 0 : // Create an initial timeline for the new tenant
1095 0 : let new_timeline_id = args.timeline_id.unwrap_or(TimelineId::generate());
1096 0 :
1097 0 : // FIXME: passing None for ancestor_start_lsn is not kosher in a sharded world: we can't have
1098 0 : // different shards picking different start lsns. Maybe we have to teach storage controller
1099 0 : // to let shard 0 branch first and then propagate the chosen LSN to other shards.
1100 0 : storage_controller
1101 0 : .tenant_timeline_create(
1102 0 : tenant_id,
1103 0 : TimelineCreateRequest {
1104 0 : new_timeline_id,
1105 0 : mode: pageserver_api::models::TimelineCreateRequestMode::Bootstrap {
1106 0 : existing_initdb_timeline_id: None,
1107 0 : pg_version: Some(args.pg_version),
1108 0 : },
1109 0 : },
1110 0 : )
1111 0 : .await?;
1112 :
1113 0 : env.register_branch_mapping(
1114 0 : DEFAULT_BRANCH_NAME.to_string(),
1115 0 : tenant_id,
1116 0 : new_timeline_id,
1117 0 : )?;
1118 :
1119 0 : println!("Created an initial timeline '{new_timeline_id}' for tenant: {tenant_id}",);
1120 0 :
1121 0 : if args.set_default {
1122 0 : println!("Setting tenant {tenant_id} as a default one");
1123 0 : env.default_tenant_id = Some(tenant_id);
1124 0 : }
1125 : }
1126 0 : TenantCmd::SetDefault(args) => {
1127 0 : println!("Setting tenant {} as a default one", args.tenant_id);
1128 0 : env.default_tenant_id = Some(args.tenant_id);
1129 0 : }
1130 0 : TenantCmd::Config(args) => {
1131 0 : let tenant_id = get_tenant_id(args.tenant_id, env)?;
1132 0 : let tenant_conf: HashMap<_, _> =
1133 0 : args.config.iter().flat_map(|c| c.split_once(':')).collect();
1134 0 :
1135 0 : pageserver
1136 0 : .tenant_config(tenant_id, tenant_conf)
1137 0 : .await
1138 0 : .with_context(|| format!("Tenant config failed for tenant with id {tenant_id}"))?;
1139 0 : println!("tenant {tenant_id} successfully configured on the pageserver");
1140 : }
1141 : }
1142 0 : Ok(())
1143 0 : }
1144 :
1145 0 : async fn handle_timeline(cmd: &TimelineCmd, env: &mut local_env::LocalEnv) -> Result<()> {
1146 0 : let pageserver = get_default_pageserver(env);
1147 0 :
1148 0 : match cmd {
1149 0 : TimelineCmd::List(args) => {
1150 : // TODO(sharding): this command shouldn't have to specify a shard ID: we should ask the storage controller
1151 : // where shard 0 is attached, and query there.
1152 0 : let tenant_shard_id = get_tenant_shard_id(args.tenant_shard_id, env)?;
1153 0 : let timelines = pageserver.timeline_list(&tenant_shard_id).await?;
1154 0 : print_timelines_tree(timelines, env.timeline_name_mappings())?;
1155 : }
1156 0 : TimelineCmd::Create(args) => {
1157 0 : let tenant_id = get_tenant_id(args.tenant_id, env)?;
1158 0 : let new_branch_name = &args.branch_name;
1159 0 : let new_timeline_id_opt = args.timeline_id;
1160 0 : let new_timeline_id = new_timeline_id_opt.unwrap_or(TimelineId::generate());
1161 0 :
1162 0 : let storage_controller = StorageController::from_env(env);
1163 0 : let create_req = TimelineCreateRequest {
1164 0 : new_timeline_id,
1165 0 : mode: pageserver_api::models::TimelineCreateRequestMode::Bootstrap {
1166 0 : existing_initdb_timeline_id: None,
1167 0 : pg_version: Some(args.pg_version),
1168 0 : },
1169 0 : };
1170 0 : let timeline_info = storage_controller
1171 0 : .tenant_timeline_create(tenant_id, create_req)
1172 0 : .await?;
1173 :
1174 0 : let last_record_lsn = timeline_info.last_record_lsn;
1175 0 : env.register_branch_mapping(new_branch_name.to_string(), tenant_id, new_timeline_id)?;
1176 :
1177 0 : println!(
1178 0 : "Created timeline '{}' at Lsn {last_record_lsn} for tenant: {tenant_id}",
1179 0 : timeline_info.timeline_id
1180 0 : );
1181 : }
1182 : // TODO: rename to import-basebackup-plus-wal
1183 0 : TimelineCmd::Import(args) => {
1184 0 : let tenant_id = get_tenant_id(args.tenant_id, env)?;
1185 0 : let timeline_id = args.timeline_id;
1186 0 : let branch_name = &args.branch_name;
1187 0 :
1188 0 : // Parse base inputs
1189 0 : let base = (args.base_lsn, args.base_tarfile.clone());
1190 0 :
1191 0 : // Parse pg_wal inputs
1192 0 : let wal_tarfile = args.wal_tarfile.clone();
1193 0 : let end_lsn = args.end_lsn;
1194 0 : // TODO validate both or none are provided
1195 0 : let pg_wal = end_lsn.zip(wal_tarfile);
1196 0 :
1197 0 : println!("Importing timeline into pageserver ...");
1198 0 : pageserver
1199 0 : .timeline_import(tenant_id, timeline_id, base, pg_wal, args.pg_version)
1200 0 : .await?;
1201 0 : env.register_branch_mapping(branch_name.to_string(), tenant_id, timeline_id)?;
1202 0 : println!("Done");
1203 : }
1204 0 : TimelineCmd::Branch(args) => {
1205 0 : let tenant_id = get_tenant_id(args.tenant_id, env)?;
1206 0 : let new_timeline_id = args.timeline_id.unwrap_or(TimelineId::generate());
1207 0 : let new_branch_name = &args.branch_name;
1208 0 : let ancestor_branch_name = args
1209 0 : .ancestor_branch_name
1210 0 : .clone()
1211 0 : .unwrap_or(DEFAULT_BRANCH_NAME.to_owned());
1212 0 : let ancestor_timeline_id = env
1213 0 : .get_branch_timeline_id(&ancestor_branch_name, tenant_id)
1214 0 : .ok_or_else(|| {
1215 0 : anyhow!("Found no timeline id for branch name '{ancestor_branch_name}'")
1216 0 : })?;
1217 :
1218 0 : let start_lsn = args.ancestor_start_lsn;
1219 0 : let storage_controller = StorageController::from_env(env);
1220 0 : let create_req = TimelineCreateRequest {
1221 0 : new_timeline_id,
1222 0 : mode: pageserver_api::models::TimelineCreateRequestMode::Branch {
1223 0 : ancestor_timeline_id,
1224 0 : ancestor_start_lsn: start_lsn,
1225 0 : pg_version: None,
1226 0 : },
1227 0 : };
1228 0 : let timeline_info = storage_controller
1229 0 : .tenant_timeline_create(tenant_id, create_req)
1230 0 : .await?;
1231 :
1232 0 : let last_record_lsn = timeline_info.last_record_lsn;
1233 0 :
1234 0 : env.register_branch_mapping(new_branch_name.to_string(), tenant_id, new_timeline_id)?;
1235 :
1236 0 : println!(
1237 0 : "Created timeline '{}' at Lsn {last_record_lsn} for tenant: {tenant_id}. Ancestor timeline: '{ancestor_branch_name}'",
1238 0 : timeline_info.timeline_id
1239 0 : );
1240 : }
1241 : }
1242 :
1243 0 : Ok(())
1244 0 : }
1245 :
1246 0 : async fn handle_endpoint(subcmd: &EndpointCmd, env: &local_env::LocalEnv) -> Result<()> {
1247 0 : let mut cplane = ComputeControlPlane::load(env.clone())?;
1248 :
1249 0 : match subcmd {
1250 0 : EndpointCmd::List(args) => {
1251 : // TODO(sharding): this command shouldn't have to specify a shard ID: we should ask the storage controller
1252 : // where shard 0 is attached, and query there.
1253 0 : let tenant_shard_id = get_tenant_shard_id(args.tenant_shard_id, env)?;
1254 0 : let timeline_infos = get_timeline_infos(env, &tenant_shard_id)
1255 0 : .await
1256 0 : .unwrap_or_else(|e| {
1257 0 : eprintln!("Failed to load timeline info: {}", e);
1258 0 : HashMap::new()
1259 0 : });
1260 0 :
1261 0 : let timeline_name_mappings = env.timeline_name_mappings();
1262 0 :
1263 0 : let mut table = comfy_table::Table::new();
1264 0 :
1265 0 : table.load_preset(comfy_table::presets::NOTHING);
1266 0 :
1267 0 : table.set_header([
1268 0 : "ENDPOINT",
1269 0 : "ADDRESS",
1270 0 : "TIMELINE",
1271 0 : "BRANCH NAME",
1272 0 : "LSN",
1273 0 : "STATUS",
1274 0 : ]);
1275 :
1276 0 : for (endpoint_id, endpoint) in cplane
1277 0 : .endpoints
1278 0 : .iter()
1279 0 : .filter(|(_, endpoint)| endpoint.tenant_id == tenant_shard_id.tenant_id)
1280 : {
1281 0 : let lsn_str = match endpoint.mode {
1282 0 : ComputeMode::Static(lsn) => {
1283 0 : // -> read-only endpoint
1284 0 : // Use the node's LSN.
1285 0 : lsn.to_string()
1286 : }
1287 : _ => {
1288 : // -> primary endpoint or hot replica
1289 : // Use the LSN at the end of the timeline.
1290 0 : timeline_infos
1291 0 : .get(&endpoint.timeline_id)
1292 0 : .map(|bi| bi.last_record_lsn.to_string())
1293 0 : .unwrap_or_else(|| "?".to_string())
1294 : }
1295 : };
1296 :
1297 0 : let branch_name = timeline_name_mappings
1298 0 : .get(&TenantTimelineId::new(
1299 0 : tenant_shard_id.tenant_id,
1300 0 : endpoint.timeline_id,
1301 0 : ))
1302 0 : .map(|name| name.as_str())
1303 0 : .unwrap_or("?");
1304 0 :
1305 0 : table.add_row([
1306 0 : endpoint_id.as_str(),
1307 0 : &endpoint.pg_address.to_string(),
1308 0 : &endpoint.timeline_id.to_string(),
1309 0 : branch_name,
1310 0 : lsn_str.as_str(),
1311 0 : &format!("{}", endpoint.status()),
1312 0 : ]);
1313 0 : }
1314 :
1315 0 : println!("{table}");
1316 : }
1317 0 : EndpointCmd::Create(args) => {
1318 0 : let tenant_id = get_tenant_id(args.tenant_id, env)?;
1319 0 : let branch_name = args
1320 0 : .branch_name
1321 0 : .clone()
1322 0 : .unwrap_or(DEFAULT_BRANCH_NAME.to_owned());
1323 0 : let endpoint_id = args
1324 0 : .endpoint_id
1325 0 : .clone()
1326 0 : .unwrap_or_else(|| format!("ep-{branch_name}"));
1327 :
1328 0 : let timeline_id = env
1329 0 : .get_branch_timeline_id(&branch_name, tenant_id)
1330 0 : .ok_or_else(|| anyhow!("Found no timeline id for branch name '{branch_name}'"))?;
1331 :
1332 0 : let mode = match (args.lsn, args.hot_standby) {
1333 0 : (Some(lsn), false) => ComputeMode::Static(lsn),
1334 0 : (None, true) => ComputeMode::Replica,
1335 0 : (None, false) => ComputeMode::Primary,
1336 0 : (Some(_), true) => anyhow::bail!("cannot specify both lsn and hot-standby"),
1337 : };
1338 :
1339 0 : match (mode, args.hot_standby) {
1340 : (ComputeMode::Static(_), true) => {
1341 0 : bail!("Cannot start a node in hot standby mode when it is already configured as a static replica")
1342 : }
1343 : (ComputeMode::Primary, true) => {
1344 0 : bail!("Cannot start a node as a hot standby replica, it is already configured as primary node")
1345 : }
1346 0 : _ => {}
1347 0 : }
1348 0 :
1349 0 : if !args.allow_multiple {
1350 0 : cplane.check_conflicting_endpoints(mode, tenant_id, timeline_id)?;
1351 0 : }
1352 :
1353 0 : cplane.new_endpoint(
1354 0 : &endpoint_id,
1355 0 : tenant_id,
1356 0 : timeline_id,
1357 0 : args.pg_port,
1358 0 : args.external_http_port,
1359 0 : args.internal_http_port,
1360 0 : args.pg_version,
1361 0 : mode,
1362 0 : !args.update_catalog,
1363 0 : false,
1364 0 : )?;
1365 : }
1366 0 : EndpointCmd::Start(args) => {
1367 0 : let endpoint_id = &args.endpoint_id;
1368 0 : let pageserver_id = args.endpoint_pageserver_id;
1369 0 : let remote_ext_config = &args.remote_ext_config;
1370 :
1371 : // If --safekeepers argument is given, use only the listed
1372 : // safekeeper nodes; otherwise all from the env.
1373 0 : let safekeepers = if let Some(safekeepers) = parse_safekeepers(&args.safekeepers)? {
1374 0 : safekeepers
1375 : } else {
1376 0 : env.safekeepers.iter().map(|sk| sk.id).collect()
1377 : };
1378 :
1379 0 : let endpoint = cplane
1380 0 : .endpoints
1381 0 : .get(endpoint_id.as_str())
1382 0 : .ok_or_else(|| anyhow::anyhow!("endpoint {endpoint_id} not found"))?;
1383 :
1384 0 : if !args.allow_multiple {
1385 0 : cplane.check_conflicting_endpoints(
1386 0 : endpoint.mode,
1387 0 : endpoint.tenant_id,
1388 0 : endpoint.timeline_id,
1389 0 : )?;
1390 0 : }
1391 :
1392 0 : let (pageservers, stripe_size) = if let Some(pageserver_id) = pageserver_id {
1393 0 : let conf = env.get_pageserver_conf(pageserver_id).unwrap();
1394 0 : let parsed = parse_host_port(&conf.listen_pg_addr).expect("Bad config");
1395 0 : (
1396 0 : vec![(parsed.0, parsed.1.unwrap_or(5432))],
1397 0 : // If caller is telling us what pageserver to use, this is not a tenant which is
1398 0 : // full managed by storage controller, therefore not sharded.
1399 0 : ShardParameters::DEFAULT_STRIPE_SIZE,
1400 0 : )
1401 : } else {
1402 : // Look up the currently attached location of the tenant, and its striping metadata,
1403 : // to pass these on to postgres.
1404 0 : let storage_controller = StorageController::from_env(env);
1405 0 : let locate_result = storage_controller.tenant_locate(endpoint.tenant_id).await?;
1406 0 : let pageservers = futures::future::try_join_all(
1407 0 : locate_result.shards.into_iter().map(|shard| async move {
1408 0 : if let ComputeMode::Static(lsn) = endpoint.mode {
1409 : // Initialize LSN leases for static computes.
1410 0 : let conf = env.get_pageserver_conf(shard.node_id).unwrap();
1411 0 : let pageserver = PageServerNode::from_env(env, conf);
1412 0 :
1413 0 : pageserver
1414 0 : .http_client
1415 0 : .timeline_init_lsn_lease(shard.shard_id, endpoint.timeline_id, lsn)
1416 0 : .await?;
1417 0 : }
1418 :
1419 0 : anyhow::Ok((
1420 0 : Host::parse(&shard.listen_pg_addr)
1421 0 : .expect("Storage controller reported bad hostname"),
1422 0 : shard.listen_pg_port,
1423 0 : ))
1424 0 : }),
1425 0 : )
1426 0 : .await?;
1427 0 : let stripe_size = locate_result.shard_params.stripe_size;
1428 0 :
1429 0 : (pageservers, stripe_size)
1430 : };
1431 0 : assert!(!pageservers.is_empty());
1432 :
1433 0 : let ps_conf = env.get_pageserver_conf(DEFAULT_PAGESERVER_ID)?;
1434 0 : let auth_token = if matches!(ps_conf.pg_auth_type, AuthType::NeonJWT) {
1435 0 : let claims = Claims::new(Some(endpoint.tenant_id), Scope::Tenant);
1436 0 :
1437 0 : Some(env.generate_auth_token(&claims)?)
1438 : } else {
1439 0 : None
1440 : };
1441 :
1442 0 : println!("Starting existing endpoint {endpoint_id}...");
1443 0 : endpoint
1444 0 : .start(
1445 0 : &auth_token,
1446 0 : safekeepers,
1447 0 : pageservers,
1448 0 : remote_ext_config.as_ref(),
1449 0 : stripe_size.0 as usize,
1450 0 : args.create_test_user,
1451 0 : )
1452 0 : .await?;
1453 : }
1454 0 : EndpointCmd::Reconfigure(args) => {
1455 0 : let endpoint_id = &args.endpoint_id;
1456 0 : let endpoint = cplane
1457 0 : .endpoints
1458 0 : .get(endpoint_id.as_str())
1459 0 : .with_context(|| format!("postgres endpoint {endpoint_id} is not found"))?;
1460 0 : let pageservers = if let Some(ps_id) = args.endpoint_pageserver_id {
1461 0 : let pageserver = PageServerNode::from_env(env, env.get_pageserver_conf(ps_id)?);
1462 0 : vec![(
1463 0 : pageserver.pg_connection_config.host().clone(),
1464 0 : pageserver.pg_connection_config.port(),
1465 0 : )]
1466 : } else {
1467 0 : let storage_controller = StorageController::from_env(env);
1468 0 : storage_controller
1469 0 : .tenant_locate(endpoint.tenant_id)
1470 0 : .await?
1471 : .shards
1472 0 : .into_iter()
1473 0 : .map(|shard| {
1474 0 : (
1475 0 : Host::parse(&shard.listen_pg_addr)
1476 0 : .expect("Storage controller reported malformed host"),
1477 0 : shard.listen_pg_port,
1478 0 : )
1479 0 : })
1480 0 : .collect::<Vec<_>>()
1481 : };
1482 : // If --safekeepers argument is given, use only the listed
1483 : // safekeeper nodes; otherwise all from the env.
1484 0 : let safekeepers = parse_safekeepers(&args.safekeepers)?;
1485 0 : endpoint.reconfigure(pageservers, None, safekeepers).await?;
1486 : }
1487 0 : EndpointCmd::Stop(args) => {
1488 0 : let endpoint_id = &args.endpoint_id;
1489 0 : let endpoint = cplane
1490 0 : .endpoints
1491 0 : .get(endpoint_id)
1492 0 : .with_context(|| format!("postgres endpoint {endpoint_id} is not found"))?;
1493 0 : endpoint.stop(&args.mode, args.destroy)?;
1494 : }
1495 : }
1496 :
1497 0 : Ok(())
1498 0 : }
1499 :
1500 : /// Parse --safekeepers as list of safekeeper ids.
1501 0 : fn parse_safekeepers(safekeepers_str: &Option<String>) -> Result<Option<Vec<NodeId>>> {
1502 0 : if let Some(safekeepers_str) = safekeepers_str {
1503 0 : let mut safekeepers: Vec<NodeId> = Vec::new();
1504 0 : for sk_id in safekeepers_str.split(',').map(str::trim) {
1505 0 : let sk_id = NodeId(
1506 0 : u64::from_str(sk_id)
1507 0 : .map_err(|_| anyhow!("invalid node ID \"{sk_id}\" in --safekeepers list"))?,
1508 : );
1509 0 : safekeepers.push(sk_id);
1510 : }
1511 0 : Ok(Some(safekeepers))
1512 : } else {
1513 0 : Ok(None)
1514 : }
1515 0 : }
1516 :
1517 0 : fn handle_mappings(subcmd: &MappingsCmd, env: &mut local_env::LocalEnv) -> Result<()> {
1518 0 : match subcmd {
1519 0 : MappingsCmd::Map(args) => {
1520 0 : env.register_branch_mapping(
1521 0 : args.branch_name.to_owned(),
1522 0 : args.tenant_id,
1523 0 : args.timeline_id,
1524 0 : )?;
1525 :
1526 0 : Ok(())
1527 : }
1528 : }
1529 0 : }
1530 :
1531 0 : fn get_pageserver(
1532 0 : env: &local_env::LocalEnv,
1533 0 : pageserver_id_arg: Option<NodeId>,
1534 0 : ) -> Result<PageServerNode> {
1535 0 : let node_id = pageserver_id_arg.unwrap_or(DEFAULT_PAGESERVER_ID);
1536 0 :
1537 0 : Ok(PageServerNode::from_env(
1538 0 : env,
1539 0 : env.get_pageserver_conf(node_id)?,
1540 : ))
1541 0 : }
1542 :
1543 0 : async fn handle_pageserver(subcmd: &PageserverCmd, env: &local_env::LocalEnv) -> Result<()> {
1544 0 : match subcmd {
1545 0 : PageserverCmd::Start(args) => {
1546 0 : if let Err(e) = get_pageserver(env, args.pageserver_id)?
1547 0 : .start(&args.start_timeout)
1548 0 : .await
1549 : {
1550 0 : eprintln!("pageserver start failed: {e}");
1551 0 : exit(1);
1552 0 : }
1553 : }
1554 :
1555 0 : PageserverCmd::Stop(args) => {
1556 0 : let immediate = match args.stop_mode {
1557 0 : StopMode::Fast => false,
1558 0 : StopMode::Immediate => true,
1559 : };
1560 0 : if let Err(e) = get_pageserver(env, args.pageserver_id)?.stop(immediate) {
1561 0 : eprintln!("pageserver stop failed: {}", e);
1562 0 : exit(1);
1563 0 : }
1564 : }
1565 :
1566 0 : PageserverCmd::Restart(args) => {
1567 0 : let pageserver = get_pageserver(env, args.pageserver_id)?;
1568 : //TODO what shutdown strategy should we use here?
1569 0 : if let Err(e) = pageserver.stop(false) {
1570 0 : eprintln!("pageserver stop failed: {}", e);
1571 0 : exit(1);
1572 0 : }
1573 :
1574 0 : if let Err(e) = pageserver.start(&args.start_timeout).await {
1575 0 : eprintln!("pageserver start failed: {e}");
1576 0 : exit(1);
1577 0 : }
1578 : }
1579 :
1580 0 : PageserverCmd::Status(args) => {
1581 0 : match get_pageserver(env, args.pageserver_id)?
1582 0 : .check_status()
1583 0 : .await
1584 : {
1585 0 : Ok(_) => println!("Page server is up and running"),
1586 0 : Err(err) => {
1587 0 : eprintln!("Page server is not available: {}", err);
1588 0 : exit(1);
1589 : }
1590 : }
1591 : }
1592 : }
1593 0 : Ok(())
1594 0 : }
1595 :
1596 0 : async fn handle_storage_controller(
1597 0 : subcmd: &StorageControllerCmd,
1598 0 : env: &local_env::LocalEnv,
1599 0 : ) -> Result<()> {
1600 0 : let svc = StorageController::from_env(env);
1601 0 : match subcmd {
1602 0 : StorageControllerCmd::Start(args) => {
1603 0 : let start_args = NeonStorageControllerStartArgs {
1604 0 : instance_id: args.instance_id,
1605 0 : base_port: args.base_port,
1606 0 : start_timeout: args.start_timeout,
1607 0 : };
1608 :
1609 0 : if let Err(e) = svc.start(start_args).await {
1610 0 : eprintln!("start failed: {e}");
1611 0 : exit(1);
1612 0 : }
1613 : }
1614 :
1615 0 : StorageControllerCmd::Stop(args) => {
1616 0 : let stop_args = NeonStorageControllerStopArgs {
1617 0 : instance_id: args.instance_id,
1618 0 : immediate: match args.stop_mode {
1619 0 : StopMode::Fast => false,
1620 0 : StopMode::Immediate => true,
1621 : },
1622 : };
1623 0 : if let Err(e) = svc.stop(stop_args).await {
1624 0 : eprintln!("stop failed: {}", e);
1625 0 : exit(1);
1626 0 : }
1627 : }
1628 : }
1629 0 : Ok(())
1630 0 : }
1631 :
1632 0 : fn get_safekeeper(env: &local_env::LocalEnv, id: NodeId) -> Result<SafekeeperNode> {
1633 0 : if let Some(node) = env.safekeepers.iter().find(|node| node.id == id) {
1634 0 : Ok(SafekeeperNode::from_env(env, node))
1635 : } else {
1636 0 : bail!("could not find safekeeper {id}")
1637 : }
1638 0 : }
1639 :
1640 0 : async fn handle_safekeeper(subcmd: &SafekeeperCmd, env: &local_env::LocalEnv) -> Result<()> {
1641 0 : match subcmd {
1642 0 : SafekeeperCmd::Start(args) => {
1643 0 : let safekeeper = get_safekeeper(env, args.id)?;
1644 :
1645 0 : if let Err(e) = safekeeper.start(&args.extra_opt, &args.start_timeout).await {
1646 0 : eprintln!("safekeeper start failed: {}", e);
1647 0 : exit(1);
1648 0 : }
1649 : }
1650 :
1651 0 : SafekeeperCmd::Stop(args) => {
1652 0 : let safekeeper = get_safekeeper(env, args.id)?;
1653 0 : let immediate = match args.stop_mode {
1654 0 : StopMode::Fast => false,
1655 0 : StopMode::Immediate => true,
1656 : };
1657 0 : if let Err(e) = safekeeper.stop(immediate) {
1658 0 : eprintln!("safekeeper stop failed: {}", e);
1659 0 : exit(1);
1660 0 : }
1661 : }
1662 :
1663 0 : SafekeeperCmd::Restart(args) => {
1664 0 : let safekeeper = get_safekeeper(env, args.id)?;
1665 0 : let immediate = match args.stop_mode {
1666 0 : StopMode::Fast => false,
1667 0 : StopMode::Immediate => true,
1668 : };
1669 :
1670 0 : if let Err(e) = safekeeper.stop(immediate) {
1671 0 : eprintln!("safekeeper stop failed: {}", e);
1672 0 : exit(1);
1673 0 : }
1674 :
1675 0 : if let Err(e) = safekeeper.start(&args.extra_opt, &args.start_timeout).await {
1676 0 : eprintln!("safekeeper start failed: {}", e);
1677 0 : exit(1);
1678 0 : }
1679 : }
1680 : }
1681 0 : Ok(())
1682 0 : }
1683 :
1684 0 : async fn handle_storage_broker(subcmd: &StorageBrokerCmd, env: &local_env::LocalEnv) -> Result<()> {
1685 0 : match subcmd {
1686 0 : StorageBrokerCmd::Start(args) => {
1687 0 : if let Err(e) = broker::start_broker_process(env, &args.start_timeout).await {
1688 0 : eprintln!("broker start failed: {e}");
1689 0 : exit(1);
1690 0 : }
1691 : }
1692 :
1693 0 : StorageBrokerCmd::Stop(_args) => {
1694 : // FIXME: stop_mode unused
1695 0 : if let Err(e) = broker::stop_broker_process(env) {
1696 0 : eprintln!("broker stop failed: {e}");
1697 0 : exit(1);
1698 0 : }
1699 : }
1700 : }
1701 0 : Ok(())
1702 0 : }
1703 :
1704 0 : async fn handle_start_all(
1705 0 : args: &StartCmdArgs,
1706 0 : env: &'static local_env::LocalEnv,
1707 0 : ) -> anyhow::Result<()> {
1708 : // FIXME: this was called "retry_timeout", is it right?
1709 0 : let Err(errors) = handle_start_all_impl(env, args.timeout).await else {
1710 0 : neon_start_status_check(env, args.timeout.as_ref())
1711 0 : .await
1712 0 : .context("status check after successful startup of all services")?;
1713 0 : return Ok(());
1714 : };
1715 :
1716 0 : eprintln!("startup failed because one or more services could not be started");
1717 :
1718 0 : for e in errors {
1719 0 : eprintln!("{e}");
1720 0 : let debug_repr = format!("{e:?}");
1721 0 : for line in debug_repr.lines() {
1722 0 : eprintln!(" {line}");
1723 0 : }
1724 : }
1725 :
1726 0 : try_stop_all(env, true).await;
1727 :
1728 0 : exit(2);
1729 0 : }
1730 :
1731 : /// Returns Ok() if and only if all services could be started successfully.
1732 : /// Otherwise, returns the list of errors that occurred during startup.
1733 0 : async fn handle_start_all_impl(
1734 0 : env: &'static local_env::LocalEnv,
1735 0 : retry_timeout: humantime::Duration,
1736 0 : ) -> Result<(), Vec<anyhow::Error>> {
1737 0 : // Endpoints are not started automatically
1738 0 :
1739 0 : let mut js = JoinSet::new();
1740 :
1741 : // force infalliblity through closure
1742 : #[allow(clippy::redundant_closure_call)]
1743 0 : (|| {
1744 0 : js.spawn(async move {
1745 0 : let retry_timeout = retry_timeout;
1746 0 : broker::start_broker_process(env, &retry_timeout).await
1747 0 : });
1748 0 :
1749 0 : js.spawn(async move {
1750 0 : let storage_controller = StorageController::from_env(env);
1751 0 : storage_controller
1752 0 : .start(NeonStorageControllerStartArgs::with_default_instance_id(
1753 0 : retry_timeout,
1754 0 : ))
1755 0 : .await
1756 0 : .map_err(|e| e.context("start storage_controller"))
1757 0 : });
1758 :
1759 0 : for ps_conf in &env.pageservers {
1760 0 : js.spawn(async move {
1761 0 : let pageserver = PageServerNode::from_env(env, ps_conf);
1762 0 : pageserver
1763 0 : .start(&retry_timeout)
1764 0 : .await
1765 0 : .map_err(|e| e.context(format!("start pageserver {}", ps_conf.id)))
1766 0 : });
1767 0 : }
1768 :
1769 0 : for node in env.safekeepers.iter() {
1770 0 : js.spawn(async move {
1771 0 : let safekeeper = SafekeeperNode::from_env(env, node);
1772 0 : safekeeper
1773 0 : .start(&[], &retry_timeout)
1774 0 : .await
1775 0 : .map_err(|e| e.context(format!("start safekeeper {}", safekeeper.id)))
1776 0 : });
1777 0 : }
1778 0 : })();
1779 0 :
1780 0 : let mut errors = Vec::new();
1781 0 : while let Some(result) = js.join_next().await {
1782 0 : let result = result.expect("we don't panic or cancel the tasks");
1783 0 : if let Err(e) = result {
1784 0 : errors.push(e);
1785 0 : }
1786 : }
1787 :
1788 0 : if !errors.is_empty() {
1789 0 : return Err(errors);
1790 0 : }
1791 0 :
1792 0 : Ok(())
1793 0 : }
1794 :
1795 0 : async fn neon_start_status_check(
1796 0 : env: &local_env::LocalEnv,
1797 0 : retry_timeout: &Duration,
1798 0 : ) -> anyhow::Result<()> {
1799 : const RETRY_INTERVAL: Duration = Duration::from_millis(100);
1800 : const NOTICE_AFTER_RETRIES: Duration = Duration::from_secs(5);
1801 :
1802 0 : let storcon = StorageController::from_env(env);
1803 0 :
1804 0 : let retries = retry_timeout.as_millis() / RETRY_INTERVAL.as_millis();
1805 0 : let notice_after_retries = retry_timeout.as_millis() / NOTICE_AFTER_RETRIES.as_millis();
1806 0 :
1807 0 : println!("\nRunning neon status check");
1808 :
1809 0 : for retry in 0..retries {
1810 0 : if retry == notice_after_retries {
1811 0 : println!("\nNeon status check has not passed yet, continuing to wait")
1812 0 : }
1813 :
1814 0 : let mut passed = true;
1815 0 : let mut nodes = storcon.node_list().await?;
1816 0 : let mut pageservers = env.pageservers.clone();
1817 0 :
1818 0 : if nodes.len() != pageservers.len() {
1819 0 : continue;
1820 0 : }
1821 0 :
1822 0 : nodes.sort_by_key(|ps| ps.id);
1823 0 : pageservers.sort_by_key(|ps| ps.id);
1824 :
1825 0 : for (idx, pageserver) in pageservers.iter().enumerate() {
1826 0 : let node = &nodes[idx];
1827 0 : if node.id != pageserver.id {
1828 0 : passed = false;
1829 0 : break;
1830 0 : }
1831 :
1832 0 : if !matches!(node.availability, NodeAvailabilityWrapper::Active) {
1833 0 : passed = false;
1834 0 : break;
1835 0 : }
1836 : }
1837 :
1838 0 : if passed {
1839 0 : println!("\nNeon started and passed status check");
1840 0 : return Ok(());
1841 0 : }
1842 0 :
1843 0 : tokio::time::sleep(RETRY_INTERVAL).await;
1844 : }
1845 :
1846 0 : anyhow::bail!("\nNeon passed status check")
1847 0 : }
1848 :
1849 0 : async fn handle_stop_all(args: &StopCmdArgs, env: &local_env::LocalEnv) -> Result<()> {
1850 0 : let immediate = match args.mode {
1851 0 : StopMode::Fast => false,
1852 0 : StopMode::Immediate => true,
1853 : };
1854 :
1855 0 : try_stop_all(env, immediate).await;
1856 :
1857 0 : Ok(())
1858 0 : }
1859 :
1860 0 : async fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
1861 0 : // Stop all endpoints
1862 0 : match ComputeControlPlane::load(env.clone()) {
1863 0 : Ok(cplane) => {
1864 0 : for (_k, node) in cplane.endpoints {
1865 0 : if let Err(e) = node.stop(if immediate { "immediate" } else { "fast" }, false) {
1866 0 : eprintln!("postgres stop failed: {e:#}");
1867 0 : }
1868 : }
1869 : }
1870 0 : Err(e) => {
1871 0 : eprintln!("postgres stop failed, could not restore control plane data from env: {e:#}")
1872 : }
1873 : }
1874 :
1875 0 : for ps_conf in &env.pageservers {
1876 0 : let pageserver = PageServerNode::from_env(env, ps_conf);
1877 0 : if let Err(e) = pageserver.stop(immediate) {
1878 0 : eprintln!("pageserver {} stop failed: {:#}", ps_conf.id, e);
1879 0 : }
1880 : }
1881 :
1882 0 : for node in env.safekeepers.iter() {
1883 0 : let safekeeper = SafekeeperNode::from_env(env, node);
1884 0 : if let Err(e) = safekeeper.stop(immediate) {
1885 0 : eprintln!("safekeeper {} stop failed: {:#}", safekeeper.id, e);
1886 0 : }
1887 : }
1888 :
1889 0 : if let Err(e) = broker::stop_broker_process(env) {
1890 0 : eprintln!("neon broker stop failed: {e:#}");
1891 0 : }
1892 :
1893 : // Stop all storage controller instances. In the most common case there's only one,
1894 : // but iterate though the base data directory in order to discover the instances.
1895 0 : let storcon_instances = env
1896 0 : .storage_controller_instances()
1897 0 : .await
1898 0 : .expect("Must inspect data dir");
1899 0 : for (instance_id, _instance_dir_path) in storcon_instances {
1900 0 : let storage_controller = StorageController::from_env(env);
1901 0 : let stop_args = NeonStorageControllerStopArgs {
1902 0 : instance_id,
1903 0 : immediate,
1904 0 : };
1905 :
1906 0 : if let Err(e) = storage_controller.stop(stop_args).await {
1907 0 : eprintln!("Storage controller instance {instance_id} stop failed: {e:#}");
1908 0 : }
1909 : }
1910 0 : }
|
