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