rust/neon
1
0
Fork 0
mirror of https://github.com/neondatabase/neon.git synced 2026-01-14 08:52:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

Polish API handler and refresh OpenAPI spec

Browse Source
This commit is contained in:
Alexey Kondratov
2023-04-04 16:50:05 +02:00
parent 69552822b4
commit 4e6a23f779
6 changed files with 134 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
compute_tools/src/compute.rs
View File

@@ -88,7 +88,7 @@ pub struct ComputeState {
impl ComputeState {
pub fn new() -> Self {
Self {
status: ComputeStatus::WaitingSpec,
status: ComputeStatus::Empty,
last_active: Utc::now(),
error: None,
spec: ComputeSpec::default(),
@@ -111,8 +111,9 @@ impl Default for ComputeState {
pub enum ComputeStatus {
// Spec wasn't provided as start, waiting for it to be
// provided by control-plane.
WaitingSpec,
// Compute node has initial spec and is starting up.
Empty,
// Compute node has spec and initial startup and
// configuration is in progress.
Init,
// Compute is configured and running.
Running,
@@ -123,7 +124,7 @@ pub enum ComputeStatus {
// Control-plane requested reconfiguration.
ConfigurationPending,
// New spec is being applied.
Reconfiguration,
Configuration,
}
#[derive(Default, Serialize)]

11
compute_tools/src/configurator.rs
View File

@@ -15,22 +15,21 @@ fn configurator_main_loop(compute: &Arc<ComputeNode>) {
let mut state = state_changed.wait(state).unwrap();
if state.status == ComputeStatus::ConfigurationPending {
info!("got reconfiguration request");
state.status = ComputeStatus::Reconfiguration;
info!("got configuration request");
state.status = ComputeStatus::Configuration;
state_changed.notify_all();
drop(state);
let mut new_status = ComputeStatus::Failed;
if let Err(e) = compute.reconfigure() {
error!("could not reconfigure compute node: {}", e);
error!("could not configure compute node: {}", e);
} else {
new_status = ComputeStatus::Running;
info!("compute node reconfigured");
info!("compute node configured");
}
// XXX: used to test that API is blocking
// TODO: remove before merge
std::thread::sleep(std::time::Duration::from_millis(2000));
// std::thread::sleep(std::time::Duration::from_millis(2000));
compute.set_status(new_status);
} else if state.status == ComputeStatus::Failed {

36
compute_tools/src/http/api.rs
View File

@@ -4,7 +4,7 @@ use std::sync::Arc;
use std::thread;
use crate::compute::{ComputeNode, ComputeStatus};
use crate::spec::ComputeSpec;
use crate::http::models::{ConfigurationRequest, GenericAPIError};
use anyhow::Result;
use hyper::service::{make_service_fn, service_fn};
@@ -82,37 +82,35 @@ async fn routes(req: Request<Body>, compute: &Arc<ComputeNode>) -> Response<Body
))
}
// Accept spec in JSON format and request compute reconfiguration from
// Accept spec in JSON format and request compute configuration from
// the configurator thread. If anything goes wrong after we set the
// compute state to `ConfigurationPending` and / or sent spec to the
// configurator thread, we basically leave compute in the potentially
// wrong state. That said, it's control-plane's responsibility to
// watch compute state after reconfiguration request and to clean
// restart in case of errors.
//
// TODO: Errors should be in JSON format with proper status codes.
(&Method::POST, "/spec") => {
info!("serving /spec POST request");
(&Method::POST, "/configure") => {
info!("serving /configure POST request");
if !compute.live_config_allowed {
let msg = "live reconfiguration is not allowed for this compute node";
error!(msg);
return Response::new(Body::from(msg));
return render_json_error(msg, StatusCode::PRECONDITION_FAILED);
}
let body_bytes = hyper::body::to_bytes(req.into_body()).await.unwrap();
let spec_raw = String::from_utf8(body_bytes.to_vec()).unwrap();
if let Ok(spec) = serde_json::from_str::<ComputeSpec>(&spec_raw) {
if let Ok(request) = serde_json::from_str::<ConfigurationRequest>(&spec_raw) {
let spec = request.spec;
let (state, state_changed) = &compute.state;
let mut state = state.lock().unwrap();
if !(state.status == ComputeStatus::WaitingSpec
|| state.status == ComputeStatus::Running)
if !(state.status == ComputeStatus::Empty || state.status == ComputeStatus::Running)
{
let msg = format!(
"invalid compute status for reconfiguration request: {}",
serde_json::to_string(&*state).unwrap()
);
error!(msg);
return Response::new(Body::from(msg));
return render_json_error(&msg, StatusCode::PRECONDITION_FAILED);
}
state.spec = spec;
state.status = ComputeStatus::ConfigurationPending;
@@ -129,11 +127,13 @@ async fn routes(req: Request<Body>, compute: &Arc<ComputeNode>) -> Response<Body
state.status
);
}
Response::new(Body::from("ok"))
// Return current compute state if everything went well.
Response::new(Body::from(serde_json::to_string(&*state).unwrap()))
} else {
let msg = "invalid spec";
error!(msg);
Response::new(Body::from(msg))
render_json_error(msg, StatusCode::BAD_REQUEST)
}
}
@@ -146,6 +146,16 @@ async fn routes(req: Request<Body>, compute: &Arc<ComputeNode>) -> Response<Body
}
}
fn render_json_error(e: &str, status: StatusCode) -> Response<Body> {
let error = GenericAPIError {
error: e.to_string(),
};
Response::builder()
.status(status)
.body(Body::from(serde_json::to_string(&error).unwrap()))
.unwrap()
}
// Main Hyper HTTP server function that runs it and blocks waiting on it forever.
#[tokio::main]
async fn serve(state: Arc<ComputeNode>) {

1
compute_tools/src/http/mod.rs
View File

@@ -1 +1,2 @@
pub mod api;
pub mod models;

16
compute_tools/src/http/models.rs Normal file
View File

@@ -0,0 +1,16 @@
use serde::{Deserialize, Serialize};
use crate::spec::ComputeSpec;
/// We now pass only `spec` in the configuration request, but later we can
/// extend it and something like `restart: bool` or something else. So put
/// `spec` into a struct initially to be more flexible in the future.
#[derive(Deserialize, Debug)]
pub struct ConfigurationRequest {
pub spec: ComputeSpec,
}
#[derive(Serialize, Debug)]
pub struct GenericAPIError {
pub error: String,
}

98
compute_tools/src/http/openapi_spec.yaml
View File

@@ -11,7 +11,7 @@ paths:
get:
tags:
- Info
summary: Get compute node internal status
summary: Get compute node internal status.
description: ""
operationId: getComputeStatus
responses:
@@ -26,7 +26,7 @@ paths:
get:
tags:
- Info
summary: Get compute node startup metrics in JSON format
summary: Get compute node startup metrics in JSON format.
description: ""
operationId: getComputeMetricsJSON
responses:
@@ -41,9 +41,9 @@ paths:
get:
tags:
- Info
summary: Get current compute insights in JSON format
summary: Get current compute insights in JSON format.
description: |
Note, that this doesn't include any historical data
Note, that this doesn't include any historical data.
operationId: getComputeInsights
responses:
200:
@@ -56,12 +56,12 @@ paths:
/info:
get:
tags:
- "info"
summary: Get info about the compute Pod/VM
- Info
summary: Get info about the compute pod / VM.
description: ""
operationId: getInfo
responses:
"200":
200:
description: Info
content:
application/json:
@@ -72,7 +72,7 @@ paths:
post:
tags:
- Check
summary: Check that we can write new data on this compute
summary: Check that we can write new data on this compute.
description: ""
operationId: checkComputeWritability
responses:
@@ -82,9 +82,57 @@ paths:
text/plain:
schema:
type: string
description: Error text or 'true' if check passed
description: Error text or 'true' if check passed.
example: "true"
/configure:
post:
tags:
- Configure
summary: Request compute node configuration.
description: |
This is a blocking API endpoint, i.e. it blocks waiting until
compute is finished configuration and is in `Running` state.
Optional non-blocking mode could be added later. Currently,
it's also assumed that reconfiguration doesn't require restart.
operationId: configureCompute
requestBody:
description: Configuration request.
required: true
content:
application/json:
schema:
type: object
required:
- spec
properties:
spec:
# XXX: I don't want to explain current spec in the OpenAPI format,
# as it could be changed really soon. Consider doing it later.
type: object
responses:
200:
description: Compute configuration finished.
content:
application/json:
schema:
$ref: "#/components/schemas/ComputeState"
400:
description: Provided spec is invalid.
content:
application/json:
schema:
$ref: "#/components/schemas/GenericError"
412:
description: |
It's not possible to do live-configuration of the compute.
It's either in the wrong state, or compute doesn't use pull
mode of configuration.
content:
application/json:
schema:
$ref: "#/components/schemas/GenericError"
components:
securitySchemes:
JWT:
@@ -95,7 +143,7 @@ components:
schemas:
ComputeMetrics:
type: object
description: Compute startup metrics
description: Compute startup metrics.
required:
- sync_safekeepers_ms
- basebackup_ms
@@ -113,7 +161,7 @@ components:
Info:
type: object
description: Information about VM/Pod
description: Information about VM/Pod.
required:
- num_cpus
properties:
@@ -130,17 +178,26 @@ components:
$ref: '#/components/schemas/ComputeStatus'
last_active:
type: string
description: The last detected compute activity timestamp in UTC and RFC3339 format
description: The last detected compute activity timestamp in UTC and RFC3339 format.
example: "2022-10-12T07:20:50.52Z"
error:
type: string
description: Text of the error during compute startup, if any
description: Text of the error during compute startup, if any.
example: ""
tenant:
type: string
description: Identifier of the current tenant served by compute node, if any.
example: c9269c359e9a199fad1ea0981246a78f
timeline:
type: string
description: Identifier of the current timeline served by compute node, if any.
example: ece7de74d4b8cbe5433a68ce4d1b97b4
ComputeInsights:
type: object
properties:
pg_stat_statements:
description: Contains raw output from pg_stat_statements in JSON format
description: Contains raw output from pg_stat_statements in JSON format.
type: array
items:
type: object
@@ -151,6 +208,19 @@ components:
- init
- failed
- running
example: running
#
# Errors
#
GenericError:
type: object
required:
- error
properties:
error:
type: string
security:
- JWT: []