Line data Source code
1 : //! Code to manage compute endpoints
2 : //!
3 : //! In the local test environment, the data for each endpoint is stored in
4 : //!
5 : //! ```text
6 : //! .neon/endpoints/<endpoint id>
7 : //! ```
8 : //!
9 : //! Some basic information about the endpoint, like the tenant and timeline IDs,
10 : //! are stored in the `endpoint.json` file. The `endpoint.json` file is created
11 : //! when the endpoint is created, and doesn't change afterwards.
12 : //!
13 : //! The endpoint is managed by the `compute_ctl` binary. When an endpoint is
14 : //! started, we launch `compute_ctl` It synchronizes the safekeepers, downloads
15 : //! the basebackup from the pageserver to initialize the the data directory, and
16 : //! finally launches the PostgreSQL process. It watches the PostgreSQL process
17 : //! until it exits.
18 : //!
19 : //! When an endpoint is created, a `postgresql.conf` file is also created in
20 : //! the endpoint's directory. The file can be modified before starting PostgreSQL.
21 : //! However, the `postgresql.conf` file in the endpoint directory is not used directly
22 : //! by PostgreSQL. It is passed to `compute_ctl`, and `compute_ctl` writes another
23 : //! copy of it in the data directory.
24 : //!
25 : //! Directory contents:
26 : //!
27 : //! ```text
28 : //! .neon/endpoints/main/
29 : //! compute.log - log output of `compute_ctl` and `postgres`
30 : //! endpoint.json - serialized `EndpointConf` struct
31 : //! postgresql.conf - postgresql settings
32 : //! spec.json - passed to `compute_ctl`
33 : //! pgdata/
34 : //! postgresql.conf - copy of postgresql.conf created by `compute_ctl`
35 : //! zenith.signal
36 : //! <other PostgreSQL files>
37 : //! ```
38 : //!
39 : use std::collections::BTreeMap;
40 : use std::net::SocketAddr;
41 : use std::net::TcpStream;
42 : use std::path::PathBuf;
43 : use std::process::Command;
44 : use std::sync::Arc;
45 : use std::time::Duration;
46 :
47 : use anyhow::{anyhow, bail, Context, Result};
48 : use compute_api::spec::RemoteExtSpec;
49 : use nix::sys::signal::kill;
50 : use nix::sys::signal::Signal;
51 : use serde::{Deserialize, Serialize};
52 : use url::Host;
53 : use utils::id::{NodeId, TenantId, TimelineId};
54 :
55 : use crate::attachment_service::AttachmentService;
56 : use crate::local_env::LocalEnv;
57 : use crate::postgresql_conf::PostgresConf;
58 :
59 : use compute_api::responses::{ComputeState, ComputeStatus};
60 : use compute_api::spec::{Cluster, ComputeFeature, ComputeMode, ComputeSpec};
61 :
62 : // contents of a endpoint.json file
63 73815 : #[derive(Serialize, Deserialize, PartialEq, Eq, Clone, Debug)]
64 : pub struct EndpointConf {
65 : endpoint_id: String,
66 : tenant_id: TenantId,
67 : timeline_id: TimelineId,
68 : mode: ComputeMode,
69 : pg_port: u16,
70 : http_port: u16,
71 : pg_version: u32,
72 : skip_pg_catalog_updates: bool,
73 : features: Vec<ComputeFeature>,
74 : }
75 :
76 : //
77 : // ComputeControlPlane
78 : //
79 : pub struct ComputeControlPlane {
80 : base_port: u16,
81 :
82 : // endpoint ID is the key
83 : pub endpoints: BTreeMap<String, Arc<Endpoint>>,
84 :
85 : env: LocalEnv,
86 : }
87 :
88 : impl ComputeControlPlane {
89 : // Load current endpoints from the endpoints/ subdirectories
90 2408 : pub fn load(env: LocalEnv) -> Result<ComputeControlPlane> {
91 2408 : let mut endpoints = BTreeMap::default();
92 3885 : for endpoint_dir in std::fs::read_dir(env.endpoints_path())
93 2408 : .with_context(|| format!("failed to list {}", env.endpoints_path().display()))?
94 3885 : {
95 3885 : let ep = Endpoint::from_dir_entry(endpoint_dir?, &env)?;
96 3885 : endpoints.insert(ep.endpoint_id.clone(), Arc::new(ep));
97 : }
98 :
99 2408 : Ok(ComputeControlPlane {
100 2408 : base_port: 55431,
101 2408 : endpoints,
102 2408 : env,
103 2408 : })
104 2408 : }
105 :
106 6 : fn get_port(&mut self) -> u16 {
107 6 : 1 + self
108 6 : .endpoints
109 6 : .values()
110 6 : .map(|ep| std::cmp::max(ep.pg_address.port(), ep.http_address.port()))
111 6 : .max()
112 6 : .unwrap_or(self.base_port)
113 6 : }
114 :
115 : #[allow(clippy::too_many_arguments)]
116 518 : pub fn new_endpoint(
117 518 : &mut self,
118 518 : endpoint_id: &str,
119 518 : tenant_id: TenantId,
120 518 : timeline_id: TimelineId,
121 518 : pg_port: Option<u16>,
122 518 : http_port: Option<u16>,
123 518 : pg_version: u32,
124 518 : mode: ComputeMode,
125 518 : ) -> Result<Arc<Endpoint>> {
126 518 : let pg_port = pg_port.unwrap_or_else(|| self.get_port());
127 518 : let http_port = http_port.unwrap_or_else(|| self.get_port() + 1);
128 518 : let ep = Arc::new(Endpoint {
129 518 : endpoint_id: endpoint_id.to_owned(),
130 518 : pg_address: SocketAddr::new("127.0.0.1".parse().unwrap(), pg_port),
131 518 : http_address: SocketAddr::new("127.0.0.1".parse().unwrap(), http_port),
132 518 : env: self.env.clone(),
133 518 : timeline_id,
134 518 : mode,
135 518 : tenant_id,
136 518 : pg_version,
137 518 : // We don't setup roles and databases in the spec locally, so we don't need to
138 518 : // do catalog updates. Catalog updates also include check availability
139 518 : // data creation. Yet, we have tests that check that size and db dump
140 518 : // before and after start are the same. So, skip catalog updates,
141 518 : // with this we basically test a case of waking up an idle compute, where
142 518 : // we also skip catalog updates in the cloud.
143 518 : skip_pg_catalog_updates: true,
144 518 : features: vec![],
145 518 : });
146 518 :
147 518 : ep.create_endpoint_dir()?;
148 : std::fs::write(
149 518 : ep.endpoint_path().join("endpoint.json"),
150 518 : serde_json::to_string_pretty(&EndpointConf {
151 518 : endpoint_id: endpoint_id.to_string(),
152 518 : tenant_id,
153 518 : timeline_id,
154 518 : mode,
155 518 : http_port,
156 518 : pg_port,
157 518 : pg_version,
158 518 : skip_pg_catalog_updates: true,
159 518 : features: vec![],
160 518 : })?,
161 0 : )?;
162 : std::fs::write(
163 518 : ep.endpoint_path().join("postgresql.conf"),
164 518 : ep.setup_pg_conf()?.to_string(),
165 0 : )?;
166 :
167 518 : self.endpoints
168 518 : .insert(ep.endpoint_id.clone(), Arc::clone(&ep));
169 518 :
170 518 : Ok(ep)
171 518 : }
172 :
173 1126 : pub fn check_conflicting_endpoints(
174 1126 : &self,
175 1126 : mode: ComputeMode,
176 1126 : tenant_id: TenantId,
177 1126 : timeline_id: TimelineId,
178 1126 : ) -> Result<()> {
179 1126 : if matches!(mode, ComputeMode::Primary) {
180 : // this check is not complete, as you could have a concurrent attempt at
181 : // creating another primary, both reading the state before checking it here,
182 : // but it's better than nothing.
183 1513 : let mut duplicates = self.endpoints.iter().filter(|(_k, v)| {
184 1513 : v.tenant_id == tenant_id
185 1392 : && v.timeline_id == timeline_id
186 956 : && v.mode == mode
187 952 : && v.status() != EndpointStatus::Stopped
188 1513 : });
189 :
190 1024 : if let Some((key, _)) = duplicates.next() {
191 29 : bail!("attempting to create a duplicate primary endpoint on tenant {tenant_id}, timeline {timeline_id}: endpoint {key:?} exists already. please don't do this, it is not supported.");
192 995 : }
193 102 : }
194 1097 : Ok(())
195 1126 : }
196 : }
197 :
198 : ///////////////////////////////////////////////////////////////////////////////
199 :
200 0 : #[derive(Debug)]
201 : pub struct Endpoint {
202 : /// used as the directory name
203 : endpoint_id: String,
204 : pub tenant_id: TenantId,
205 : pub timeline_id: TimelineId,
206 : pub mode: ComputeMode,
207 :
208 : // port and address of the Postgres server and `compute_ctl`'s HTTP API
209 : pub pg_address: SocketAddr,
210 : pub http_address: SocketAddr,
211 :
212 : // postgres major version in the format: 14, 15, etc.
213 : pg_version: u32,
214 :
215 : // These are not part of the endpoint as such, but the environment
216 : // the endpoint runs in.
217 : pub env: LocalEnv,
218 :
219 : // Optimizations
220 : skip_pg_catalog_updates: bool,
221 :
222 : // Feature flags
223 : features: Vec<ComputeFeature>,
224 : }
225 :
226 1543 : #[derive(PartialEq, Eq)]
227 : pub enum EndpointStatus {
228 : Running,
229 : Stopped,
230 : Crashed,
231 : RunningNoPidfile,
232 : }
233 :
234 : impl std::fmt::Display for EndpointStatus {
235 0 : fn fmt(&self, writer: &mut std::fmt::Formatter) -> std::fmt::Result {
236 0 : let s = match self {
237 0 : Self::Running => "running",
238 0 : Self::Stopped => "stopped",
239 0 : Self::Crashed => "crashed",
240 0 : Self::RunningNoPidfile => "running, no pidfile",
241 : };
242 0 : write!(writer, "{}", s)
243 0 : }
244 : }
245 :
246 : impl Endpoint {
247 3885 : fn from_dir_entry(entry: std::fs::DirEntry, env: &LocalEnv) -> Result<Endpoint> {
248 3885 : if !entry.file_type()?.is_dir() {
249 0 : anyhow::bail!(
250 0 : "Endpoint::from_dir_entry failed: '{}' is not a directory",
251 0 : entry.path().display()
252 0 : );
253 3885 : }
254 3885 :
255 3885 : // parse data directory name
256 3885 : let fname = entry.file_name();
257 3885 : let endpoint_id = fname.to_str().unwrap().to_string();
258 :
259 : // Read the endpoint.json file
260 3885 : let conf: EndpointConf =
261 3885 : serde_json::from_slice(&std::fs::read(entry.path().join("endpoint.json"))?)?;
262 :
263 3885 : Ok(Endpoint {
264 3885 : pg_address: SocketAddr::new("127.0.0.1".parse().unwrap(), conf.pg_port),
265 3885 : http_address: SocketAddr::new("127.0.0.1".parse().unwrap(), conf.http_port),
266 3885 : endpoint_id,
267 3885 : env: env.clone(),
268 3885 : timeline_id: conf.timeline_id,
269 3885 : mode: conf.mode,
270 3885 : tenant_id: conf.tenant_id,
271 3885 : pg_version: conf.pg_version,
272 3885 : skip_pg_catalog_updates: conf.skip_pg_catalog_updates,
273 3885 : features: conf.features,
274 3885 : })
275 3885 : }
276 :
277 518 : fn create_endpoint_dir(&self) -> Result<()> {
278 518 : std::fs::create_dir_all(self.endpoint_path()).with_context(|| {
279 0 : format!(
280 0 : "could not create endpoint directory {}",
281 0 : self.endpoint_path().display()
282 0 : )
283 518 : })
284 518 : }
285 :
286 : // Generate postgresql.conf with default configuration
287 518 : fn setup_pg_conf(&self) -> Result<PostgresConf> {
288 518 : let mut conf = PostgresConf::new();
289 518 : conf.append("max_wal_senders", "10");
290 518 : conf.append("wal_log_hints", "off");
291 518 : conf.append("max_replication_slots", "10");
292 518 : conf.append("hot_standby", "on");
293 518 : conf.append("shared_buffers", "1MB");
294 518 : conf.append("fsync", "off");
295 518 : conf.append("max_connections", "100");
296 518 : conf.append("wal_level", "logical");
297 518 : // wal_sender_timeout is the maximum time to wait for WAL replication.
298 518 : // It also defines how often the walreciever will send a feedback message to the wal sender.
299 518 : conf.append("wal_sender_timeout", "5s");
300 518 : conf.append("listen_addresses", &self.pg_address.ip().to_string());
301 518 : conf.append("port", &self.pg_address.port().to_string());
302 518 : conf.append("wal_keep_size", "0");
303 518 : // walproposer panics when basebackup is invalid, it is pointless to restart in this case.
304 518 : conf.append("restart_after_crash", "off");
305 518 :
306 518 : // Load the 'neon' extension
307 518 : conf.append("shared_preload_libraries", "neon");
308 518 :
309 518 : conf.append_line("");
310 518 : // Replication-related configurations, such as WAL sending
311 518 : match &self.mode {
312 : ComputeMode::Primary => {
313 : // Configure backpressure
314 : // - Replication write lag depends on how fast the walreceiver can process incoming WAL.
315 : // This lag determines latency of get_page_at_lsn. Speed of applying WAL is about 10MB/sec,
316 : // so to avoid expiration of 1 minute timeout, this lag should not be larger than 600MB.
317 : // Actually latency should be much smaller (better if < 1sec). But we assume that recently
318 : // updates pages are not requested from pageserver.
319 : // - Replication flush lag depends on speed of persisting data by checkpointer (creation of
320 : // delta/image layers) and advancing disk_consistent_lsn. Safekeepers are able to
321 : // remove/archive WAL only beyond disk_consistent_lsn. Too large a lag can cause long
322 : // recovery time (in case of pageserver crash) and disk space overflow at safekeepers.
323 : // - Replication apply lag depends on speed of uploading changes to S3 by uploader thread.
324 : // To be able to restore database in case of pageserver node crash, safekeeper should not
325 : // remove WAL beyond this point. Too large lag can cause space exhaustion in safekeepers
326 : // (if they are not able to upload WAL to S3).
327 467 : conf.append("max_replication_write_lag", "15MB");
328 467 : conf.append("max_replication_flush_lag", "10GB");
329 467 :
330 467 : if !self.env.safekeepers.is_empty() {
331 467 : // Configure Postgres to connect to the safekeepers
332 467 : conf.append("synchronous_standby_names", "walproposer");
333 467 :
334 467 : let safekeepers = self
335 467 : .env
336 467 : .safekeepers
337 467 : .iter()
338 608 : .map(|sk| format!("localhost:{}", sk.get_compute_port()))
339 467 : .collect::<Vec<String>>()
340 467 : .join(",");
341 467 : conf.append("neon.safekeepers", &safekeepers);
342 467 : } else {
343 0 : // We only use setup without safekeepers for tests,
344 0 : // and don't care about data durability on pageserver,
345 0 : // so set more relaxed synchronous_commit.
346 0 : conf.append("synchronous_commit", "remote_write");
347 0 :
348 0 : // Configure the node to stream WAL directly to the pageserver
349 0 : // This isn't really a supported configuration, but can be useful for
350 0 : // testing.
351 0 : conf.append("synchronous_standby_names", "pageserver");
352 0 : }
353 : }
354 49 : ComputeMode::Static(lsn) => {
355 49 : conf.append("recovery_target_lsn", &lsn.to_string());
356 49 : }
357 : ComputeMode::Replica => {
358 2 : assert!(!self.env.safekeepers.is_empty());
359 :
360 : // TODO: use future host field from safekeeper spec
361 : // Pass the list of safekeepers to the replica so that it can connect to any of them,
362 : // whichever is available.
363 2 : let sk_ports = self
364 2 : .env
365 2 : .safekeepers
366 2 : .iter()
367 2 : .map(|x| x.get_compute_port().to_string())
368 2 : .collect::<Vec<_>>()
369 2 : .join(",");
370 2 : let sk_hosts = vec!["localhost"; self.env.safekeepers.len()].join(",");
371 2 :
372 2 : let connstr = format!(
373 2 : "host={} port={} options='-c timeline_id={} tenant_id={}' application_name=replica replication=true",
374 2 : sk_hosts,
375 2 : sk_ports,
376 2 : &self.timeline_id.to_string(),
377 2 : &self.tenant_id.to_string(),
378 2 : );
379 2 :
380 2 : let slot_name = format!("repl_{}_", self.timeline_id);
381 2 : conf.append("primary_conninfo", connstr.as_str());
382 2 : conf.append("primary_slot_name", slot_name.as_str());
383 2 : conf.append("hot_standby", "on");
384 2 : // prefetching of blocks referenced in WAL doesn't make sense for us
385 2 : // Neon hot standby ignores pages that are not in the shared_buffers
386 2 : if self.pg_version >= 15 {
387 0 : conf.append("recovery_prefetch", "off");
388 2 : }
389 : }
390 : }
391 :
392 518 : Ok(conf)
393 518 : }
394 :
395 9480 : pub fn endpoint_path(&self) -> PathBuf {
396 9480 : self.env.endpoints_path().join(&self.endpoint_id)
397 9480 : }
398 :
399 3374 : pub fn pgdata(&self) -> PathBuf {
400 3374 : self.endpoint_path().join("pgdata")
401 3374 : }
402 :
403 1543 : pub fn status(&self) -> EndpointStatus {
404 1543 : let timeout = Duration::from_millis(300);
405 1543 : let has_pidfile = self.pgdata().join("postmaster.pid").exists();
406 1543 : let can_connect = TcpStream::connect_timeout(&self.pg_address, timeout).is_ok();
407 1543 :
408 1543 : match (has_pidfile, can_connect) {
409 35 : (true, true) => EndpointStatus::Running,
410 1508 : (false, false) => EndpointStatus::Stopped,
411 0 : (true, false) => EndpointStatus::Crashed,
412 0 : (false, true) => EndpointStatus::RunningNoPidfile,
413 : }
414 1543 : }
415 :
416 574 : fn pg_ctl(&self, args: &[&str], auth_token: &Option<String>) -> Result<()> {
417 574 : let pg_ctl_path = self.env.pg_bin_dir(self.pg_version)?.join("pg_ctl");
418 574 : let mut cmd = Command::new(&pg_ctl_path);
419 574 : cmd.args(
420 574 : [
421 574 : &[
422 574 : "-D",
423 574 : self.pgdata().to_str().unwrap(),
424 574 : "-w", //wait till pg_ctl actually does what was asked
425 574 : ],
426 574 : args,
427 574 : ]
428 574 : .concat(),
429 574 : )
430 574 : .env_clear()
431 574 : .env(
432 574 : "LD_LIBRARY_PATH",
433 574 : self.env.pg_lib_dir(self.pg_version)?.to_str().unwrap(),
434 574 : )
435 574 : .env(
436 574 : "DYLD_LIBRARY_PATH",
437 574 : self.env.pg_lib_dir(self.pg_version)?.to_str().unwrap(),
438 : );
439 :
440 : // Pass authentication token used for the connections to pageserver and safekeepers
441 574 : if let Some(token) = auth_token {
442 0 : cmd.env("NEON_AUTH_TOKEN", token);
443 574 : }
444 :
445 574 : let pg_ctl = cmd
446 574 : .output()
447 574 : .context(format!("{} failed", pg_ctl_path.display()))?;
448 574 : if !pg_ctl.status.success() {
449 2 : anyhow::bail!(
450 2 : "pg_ctl failed, exit code: {}, stdout: {}, stderr: {}",
451 2 : pg_ctl.status,
452 2 : String::from_utf8_lossy(&pg_ctl.stdout),
453 2 : String::from_utf8_lossy(&pg_ctl.stderr),
454 2 : );
455 572 : }
456 572 :
457 572 : Ok(())
458 574 : }
459 :
460 572 : fn wait_for_compute_ctl_to_exit(&self, send_sigterm: bool) -> Result<()> {
461 572 : // TODO use background_process::stop_process instead: https://github.com/neondatabase/neon/pull/6482
462 572 : let pidfile_path = self.endpoint_path().join("compute_ctl.pid");
463 572 : let pid: u32 = std::fs::read_to_string(pidfile_path)?.parse()?;
464 572 : let pid = nix::unistd::Pid::from_raw(pid as i32);
465 572 : if send_sigterm {
466 26 : kill(pid, Signal::SIGTERM).ok();
467 546 : }
468 572 : crate::background_process::wait_until_stopped("compute_ctl", pid)?;
469 572 : Ok(())
470 572 : }
471 :
472 813 : fn read_postgresql_conf(&self) -> Result<String> {
473 813 : // Slurp the endpoints/<endpoint id>/postgresql.conf file into
474 813 : // memory. We will include it in the spec file that we pass to
475 813 : // `compute_ctl`, and `compute_ctl` will write it to the postgresql.conf
476 813 : // in the data directory.
477 813 : let postgresql_conf_path = self.endpoint_path().join("postgresql.conf");
478 813 : match std::fs::read(&postgresql_conf_path) {
479 813 : Ok(content) => Ok(String::from_utf8(content)?),
480 0 : Err(e) if e.kind() == std::io::ErrorKind::NotFound => Ok("".to_string()),
481 0 : Err(e) => Err(anyhow::Error::new(e).context(format!(
482 0 : "failed to read config file in {}",
483 0 : postgresql_conf_path.to_str().unwrap()
484 0 : ))),
485 : }
486 813 : }
487 :
488 813 : fn build_pageserver_connstr(pageservers: &[(Host, u16)]) -> String {
489 813 : pageservers
490 813 : .iter()
491 912 : .map(|(host, port)| format!("postgresql://no_user@{host}:{port}"))
492 813 : .collect::<Vec<_>>()
493 813 : .join(",")
494 813 : }
495 :
496 582 : pub async fn start(
497 582 : &self,
498 582 : auth_token: &Option<String>,
499 582 : safekeepers: Vec<NodeId>,
500 582 : pageservers: Vec<(Host, u16)>,
501 582 : remote_ext_config: Option<&String>,
502 582 : shard_stripe_size: usize,
503 582 : ) -> Result<()> {
504 582 : if self.status() == EndpointStatus::Running {
505 0 : anyhow::bail!("The endpoint is already running");
506 582 : }
507 :
508 582 : let postgresql_conf = self.read_postgresql_conf()?;
509 :
510 : // We always start the compute node from scratch, so if the Postgres
511 : // data dir exists from a previous launch, remove it first.
512 582 : if self.pgdata().exists() {
513 67 : std::fs::remove_dir_all(self.pgdata())?;
514 515 : }
515 :
516 582 : let pageserver_connstring = Self::build_pageserver_connstr(&pageservers);
517 582 : assert!(!pageserver_connstring.is_empty());
518 :
519 582 : let mut safekeeper_connstrings = Vec::new();
520 582 : if self.mode == ComputeMode::Primary {
521 1205 : for sk_id in safekeepers {
522 674 : let sk = self
523 674 : .env
524 674 : .safekeepers
525 674 : .iter()
526 892 : .find(|node| node.id == sk_id)
527 674 : .ok_or_else(|| anyhow!("safekeeper {sk_id} does not exist"))?;
528 674 : safekeeper_connstrings.push(format!("127.0.0.1:{}", sk.get_compute_port()));
529 : }
530 51 : }
531 :
532 : // check for file remote_extensions_spec.json
533 : // if it is present, read it and pass to compute_ctl
534 582 : let remote_extensions_spec_path = self.endpoint_path().join("remote_extensions_spec.json");
535 582 : let remote_extensions_spec = std::fs::File::open(remote_extensions_spec_path);
536 : let remote_extensions: Option<RemoteExtSpec>;
537 :
538 582 : if let Ok(spec_file) = remote_extensions_spec {
539 1 : remote_extensions = serde_json::from_reader(spec_file).ok();
540 581 : } else {
541 581 : remote_extensions = None;
542 581 : };
543 :
544 : // Create spec file
545 582 : let spec = ComputeSpec {
546 582 : skip_pg_catalog_updates: self.skip_pg_catalog_updates,
547 582 : format_version: 1.0,
548 582 : operation_uuid: None,
549 582 : features: self.features.clone(),
550 582 : cluster: Cluster {
551 582 : cluster_id: None, // project ID: not used
552 582 : name: None, // project name: not used
553 582 : state: None,
554 582 : roles: vec![],
555 582 : databases: vec![],
556 582 : settings: None,
557 582 : postgresql_conf: Some(postgresql_conf),
558 582 : },
559 582 : delta_operations: None,
560 582 : tenant_id: Some(self.tenant_id),
561 582 : timeline_id: Some(self.timeline_id),
562 582 : mode: self.mode,
563 582 : pageserver_connstring: Some(pageserver_connstring),
564 582 : safekeeper_connstrings,
565 582 : storage_auth_token: auth_token.clone(),
566 582 : remote_extensions,
567 582 : pgbouncer_settings: None,
568 582 : shard_stripe_size: Some(shard_stripe_size),
569 582 : };
570 582 : let spec_path = self.endpoint_path().join("spec.json");
571 582 : std::fs::write(spec_path, serde_json::to_string_pretty(&spec)?)?;
572 :
573 : // Open log file. We'll redirect the stdout and stderr of `compute_ctl` to it.
574 582 : let logfile = std::fs::OpenOptions::new()
575 582 : .create(true)
576 582 : .append(true)
577 582 : .open(self.endpoint_path().join("compute.log"))?;
578 :
579 : // Launch compute_ctl
580 582 : println!("Starting postgres node at '{}'", self.connstr());
581 582 : let mut cmd = Command::new(self.env.neon_distrib_dir.join("compute_ctl"));
582 582 : cmd.args(["--http-port", &self.http_address.port().to_string()])
583 582 : .args(["--pgdata", self.pgdata().to_str().unwrap()])
584 582 : .args(["--connstr", &self.connstr()])
585 582 : .args([
586 582 : "--spec-path",
587 582 : self.endpoint_path().join("spec.json").to_str().unwrap(),
588 582 : ])
589 582 : .args([
590 582 : "--pgbin",
591 582 : self.env
592 582 : .pg_bin_dir(self.pg_version)?
593 582 : .join("postgres")
594 582 : .to_str()
595 582 : .unwrap(),
596 582 : ])
597 582 : .stdin(std::process::Stdio::null())
598 582 : .stderr(logfile.try_clone()?)
599 582 : .stdout(logfile);
600 :
601 582 : if let Some(remote_ext_config) = remote_ext_config {
602 1 : cmd.args(["--remote-ext-config", remote_ext_config]);
603 581 : }
604 :
605 582 : let child = cmd.spawn()?;
606 : // set up a scopeguard to kill & wait for the child in case we panic or bail below
607 582 : let child = scopeguard::guard(child, |mut child| {
608 8 : println!("SIGKILL & wait the started process");
609 8 : (|| {
610 8 : // TODO: use another signal that can be caught by the child so it can clean up any children it spawned
611 8 : child.kill().context("SIGKILL child")?;
612 8 : child.wait().context("wait() for child process")?;
613 8 : anyhow::Ok(())
614 8 : })()
615 8 : .with_context(|| format!("scopeguard kill&wait child {child:?}"))
616 8 : .unwrap();
617 582 : });
618 582 :
619 582 : // Write down the pid so we can wait for it when we want to stop
620 582 : // TODO use background_process::start_process instead: https://github.com/neondatabase/neon/pull/6482
621 582 : let pid = child.id();
622 582 : let pidfile_path = self.endpoint_path().join("compute_ctl.pid");
623 582 : std::fs::write(pidfile_path, pid.to_string())?;
624 :
625 : // Wait for it to start
626 582 : let mut attempt = 0;
627 : const ATTEMPT_INTERVAL: Duration = Duration::from_millis(100);
628 : const MAX_ATTEMPTS: u32 = 10 * 30; // Wait up to 30 s
629 : loop {
630 3533 : attempt += 1;
631 9431 : match self.get_status().await {
632 2949 : Ok(state) => {
633 2949 : match state.status {
634 : ComputeStatus::Init => {
635 2367 : if attempt == MAX_ATTEMPTS {
636 0 : bail!("compute startup timed out; still in Init state");
637 2367 : }
638 : // keep retrying
639 : }
640 : ComputeStatus::Running => {
641 : // All good!
642 574 : break;
643 : }
644 : ComputeStatus::Failed => {
645 8 : bail!(
646 8 : "compute startup failed: {}",
647 8 : state
648 8 : .error
649 8 : .as_deref()
650 8 : .unwrap_or("<no error from compute_ctl>")
651 8 : );
652 : }
653 : ComputeStatus::Empty
654 : | ComputeStatus::ConfigurationPending
655 : | ComputeStatus::Configuration => {
656 0 : bail!("unexpected compute status: {:?}", state.status)
657 : }
658 : }
659 : }
660 584 : Err(e) => {
661 584 : if attempt == MAX_ATTEMPTS {
662 0 : return Err(e).context("timed out waiting to connect to compute_ctl HTTP");
663 584 : }
664 : }
665 : }
666 2951 : std::thread::sleep(ATTEMPT_INTERVAL);
667 : }
668 :
669 : // disarm the scopeguard, let the child outlive this function (and neon_local invoction)
670 574 : drop(scopeguard::ScopeGuard::into_inner(child));
671 574 :
672 574 : Ok(())
673 582 : }
674 :
675 : // Call the /status HTTP API
676 3533 : pub async fn get_status(&self) -> Result<ComputeState> {
677 3533 : let client = reqwest::Client::new();
678 :
679 3533 : let response = client
680 3533 : .request(
681 3533 : reqwest::Method::GET,
682 3533 : format!(
683 3533 : "http://{}:{}/status",
684 3533 : self.http_address.ip(),
685 3533 : self.http_address.port()
686 3533 : ),
687 3533 : )
688 3533 : .send()
689 9431 : .await?;
690 :
691 : // Interpret the response
692 2949 : let status = response.status();
693 2949 : if !(status.is_client_error() || status.is_server_error()) {
694 2949 : Ok(response.json().await?)
695 : } else {
696 : // reqwest does not export its error construction utility functions, so let's craft the message ourselves
697 0 : let url = response.url().to_owned();
698 0 : let msg = match response.text().await {
699 0 : Ok(err_body) => format!("Error: {}", err_body),
700 0 : Err(_) => format!("Http error ({}) at {}.", status.as_u16(), url),
701 : };
702 0 : Err(anyhow::anyhow!(msg))
703 : }
704 3533 : }
705 :
706 231 : pub async fn reconfigure(&self, mut pageservers: Vec<(Host, u16)>) -> Result<()> {
707 231 : let mut spec: ComputeSpec = {
708 231 : let spec_path = self.endpoint_path().join("spec.json");
709 231 : let file = std::fs::File::open(spec_path)?;
710 231 : serde_json::from_reader(file)?
711 : };
712 :
713 231 : let postgresql_conf = self.read_postgresql_conf()?;
714 231 : spec.cluster.postgresql_conf = Some(postgresql_conf);
715 231 :
716 231 : // If we weren't given explicit pageservers, query the attachment service
717 231 : if pageservers.is_empty() {
718 0 : let attachment_service = AttachmentService::from_env(&self.env);
719 0 : let locate_result = attachment_service.tenant_locate(self.tenant_id).await?;
720 0 : pageservers = locate_result
721 0 : .shards
722 0 : .into_iter()
723 0 : .map(|shard| {
724 0 : (
725 0 : Host::parse(&shard.listen_pg_addr)
726 0 : .expect("Attachment service reported bad hostname"),
727 0 : shard.listen_pg_port,
728 0 : )
729 0 : })
730 0 : .collect::<Vec<_>>();
731 231 : }
732 :
733 231 : let pageserver_connstr = Self::build_pageserver_connstr(&pageservers);
734 231 : assert!(!pageserver_connstr.is_empty());
735 231 : spec.pageserver_connstring = Some(pageserver_connstr);
736 231 :
737 231 : let client = reqwest::Client::new();
738 231 : let response = client
739 231 : .post(format!(
740 231 : "http://{}:{}/configure",
741 231 : self.http_address.ip(),
742 231 : self.http_address.port()
743 231 : ))
744 231 : .body(format!(
745 231 : "{{\"spec\":{}}}",
746 231 : serde_json::to_string_pretty(&spec)?
747 : ))
748 231 : .send()
749 693 : .await?;
750 :
751 231 : let status = response.status();
752 231 : if !(status.is_client_error() || status.is_server_error()) {
753 231 : Ok(())
754 : } else {
755 0 : let url = response.url().to_owned();
756 0 : let msg = match response.text().await {
757 0 : Ok(err_body) => format!("Error: {}", err_body),
758 0 : Err(_) => format!("Http error ({}) at {}.", status.as_u16(), url),
759 : };
760 0 : Err(anyhow::anyhow!(msg))
761 : }
762 231 : }
763 :
764 574 : pub fn stop(&self, mode: &str, destroy: bool) -> Result<()> {
765 574 : self.pg_ctl(&["-m", mode, "stop"], &None)?;
766 :
767 : // Also wait for the compute_ctl process to die. It might have some
768 : // cleanup work to do after postgres stops, like syncing safekeepers,
769 : // etc.
770 : //
771 : // If destroying, send it SIGTERM before waiting. Sometimes we do *not*
772 : // want this cleanup: tests intentionally do stop when majority of
773 : // safekeepers is down, so sync-safekeepers would hang otherwise. This
774 : // could be a separate flag though.
775 572 : self.wait_for_compute_ctl_to_exit(destroy)?;
776 572 : if destroy {
777 26 : println!(
778 26 : "Destroying postgres data directory '{}'",
779 26 : self.pgdata().to_str().unwrap()
780 26 : );
781 26 : std::fs::remove_dir_all(self.endpoint_path())?;
782 546 : }
783 572 : Ok(())
784 574 : }
785 :
786 1164 : pub fn connstr(&self) -> String {
787 1164 : format!(
788 1164 : "postgresql://{}@{}:{}/{}",
789 1164 : "cloud_admin",
790 1164 : self.pg_address.ip(),
791 1164 : self.pg_address.port(),
792 1164 : "postgres"
793 1164 : )
794 1164 : }
795 : }
|
