split TerminationPending into two values, keeping ComputeStatus stateless (#12506)

After https://github.com/neondatabase/neon/pull/12240 we observed
issues in our go code as `ComputeStatus` is not stateless, thus doesn't
deserialize as string.

```
could not check compute activity: json: cannot unmarshal object into Go struct field
ComputeState.status of type computeclient.ComputeStatus
```

- Fix this by splitting this status into two.
- Update compute OpenApi spec to reflect changes to `/terminate` in
previous PR
This commit is contained in:
Mikhail
2025-07-10 20:28:10 +01:00
committed by GitHub
parent c5aaf1ae21
commit 3593fe195a
7 changed files with 81 additions and 22 deletions

View File

@@ -46,11 +46,14 @@ stateDiagram-v2
Configuration --> Failed : Failed to configure the compute
Configuration --> Running : Compute has been configured
Empty --> Init : Compute spec is immediately available
Empty --> TerminationPending : Requested termination
Empty --> TerminationPendingFast : Requested termination
Empty --> TerminationPendingImmediate : Requested termination
Init --> Failed : Failed to start Postgres
Init --> Running : Started Postgres
Running --> TerminationPending : Requested termination
TerminationPending --> Terminated : Terminated compute
Running --> TerminationPendingFast : Requested termination
Running --> TerminationPendingImmediate : Requested termination
TerminationPendingFast --> Terminated compute with 30s delay for cplane to inspect status
TerminationPendingImmediate --> Terminated : Terminated compute immediately
Failed --> [*] : Compute exited
Terminated --> [*] : Compute exited
```

View File

@@ -956,14 +956,20 @@ impl ComputeNode {
None
};
let mut delay_exit = false;
let mut state = self.state.lock().unwrap();
state.terminate_flush_lsn = lsn;
if let ComputeStatus::TerminationPending { mode } = state.status {
let delay_exit = state.status == ComputeStatus::TerminationPendingFast;
if state.status == ComputeStatus::TerminationPendingFast
|| state.status == ComputeStatus::TerminationPendingImmediate
{
info!(
"Changing compute status from {} to {}",
state.status,
ComputeStatus::Terminated
);
state.status = ComputeStatus::Terminated;
self.state_changed.notify_all();
// we were asked to terminate gracefully, don't exit to avoid restart
delay_exit = mode == compute_api::responses::TerminateMode::Fast
}
drop(state);
@@ -1901,7 +1907,8 @@ impl ComputeNode {
// exit loop
ComputeStatus::Failed
| ComputeStatus::TerminationPending { .. }
| ComputeStatus::TerminationPendingFast
| ComputeStatus::TerminationPendingImmediate
| ComputeStatus::Terminated => break 'cert_update,
// wait

View File

@@ -371,9 +371,28 @@ paths:
summary: Terminate Postgres and wait for it to exit
description: ""
operationId: terminate
parameters:
- name: mode
in: query
description: "Terminate mode: fast (wait 30s before returning) and immediate"
required: false
schema:
type: string
enum: ["fast", "immediate"]
default: fast
responses:
200:
description: Result
content:
application/json:
schema:
$ref: "#/components/schemas/TerminateResponse"
201:
description: Result if compute is already terminated
content:
application/json:
schema:
$ref: "#/components/schemas/TerminateResponse"
412:
description: "wrong state"
content:
@@ -530,11 +549,14 @@ components:
type: string
enum:
- empty
- init
- failed
- running
- configuration_pending
- init
- running
- configuration
- failed
- termination_pending_fast
- termination_pending_immediate
- terminated
example: running
ExtensionInstallRequest:
@@ -660,6 +682,17 @@ components:
description: Role name.
example: "neon"
TerminateResponse:
type: object
required:
- lsn
properties:
lsn:
type: string
nullable: true
description: "last WAL flush LSN"
example: "0/028F10D8"
SetRoleGrantsResponse:
type: object
required:

View File

@@ -3,7 +3,7 @@ use crate::http::JsonResponse;
use axum::extract::State;
use axum::response::Response;
use axum_extra::extract::OptionalQuery;
use compute_api::responses::{ComputeStatus, TerminateResponse};
use compute_api::responses::{ComputeStatus, TerminateMode, TerminateResponse};
use http::StatusCode;
use serde::Deserialize;
use std::sync::Arc;
@@ -12,7 +12,7 @@ use tracing::info;
#[derive(Deserialize, Default)]
pub struct TerminateQuery {
mode: compute_api::responses::TerminateMode,
mode: TerminateMode,
}
/// Terminate the compute.
@@ -24,16 +24,16 @@ pub(in crate::http) async fn terminate(
{
let mut state = compute.state.lock().unwrap();
if state.status == ComputeStatus::Terminated {
return JsonResponse::success(StatusCode::CREATED, state.terminate_flush_lsn);
let response = TerminateResponse {
lsn: state.terminate_flush_lsn,
};
return JsonResponse::success(StatusCode::CREATED, response);
}
if !matches!(state.status, ComputeStatus::Empty | ComputeStatus::Running) {
return JsonResponse::invalid_status(state.status);
}
state.set_status(
ComputeStatus::TerminationPending { mode },
&compute.state_changed,
);
state.set_status(mode.into(), &compute.state_changed);
}
forward_termination_signal(false);

View File

@@ -84,7 +84,8 @@ impl ComputeMonitor {
if matches!(
compute_status,
ComputeStatus::Terminated
| ComputeStatus::TerminationPending { .. }
| ComputeStatus::TerminationPendingFast
| ComputeStatus::TerminationPendingImmediate
| ComputeStatus::Failed
) {
info!(

View File

@@ -922,7 +922,8 @@ impl Endpoint {
ComputeStatus::Empty
| ComputeStatus::ConfigurationPending
| ComputeStatus::Configuration
| ComputeStatus::TerminationPending { .. }
| ComputeStatus::TerminationPendingFast
| ComputeStatus::TerminationPendingImmediate
| ComputeStatus::Terminated => {
bail!("unexpected compute status: {:?}", state.status)
}

View File

@@ -121,6 +121,15 @@ pub enum TerminateMode {
Immediate,
}
impl From<TerminateMode> for ComputeStatus {
fn from(mode: TerminateMode) -> Self {
match mode {
TerminateMode::Fast => ComputeStatus::TerminationPendingFast,
TerminateMode::Immediate => ComputeStatus::TerminationPendingImmediate,
}
}
}
#[derive(Serialize, Clone, Copy, Debug, Deserialize, PartialEq, Eq)]
#[serde(rename_all = "snake_case")]
pub enum ComputeStatus {
@@ -141,7 +150,9 @@ pub enum ComputeStatus {
// control-plane to terminate it.
Failed,
// Termination requested
TerminationPending { mode: TerminateMode },
TerminationPendingFast,
// Termination requested, without waiting 30s before returning from /terminate
TerminationPendingImmediate,
// Terminated Postgres
Terminated,
}
@@ -160,7 +171,10 @@ impl Display for ComputeStatus {
ComputeStatus::Running => f.write_str("running"),
ComputeStatus::Configuration => f.write_str("configuration"),
ComputeStatus::Failed => f.write_str("failed"),
ComputeStatus::TerminationPending { .. } => f.write_str("termination-pending"),
ComputeStatus::TerminationPendingFast => f.write_str("termination-pending-fast"),
ComputeStatus::TerminationPendingImmediate => {
f.write_str("termination-pending-immediate")
}
ComputeStatus::Terminated => f.write_str("terminated"),
}
}