Read cplane events from regional redis (#7352)

## Problem

Actually read redis events.

## Summary of changes

This is revert of https://github.com/neondatabase/neon/pull/7350 +
fixes.
* Fixed events parsing
* Added timeout after connection failure
* Separated regional and global redis clients.
This commit is contained in:
Anna Khanova
2024-04-11 20:24:34 +02:00
committed by GitHub
parent 5299f917d6
commit 40f15c3123
16 changed files with 479 additions and 144 deletions

View File

@@ -8,9 +8,9 @@ use crate::{
backend::{ComputeCredentialKeys, ComputeUserInfo},
IpPattern,
},
cache::{project_info::ProjectInfoCacheImpl, Cached, TimedLru},
cache::{endpoints::EndpointsCache, project_info::ProjectInfoCacheImpl, Cached, TimedLru},
compute,
config::{CacheOptions, ProjectInfoCacheOptions},
config::{CacheOptions, EndpointCacheConfig, ProjectInfoCacheOptions},
context::RequestMonitoring,
intern::ProjectIdInt,
metrics::ApiLockMetrics,
@@ -417,12 +417,15 @@ pub struct ApiCaches {
pub node_info: NodeInfoCache,
/// Cache which stores project_id -> endpoint_ids mapping.
pub project_info: Arc<ProjectInfoCacheImpl>,
/// List of all valid endpoints.
pub endpoints_cache: Arc<EndpointsCache>,
}
impl ApiCaches {
pub fn new(
wake_compute_cache_config: CacheOptions,
project_info_cache_config: ProjectInfoCacheOptions,
endpoint_cache_config: EndpointCacheConfig,
) -> Self {
Self {
node_info: NodeInfoCache::new(
@@ -432,6 +435,7 @@ impl ApiCaches {
true,
),
project_info: Arc::new(ProjectInfoCacheImpl::new(project_info_cache_config)),
endpoints_cache: Arc::new(EndpointsCache::new(endpoint_cache_config)),
}
}
}
@@ -442,6 +446,7 @@ pub struct ApiLocks {
node_locks: DashMap<EndpointCacheKey, Arc<Semaphore>>,
permits: usize,
timeout: Duration,
epoch: std::time::Duration,
metrics: &'static ApiLockMetrics,
}
@@ -451,6 +456,7 @@ impl ApiLocks {
permits: usize,
shards: usize,
timeout: Duration,
epoch: std::time::Duration,
metrics: &'static ApiLockMetrics,
) -> prometheus::Result<Self> {
Ok(Self {
@@ -458,6 +464,7 @@ impl ApiLocks {
node_locks: DashMap::with_shard_amount(shards),
permits,
timeout,
epoch,
metrics,
})
}
@@ -495,12 +502,12 @@ impl ApiLocks {
})
}
pub async fn garbage_collect_worker(&self, epoch: std::time::Duration) {
pub async fn garbage_collect_worker(&self) {
if self.permits == 0 {
return;
}
let mut interval = tokio::time::interval(epoch / (self.node_locks.shards().len()) as u32);
let mut interval =
tokio::time::interval(self.epoch / (self.node_locks.shards().len()) as u32);
loop {
for (i, shard) in self.node_locks.shards().iter().enumerate() {
interval.tick().await;

View File

@@ -12,7 +12,7 @@ use crate::{
console::messages::ColdStartInfo,
http,
metrics::{CacheOutcome, Metrics},
scram,
scram, Normalize,
};
use crate::{cache::Cached, context::RequestMonitoring};
use futures::TryFutureExt;
@@ -24,7 +24,7 @@ use tracing::{error, info, info_span, warn, Instrument};
pub struct Api {
endpoint: http::Endpoint,
pub caches: &'static ApiCaches,
locks: &'static ApiLocks,
pub locks: &'static ApiLocks,
jwt: String,
}
@@ -56,6 +56,15 @@ impl Api {
ctx: &mut RequestMonitoring,
user_info: &ComputeUserInfo,
) -> Result<AuthInfo, GetAuthInfoError> {
if !self
.caches
.endpoints_cache
.is_valid(ctx, &user_info.endpoint.normalize())
.await
{
info!("endpoint is not valid, skipping the request");
return Ok(AuthInfo::default());
}
let request_id = ctx.session_id.to_string();
let application_name = ctx.console_application_name();
async {
@@ -82,7 +91,9 @@ impl Api {
Ok(body) => body,
// Error 404 is special: it's ok not to have a secret.
Err(e) => match e.http_status_code() {
Some(http::StatusCode::NOT_FOUND) => return Ok(AuthInfo::default()),
Some(http::StatusCode::NOT_FOUND) => {
return Ok(AuthInfo::default());
}
_otherwise => return Err(e.into()),
},
};
@@ -178,23 +189,27 @@ impl super::Api for Api {
ctx: &mut RequestMonitoring,
user_info: &ComputeUserInfo,
) -> Result<CachedRoleSecret, GetAuthInfoError> {
let ep = &user_info.endpoint;
let normalized_ep = &user_info.endpoint.normalize();
let user = &user_info.user;
if let Some(role_secret) = self.caches.project_info.get_role_secret(ep, user) {
if let Some(role_secret) = self
.caches
.project_info
.get_role_secret(normalized_ep, user)
{
return Ok(role_secret);
}
let auth_info = self.do_get_auth_info(ctx, user_info).await?;
if let Some(project_id) = auth_info.project_id {
let ep_int = ep.into();
let normalized_ep_int = normalized_ep.into();
self.caches.project_info.insert_role_secret(
project_id,
ep_int,
normalized_ep_int,
user.into(),
auth_info.secret.clone(),
);
self.caches.project_info.insert_allowed_ips(
project_id,
ep_int,
normalized_ep_int,
Arc::new(auth_info.allowed_ips),
);
ctx.set_project_id(project_id);
@@ -208,8 +223,8 @@ impl super::Api for Api {
ctx: &mut RequestMonitoring,
user_info: &ComputeUserInfo,
) -> Result<(CachedAllowedIps, Option<CachedRoleSecret>), GetAuthInfoError> {
let ep = &user_info.endpoint;
if let Some(allowed_ips) = self.caches.project_info.get_allowed_ips(ep) {
let normalized_ep = &user_info.endpoint.normalize();
if let Some(allowed_ips) = self.caches.project_info.get_allowed_ips(normalized_ep) {
Metrics::get()
.proxy
.allowed_ips_cache_misses
@@ -224,16 +239,18 @@ impl super::Api for Api {
let allowed_ips = Arc::new(auth_info.allowed_ips);
let user = &user_info.user;
if let Some(project_id) = auth_info.project_id {
let ep_int = ep.into();
let normalized_ep_int = normalized_ep.into();
self.caches.project_info.insert_role_secret(
project_id,
ep_int,
normalized_ep_int,
user.into(),
auth_info.secret.clone(),
);
self.caches
.project_info
.insert_allowed_ips(project_id, ep_int, allowed_ips.clone());
self.caches.project_info.insert_allowed_ips(
project_id,
normalized_ep_int,
allowed_ips.clone(),
);
ctx.set_project_id(project_id);
}
Ok((