Line data Source code
1 : use std::fs::File;
2 : use std::path::Path;
3 :
4 : use anyhow::{Result, anyhow, bail};
5 : use compute_api::responses::{
6 : ComputeConfig, ControlPlaneComputeStatus, ControlPlaneConfigResponse,
7 : };
8 : use reqwest::StatusCode;
9 : use tokio_postgres::Client;
10 : use tracing::{error, info, instrument};
11 :
12 : use crate::config;
13 : use crate::metrics::{CPLANE_REQUESTS_TOTAL, CPlaneRequestRPC, UNKNOWN_HTTP_STATUS};
14 : use crate::migration::MigrationRunner;
15 : use crate::params::PG_HBA_ALL_MD5;
16 :
17 : // Do control plane request and return response if any. In case of error it
18 : // returns a bool flag indicating whether it makes sense to retry the request
19 : // and a string with error message.
20 0 : fn do_control_plane_request(
21 0 : uri: &str,
22 0 : jwt: &str,
23 0 : ) -> Result<ControlPlaneConfigResponse, (bool, String, String)> {
24 0 : let resp = reqwest::blocking::Client::new()
25 0 : .get(uri)
26 0 : .header("Authorization", format!("Bearer {}", jwt))
27 0 : .send()
28 0 : .map_err(|e| {
29 0 : (
30 0 : true,
31 0 : format!("could not perform request to control plane: {:?}", e),
32 0 : UNKNOWN_HTTP_STATUS.to_string(),
33 0 : )
34 0 : })?;
35 :
36 0 : let status = resp.status();
37 0 : match status {
38 0 : StatusCode::OK => match resp.json::<ControlPlaneConfigResponse>() {
39 0 : Ok(spec_resp) => Ok(spec_resp),
40 0 : Err(e) => Err((
41 0 : true,
42 0 : format!("could not deserialize control plane response: {:?}", e),
43 0 : status.to_string(),
44 0 : )),
45 : },
46 0 : StatusCode::SERVICE_UNAVAILABLE => Err((
47 0 : true,
48 0 : "control plane is temporarily unavailable".to_string(),
49 0 : status.to_string(),
50 0 : )),
51 : StatusCode::BAD_GATEWAY => {
52 : // We have a problem with intermittent 502 errors now
53 : // https://github.com/neondatabase/cloud/issues/2353
54 : // It's fine to retry GET request in this case.
55 0 : Err((
56 0 : true,
57 0 : "control plane request failed with 502".to_string(),
58 0 : status.to_string(),
59 0 : ))
60 : }
61 : // Another code, likely 500 or 404, means that compute is unknown to the control plane
62 : // or some internal failure happened. Doesn't make much sense to retry in this case.
63 0 : _ => Err((
64 0 : false,
65 0 : format!("unexpected control plane response status code: {}", status),
66 0 : status.to_string(),
67 0 : )),
68 : }
69 0 : }
70 :
71 : /// Request config from the control-plane by compute_id. If
72 : /// `NEON_CONTROL_PLANE_TOKEN` env variable is set, it will be used for
73 : /// authorization.
74 0 : pub fn get_config_from_control_plane(base_uri: &str, compute_id: &str) -> Result<ComputeConfig> {
75 0 : let cp_uri = format!("{base_uri}/compute/api/v2/computes/{compute_id}/spec");
76 0 : let jwt: String = std::env::var("NEON_CONTROL_PLANE_TOKEN").unwrap_or_default();
77 0 : let mut attempt = 1;
78 0 :
79 0 : info!("getting config from control plane: {}", cp_uri);
80 :
81 : // Do 3 attempts to get spec from the control plane using the following logic:
82 : // - network error -> then retry
83 : // - compute id is unknown or any other error -> bail out
84 : // - no spec for compute yet (Empty state) -> return Ok(None)
85 : // - got config -> return Ok(Some(config))
86 0 : while attempt < 4 {
87 0 : let result = match do_control_plane_request(&cp_uri, &jwt) {
88 0 : Ok(config_resp) => {
89 0 : CPLANE_REQUESTS_TOTAL
90 0 : .with_label_values(&[
91 0 : CPlaneRequestRPC::GetConfig.as_str(),
92 0 : &StatusCode::OK.to_string(),
93 0 : ])
94 0 : .inc();
95 0 : match config_resp.status {
96 0 : ControlPlaneComputeStatus::Empty => Ok(config_resp.into()),
97 : ControlPlaneComputeStatus::Attached => {
98 0 : if config_resp.spec.is_some() {
99 0 : Ok(config_resp.into())
100 : } else {
101 0 : bail!("compute is attached, but spec is empty")
102 : }
103 : }
104 : }
105 : }
106 0 : Err((retry, msg, status)) => {
107 0 : CPLANE_REQUESTS_TOTAL
108 0 : .with_label_values(&[CPlaneRequestRPC::GetConfig.as_str(), &status])
109 0 : .inc();
110 0 : if retry {
111 0 : Err(anyhow!(msg))
112 : } else {
113 0 : bail!(msg);
114 : }
115 : }
116 : };
117 :
118 0 : if let Err(e) = &result {
119 0 : error!("attempt {} to get config failed with: {}", attempt, e);
120 : } else {
121 0 : return result;
122 : }
123 :
124 0 : attempt += 1;
125 0 : std::thread::sleep(std::time::Duration::from_millis(100));
126 : }
127 :
128 : // All attempts failed, return error.
129 0 : Err(anyhow::anyhow!(
130 0 : "Exhausted all attempts to retrieve the config from the control plane"
131 0 : ))
132 0 : }
133 :
134 : /// Check `pg_hba.conf` and update if needed to allow external connections.
135 0 : pub fn update_pg_hba(pgdata_path: &Path) -> Result<()> {
136 0 : // XXX: consider making it a part of config.json
137 0 : let pghba_path = pgdata_path.join("pg_hba.conf");
138 0 :
139 0 : if config::line_in_file(&pghba_path, PG_HBA_ALL_MD5)? {
140 0 : info!("updated pg_hba.conf to allow external connections");
141 : } else {
142 0 : info!("pg_hba.conf is up-to-date");
143 : }
144 :
145 0 : Ok(())
146 0 : }
147 :
148 : /// Create a standby.signal file
149 0 : pub fn add_standby_signal(pgdata_path: &Path) -> Result<()> {
150 0 : // XXX: consider making it a part of config.json
151 0 : let signalfile = pgdata_path.join("standby.signal");
152 0 :
153 0 : if !signalfile.exists() {
154 0 : File::create(signalfile)?;
155 0 : info!("created standby.signal");
156 : } else {
157 0 : info!("reused pre-existing standby.signal");
158 : }
159 0 : Ok(())
160 0 : }
161 :
162 : #[instrument(skip_all)]
163 : pub async fn handle_neon_extension_upgrade(client: &mut Client) -> Result<()> {
164 : let query = "ALTER EXTENSION neon UPDATE";
165 : info!("update neon extension version with query: {}", query);
166 : client.simple_query(query).await?;
167 :
168 : Ok(())
169 : }
170 :
171 : #[instrument(skip_all)]
172 : pub async fn handle_migrations(client: &mut Client) -> Result<()> {
173 : info!("handle migrations");
174 :
175 : // !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
176 : // !BE SURE TO ONLY ADD MIGRATIONS TO THE END OF THIS ARRAY. IF YOU DO NOT, VERY VERY BAD THINGS MAY HAPPEN!
177 : // !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
178 :
179 : // Add new migrations in numerical order.
180 : let migrations = [
181 : include_str!("./migrations/0001-neon_superuser_bypass_rls.sql"),
182 : include_str!("./migrations/0002-alter_roles.sql"),
183 : include_str!("./migrations/0003-grant_pg_create_subscription_to_neon_superuser.sql"),
184 : include_str!("./migrations/0004-grant_pg_monitor_to_neon_superuser.sql"),
185 : include_str!("./migrations/0005-grant_all_on_tables_to_neon_superuser.sql"),
186 : include_str!("./migrations/0006-grant_all_on_sequences_to_neon_superuser.sql"),
187 : include_str!(
188 : "./migrations/0007-grant_all_on_tables_to_neon_superuser_with_grant_option.sql"
189 : ),
190 : include_str!(
191 : "./migrations/0008-grant_all_on_sequences_to_neon_superuser_with_grant_option.sql"
192 : ),
193 : include_str!("./migrations/0009-revoke_replication_for_previously_allowed_roles.sql"),
194 : include_str!(
195 : "./migrations/0010-grant_snapshot_synchronization_funcs_to_neon_superuser.sql"
196 : ),
197 : include_str!(
198 : "./migrations/0011-grant_pg_show_replication_origin_status_to_neon_superuser.sql"
199 : ),
200 : ];
201 :
202 : MigrationRunner::new(client, &migrations)
203 : .run_migrations()
204 : .await?;
205 :
206 : Ok(())
207 : }
|
